Package 'prolfqua'

Title:	Proteomics Label Free Quantification
Description:	The prolfqua package integrates essential steps of the mass spectrometry based label free quantification data analysis workflow: quality control, data normalization, protein aggregation, modeling, hypothesis testing and sample size estimation. We use prolfqua to : develop highly customizable, visually appealing, and interactive data analysis reports in pdf or HTML format; to visualize and model simple experimental designs with a single explanatory variable and complex experiments with multiple factors; to benchmark data acquisition, data preprocessing or data modeling methods. We developed and improved the package by applying the "Eating your own dog food" principle, when analysing dozens of proteomics datasets.
Authors:	Witold Wolski [aut] (ORCID: <https://orcid.org/0000-0002-6468-120X>), Functional Genomics Center Zurich [fnd], Christian Panse [cre] (ORCID: <https://orcid.org/0000-0003-1975-3064>)
Maintainer:	Christian Panse <[email protected]>
License:	MIT + file LICENSE
Version:	0.99.2
Built:	2026-10-09 16:31:17 UTC
Source:	https://github.com/BiocStaging/prolfqua

Help Index

adjust columns
AggregateLimpa
AggregateMedpolish
AggregateRlm
AggregateTopN
Analysis Configuration
DRY function: process and export annotated contrasts
R6 class for extracting ANOVA results as a data frame
Build a contrast analysis using one of several statistical methods
Build protein models from data
Build Firth logistic model for nested LFQData
Build Firth logistic model for aggregated LFQData
Build protein models with LOD imputation for failed fits
Build limma model from LFQData
Build limma model with LOD imputation for failed proteins
Build limma model with vooma precision weights (proteomics)
Build limma-voom model with LOD imputation for failed proteins
Build limpa vooma model from aggregated LFQData
center to reference
Contrast Configuration
Estimate contrasts using Wald Test
Quasibinomial detection-count facade for nested input
DEqMS contrast analysis facade
DEqMS contrast analysis with vooma weights facade
Estimate contrasts using Wald Test
Firth logistic missingness contrast analysis facade
Firth logistic missingness contrast analysis facade for nested input
Base class for all Contrasts classes
Limma-based contrasts (direct limma pipeline)
Limma contrast analysis facade
Limma contrast analysis with LOD imputation facade
Limma-voom contrast analysis facade
Limma-voom contrast analysis with LOD imputation facade
Limpa contrast analysis facade
Limpa contrast analysis facade for nested input
Lmer (mixed-model) contrast analysis facade for nested input
LM contrast analysis facade
LM contrast analysis with LOD imputation facade
Limma moderated contrasts
DEqMS count-dependent moderated contrasts
plot contrasts
Rfit rank-based regression contrast analysis facade
Rfit rank-regression contrast analysis with LOD imputation facade
RLM contrast analysis facade
ROPECA reproducibility-optimization method
ROPECA contrast analysis facade for nested input
holds results when contrasts are added.
Residual degrees of freedom for an augmented rfit fit
Registry of available contrast facade classes
find file stored in package
Combined generate_contrasts
Single-factor contrasts (pairwise comparisons)
Extracts uniprot ID
group label function
Count distinct elements for each level of hierarchy per sample
Fill missing responses with predictions from per-subject linear models
Interaction contrasts (difference of differences)
Internal Functions by category
Detect contaminant identifiers
Detect decoy / reverse-database identifiers
Level-specific contrasts (per secondary level)
LFQData R6 class
LFQDataPlotter —- Create various visualization of the LFQdata
Decorates LFQData with methods to compute statistics of interactions
Summarize LFQData
converts LFQData object to SummarizedExperiment
Decorate LFQData with Methods for transforming Intensities
List currently registered contrast facades
read minimal yaml to reconstruct configuration
Rebuild a ContrastConfiguration from a list
Look up a contrast facade by short name
Likelihood ratio test
main effects contrasts
matrix or data.frame to tibble (taken from tidyquant)
Merge contrast results coming from two different model.
Group statistics and limit of detection
R6 class representing modelling result
R6 class representing modelling result
R6 interface class representing modelling result
R6 class representing a limma modelling result
Max nr_children across samples per hierarchy unit
Count distinct child features per hierarchy unit
Aggregates e.g. protein abundances from peptide abundances
Package Data
pivot model contrasts matrix to wide format produced by 'contrasts_linfct' and ...
load data from prolfqua
Extract all value slots in an R6 object
Register a contrast facade class
Setup a tidy table compatible with a AnalysisConfiguration
Scale estimate for an augmented rfit fit
simulate protein level data with two groups
Simulate data, protein, with config with 2 factors Treatment and Background
Simulate data, protein and peptide, with config
Simulate data, protein, with config
Create a quasibinomial detection-count strategy
Create limma modelling strategy
Create limpa modelling strategy
Create Firth's logistic regression strategy
Quasibinomial detection-count strategy
R6 class for limma modelling strategy
R6 class for limpa modelling strategy
Linear model strategy (R6 class)
Linear mixed-effects model strategy (R6 class)
Firth's logistic regression strategy (R6 class)
Rank-based regression strategy (R6 class)
Robust linear model strategy (R6 class)
compute var sd etc for all factor levels
table facade to easily switch implementations
UpSetR plot from interaction_missing_stats
Named variance-covariance matrix for an augmented rfit fit
volcano plotly
add missing values to x vector based on the values of x

adjust columns

Description

adjust columns

Usage

adjust_p_values(
  contrast_df,
  column = "p.value",
  group_by_col = "contrast",
  newname = "FDR"
)
adjust_p_values(
  contrast_df,
  column = "p.value",
  group_by_col = "contrast",
  newname = "FDR"
)

Arguments

contrast_df

data.frame with p-values to adjust

column

name of column containing p-values

group_by_col

column(s) to group by before adjusting (e.g. contrast), or NULL for no grouping

newname

name of the new column with adjusted p-values

Value

The computed result.

Examples


bb <- c(runif(1000), rexp(1500,rate=5))
length(bb)
bb <- bb[bb < 1]
length(bb)
bb <- bb[1:2000]
hist(bb)
data <- data.frame(contrast = rep(LETTERS[1:5],400), p.value = bb)

dataX <- adjust_p_values(data)
Adata <- dataX |> dplyr::filter(contrast == "A")
stopifnot(all.equal(Adata$FDR, p.adjust(Adata$p.value, method="BH")))
data2 <- adjust_p_values(data, group_by_col = NULL)
stopifnot(all.equal(data2$FDR, p.adjust(data2$p.value, method="BH")))


bb <- c(runif(1000), rexp(1500,rate=5))
length(bb)
bb <- bb[bb < 1]
length(bb)
bb <- bb[1:2000]
hist(bb)
data <- data.frame(contrast = rep(LETTERS[1:5],400), p.value = bb)

dataX <- adjust_p_values(data)
Adata <- dataX |> dplyr::filter(contrast == "A")
stopifnot(all.equal(Adata$FDR, p.adjust(Adata$p.value, method="BH")))
data2 <- adjust_p_values(data, group_by_col = NULL)
stopifnot(all.equal(data2$FDR, p.adjust(data2$p.value, method="BH")))

Aggregates peptide/precursor intensities to protein level using limpa's probabilistic DPC-based quantification (limpa::dpcQuant). With impute_only = TRUE, performs same-level imputation without aggregation using limpa::dpcQuantByRow.

The output LFQData contains three value columns:

intensity ("limpa") — protein-level log-expression (complete, no NAs)
standard error (in config$opt_se) — per-protein, per-sample posterior SEs
observation count (in config$nr_children) — number of observed precursors

Value

An R6 class generator.

Super class

prolfqua::AggregatorBase -> AggregateLimpa

Public fields

dpc_result: estimated DPC object from limpa::dpc
dpc_slope: DPC slope parameter (default 0.8)
impute_only: if TRUE use dpcQuantByRow (no aggregation)

Methods

Method `new()`

initialize

Usage

AggregateLimpa$new(
  lfq,
  prefix = "protein",
  dpc_slope = 0.8,
  impute_only = FALSE
)

Arguments

lfq: LFQData with log2-transformed intensities
prefix: default "protein"
dpc_slope: DPC slope parameter passed to limpa (default 0.8)
impute_only: if TRUE, use dpcQuantByRow instead of dpcQuant

Method `aggregate()`

run limpa DPC-based aggregation (or imputation)

Usage

AggregateLimpa$aggregate()

Returns

LFQData

Method `plot()`

creates aggregation plots (only for aggregation mode, not impute_only)

Usage

AggregateLimpa$plot(subset = NULL, show.legend = FALSE)

Arguments

subset: create plots for a subset of the data only
show.legend: default FALSE

Returns

data.frame

Method `write_plots()`

writes plots to folder

Usage

AggregateLimpa$write_plots(
  qcpath,
  subset = NULL,
  show.legend = FALSE,
  width = 6,
  height = 6
)

Arguments

qcpath: qcpath
subset: write plots only for some
show.legend: legend
width: figure width
height: figure height

Returns

file path

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AggregateLimpa$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq

agg <- AggregateLimpa$new(lfqdata, "protein")
agg$aggregate()
agg$lfq_agg$data_wide()
istar <- prolfqua::sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq

agg <- AggregateLimpa$new(lfqdata, "protein")
agg$aggregate()
agg$lfq_agg$data_wide()

AggregateMedpolish

Description

AggregateMedpolish

Details

Aggregates peptide intensities to protein level using median polish. Works best with variance-stabilized (log-transformed) intensities.

Value

An R6 class generator.

Super class

prolfqua::AggregatorBase -> AggregateMedpolish

Methods

Public methods

AggregateMedpolish$aggregate()
AggregateMedpolish$clone()

Inherited methods

prolfqua::AggregatorBase$initialize()
prolfqua::AggregatorBase$plot()
prolfqua::AggregatorBase$write_plots()

Method `aggregate()`

run median polish aggregation

Usage

AggregateMedpolish$aggregate()

Returns

LFQData

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AggregateMedpolish$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)
lfqTrans <- lfqdata$clone()$get_Transformer()$log2()$robscale()$lfq

agg <- AggregateMedpolish$new(lfqTrans, "protein")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)
lfqTrans <- lfqdata$clone()$get_Transformer()$log2()$robscale()$lfq

agg <- AggregateMedpolish$new(lfqTrans, "protein")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

AggregateRlm

Description

AggregateRlm

Details

Aggregates peptide intensities to protein level using robust regression (rlm). Works best with variance-stabilized (log-transformed) intensities.

Value

An R6 class generator.

Super class

prolfqua::AggregatorBase -> AggregateRlm

Methods

Public methods

AggregateRlm$aggregate()
AggregateRlm$clone()

Inherited methods

prolfqua::AggregatorBase$initialize()
prolfqua::AggregatorBase$plot()
prolfqua::AggregatorBase$write_plots()

Method `aggregate()`

run robust regression aggregation

Usage

AggregateRlm$aggregate()

Returns

LFQData

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AggregateRlm$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)
lfqTrans <- lfqdata$clone()$get_Transformer()$log2()$robscale()$lfq

agg <- AggregateRlm$new(lfqTrans, "protein")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)
lfqTrans <- lfqdata$clone()$get_Transformer()$log2()$robscale()$lfq

agg <- AggregateRlm$new(lfqTrans, "protein")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

AggregateTopN

Description

AggregateTopN

Details

Aggregates peptide intensities to protein level using top N peptides. Works with raw (untransformed) intensities.

Value

An R6 class generator.

Super class

prolfqua::AggregatorBase -> AggregateTopN

Public fields

N: top N peptides by intensity
func: aggregation function name: "sum" or "mean"

Methods

Public methods

AggregateTopN$new()
AggregateTopN$aggregate()
AggregateTopN$clone()

Inherited methods

prolfqua::AggregatorBase$plot()
prolfqua::AggregatorBase$write_plots()

Method `new()`

initialize

Usage

AggregateTopN$new(lfq, prefix = "protein", N = 3, func = "sum")

Arguments

lfq: LFQData
prefix: default protein
N: top N peptides (default 3)
func: "sum" or "mean" (default "sum")

Method `aggregate()`

run top N aggregation

Usage

AggregateTopN$aggregate()

Returns

LFQData

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AggregateTopN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)

agg <- AggregateTopN$new(lfqdata, "protein", N = 3, func = "sum")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

agg_mean <- AggregateTopN$new(lfqdata, "protein", N = 3, func = "mean")
agg_mean$aggregate()
protPlotter <- agg_mean$lfq_agg$get_Plotter()
protPlotter$heatmap()
agg_mean$write_plots(tempdir())

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data |> dplyr::filter(protein_Id %in% sample(protein_Id, 100))
lfqdata <- LFQData$new(data, istar$config)

agg <- AggregateTopN$new(lfqdata, "protein", N = 3, func = "sum")
agg$aggregate()
p <- agg$plot()
p$plots[[1]]

agg_mean <- AggregateTopN$new(lfqdata, "protein", N = 3, func = "mean")
agg_mean$aggregate()
protPlotter <- agg_mean$lfq_agg$get_Plotter()
protPlotter$heatmap()
agg_mean$write_plots(tempdir())

Analysis Configuration

Description

Analysis Configuration — holds all table annotations, hierarchy definitions, factor definitions, and analysis parameters in a single flat object.

Value

An R6 class generator.

Public fields

sep: separator to use when uniting columns is necessary
file_name: column name of column containing raw file names
sample_name: (will be generated from factors or file_name)
norm_value: optional column with normalization values (e.g., Creatinine)
isotope_label: which column contains the isotope label (e.g. heavy or light), or light only if LFQ.
ident_q_value: column name with identification QValues (smaller better)
ident_score: column with identification score (larger better)
opt_rt: optional column with rt information
opt_mz: optional column with mz information
opt_se: optional column with standard errors (e.g. from limpa aggregation)
nr_children: optional column containing for instance the number of peptides
work_intensity: column which contains the intensities
is_response_transformed: are the intensities transformed for constant variance
bin_resp: column with encoded missing information
pattern_decoys: optional regex identifying decoy / reverse-database ids, unioned with the built-in default prefixes (see is_decoy). NULL (default) leaves decoy handling off; set it (even to "" for defaults only) to opt into flagging and the targets-only model fit.
pattern_contaminants: optional regex identifying contaminant ids (keratin, trypsin, BSA, ...). NULL (default) means contaminants are not flagged; set it to enable contaminant flagging / removal.
factors: Names of columns containing factors (annotations)
factor_depth: number of relevant factors (used by plotting functions etc)
hierarchy: list with columns describing the measurement hierarchy (i.e. protein peptide precursor fragment)
hierarchy_depth: At which depth do you want to model i.e. protein than 1
min_peptides_protein: minimum number of peptides per protein

Methods

Public methods

AnalysisConfiguration$set_response()
AnalysisConfiguration$get_response()
AnalysisConfiguration$pop_response()
AnalysisConfiguration$factor_keys()
AnalysisConfiguration$factor_keys_depth()
AnalysisConfiguration$hierarchy_keys()
AnalysisConfiguration$hierarchy_keys_depth()
AnalysisConfiguration$id_vars()
AnalysisConfiguration$value_vars()
AnalysisConfiguration$annotation_vars()
AnalysisConfiguration$clone()

Method `set_response()`

Add name of intensity column

Usage

AnalysisConfiguration$set_response(col_name)

Arguments

col_name: name of intensity column

Method `get_response()`

Get name of working intensity column

Usage

AnalysisConfiguration$get_response()

Method `pop_response()`

Remove last name in array of working intensity column names

Usage

AnalysisConfiguration$pop_response()

Method `factor_keys()`

Get factor keys

Usage

AnalysisConfiguration$factor_keys()

Returns

array with keys

Method `factor_keys_depth()`

Get factor keys till factorDepth

Usage

AnalysisConfiguration$factor_keys_depth()

Method `hierarchy_keys()`

get hierarchy keys

Usage

AnalysisConfiguration$hierarchy_keys()

Returns

array of column names

Method `hierarchy_keys_depth()`

get hierarchy keys up to depth

Usage

AnalysisConfiguration$hierarchy_keys_depth(names = TRUE)

Arguments

names: if TRUE names only if FALSE key value pairs

Returns

array of column names

Method `id_vars()`

get names of columns annotating values (e.g. intensities)

Usage

AnalysisConfiguration$id_vars()

Returns

character array

Method `value_vars()`

get names of columns containing observations e.g. (intensity, qValue, mz or rt)

Usage

AnalysisConfiguration$value_vars()

Method `annotation_vars()`

get names of columns with sample annotations

Usage

AnalysisConfiguration$annotation_vars()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AnalysisConfiguration$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_peptide_config()
config <- istar$config
stopifnot("AnalysisConfiguration" %in% class(config))
stopifnot(length(config$hierarchy_keys()) > 0)
stopifnot(length(config$factor_keys()) > 0)
stopifnot(length(config$get_response()) == 1)

istar <- sim_lfq_data_peptide_config()
config <- istar$config
stopifnot("AnalysisConfiguration" %in% class(config))
stopifnot(length(config$hierarchy_keys()) > 0)
stopifnot(length(config$factor_keys()) > 0)
stopifnot(length(config$get_response()) == 1)

DRY function: process and export annotated contrasts

Description

DRY function: process and export annotated contrasts

Usage

annotation_add_contrasts(
  df,
  primary_col,
  secondary_col,
  prefix = "G_",
  dataset_id = "dataset",
  decreasing = FALSE,
  interactions = TRUE
)
annotation_add_contrasts(
  df,
  primary_col,
  secondary_col,
  prefix = "G_",
  dataset_id = "dataset",
  decreasing = FALSE,
  interactions = TRUE
)

Arguments

df

data.frame with annotation data

primary_col

character, name of the primary factor column

secondary_col

character, name of the secondary factor column

prefix

character, prefix for output naming (default "G_")

dataset_id

character, dataset identifier for output naming

decreasing

logical, if TRUE sort factor levels in decreasing order

interactions

logical, if TRUE include interaction contrasts

Value

Contrast definitions or contrast results.

Examples


data("x5463yzwer453bbb", package = "prolfqua")
annotation_add_contrasts(x5463yzwer453bbb, "factor_A", "factor_B", "primary")
annotation_add_contrasts(x5463yzwer453bbb, "factor_B", "factor_A", "secondary")
data("x5463yzwer453bbb", package = "prolfqua")
annotation_add_contrasts(x5463yzwer453bbb, "factor_A", "factor_B", "primary")
annotation_add_contrasts(x5463yzwer453bbb, "factor_B", "factor_A", "secondary")

R6 class for extracting ANOVA results as a data frame

Description

R6 class for extracting ANOVA results as a data frame

Details

Wraps the ANOVA extraction logic and associated column names.

Value

An R6 class generator.

Public fields

test: statistical test type ("F", "Chisq", etc.)
col_pval: p-value column name in ANOVA output

Methods

Public methods

AnovaExtractor$new()
AnovaExtractor$fun()
AnovaExtractor$clone()

Method `new()`

Create a new AnovaExtractor

Usage

AnovaExtractor$new(test = "F")

Arguments

test: statistical test type

Method `fun()`

Extract ANOVA table from a fitted model

Usage

AnovaExtractor$fun(x)

Arguments

x: a fitted model

Returns

data.frame with factor column and ANOVA statistics

Method `clone()`

The objects of this class are cloneable with this method.

Usage

AnovaExtractor$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

extractor <- AnovaExtractor$new(test = "F")
fit <- stats::lm(Sepal.Length ~ Species, data = iris)
res <- extractor$fun(fit)
stopifnot("factor" %in% colnames(res))
extractor <- AnovaExtractor$new(test = "F")
fit <- stats::lm(Sepal.Length ~ Species, data = iris)
res <- extractor$fun(fit)
stopifnot("factor" %in% colnames(res))

Build a contrast analysis using one of several statistical methods

Description

A builder function that dispatches to the appropriate facade class based on the chosen method. Each facade encapsulates the full pipeline from strategy construction through modelling to contrast computation.

Usage

build_contrast_analysis(lfqdata, modelstr, contrasts, method = "lm", ...)
build_contrast_analysis(lfqdata, modelstr, contrasts, method = "lm", ...)

Arguments

lfqdata

an LFQData object

modelstr

model formula string without the response variable (e.g. "~ group_"). The response is taken automatically from lfqdata$get_config()$get_response().

contrasts

named character vector of contrasts (e.g. c("A_vs_B" = "group_A - group_B"))

method

a registered facade key. The built-in keys are "lm", "lm_impute", "limma", "limma_impute", "limma_voom", "limma_voom_impute", "limpa", "limpa_nested", "rlm", "rfit", "rfit_impute", "deqms", "deqms_voom", "firth", "firth_nested", "binomial_nested", "lmer_nested", "ropeca_nested"; downstream packages may add more via register_facade. The authoritative list is names(list_facades()). Defaults to "lm".

...

additional arguments forwarded to the underlying strategy function (e.g. trend, robust for strategy_limma)

Value

one of ContrastsLimmaFacade, ContrastsLMFacade, ContrastsRLMFacade, ContrastsRfitFacade, ContrastsRfitImputeFacade, ContrastsLmerNestedFacade, ContrastsLMImputeFacade, ContrastsDEqMSFacade, ContrastsROPECANestedFacade, ContrastsFirthFacade, ContrastsFirthNestedFacade, ContrastsLimpaFacade, ContrastsBinomialNestedFacade, or ContrastsLimpaNestedFacade

Examples

istar <- sim_lfq_data_protein_config(Nprot = 20)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")

fa_lm    <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "lm")
head(fa_lm$get_contrasts())

fa_limma <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "limma")
head(fa_limma$get_contrasts())

fa_deqms <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "deqms")
head(fa_deqms$get_contrasts())

istar_pep <- sim_lfq_data_peptide_config()
lfqdata_pep <- LFQData$new(istar_pep$data, istar_pep$config)
lfqdata_pep <- lfqdata_pep$get_Transformer()$log2()$lfq

fa_lmer <- build_contrast_analysis(
  lfqdata_pep,
  "~ group_ + (1 | peptide_Id) + (1 | sampleName)",
  contrasts,
  method = "lmer_nested"
)
head(fa_lmer$get_contrasts())

fa_ropeca <- build_contrast_analysis(lfqdata_pep, "~ group_", contrasts, method = "ropeca_nested")
head(fa_ropeca$get_contrasts())

fa_firth <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "firth")
head(fa_firth$get_contrasts())
istar <- sim_lfq_data_protein_config(Nprot = 20)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")

fa_lm    <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "lm")
head(fa_lm$get_contrasts())

fa_limma <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "limma")
head(fa_limma$get_contrasts())

fa_deqms <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "deqms")
head(fa_deqms$get_contrasts())

istar_pep <- sim_lfq_data_peptide_config()
lfqdata_pep <- LFQData$new(istar_pep$data, istar_pep$config)
lfqdata_pep <- lfqdata_pep$get_Transformer()$log2()$lfq

fa_lmer <- build_contrast_analysis(
  lfqdata_pep,
  "~ group_ + (1 | peptide_Id) + (1 | sampleName)",
  contrasts,
  method = "lmer_nested"
)
head(fa_lmer$get_contrasts())

fa_ropeca <- build_contrast_analysis(lfqdata_pep, "~ group_", contrasts, method = "ropeca_nested")
head(fa_ropeca$get_contrasts())

fa_firth <- build_contrast_analysis(lfqdata, "~ group_", contrasts, method = "firth")
head(fa_firth$get_contrasts())

Build protein models from data

Description

Build protein models from data

Usage

build_model(
  data,
  model_strategy,
  subject_id = if ("LFQData" %in% class(data)) {
     data$subject_id()
 } else {
    
    "protein_Id"
 },
  model_name = model_strategy$model_name
)
build_model(
  data,
  model_strategy,
  subject_id = if ("LFQData" %in% class(data)) {
     data$subject_id()
 } else {
    
    "protein_Id"
 },
  model_name = model_strategy$model_name
)

Arguments

data

data - a data frame or LFQData object

model_strategy

model strategy object (e.g. from strategy_lmer or strategy_lm)

subject_id

grouping variable

model_name

model name

Value

a object of class Model

Examples

D <- prolfqua::sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.1)
D$data$abundance |> is.na() |> sum()
D <- prolfqua::sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.1, seed =3)
D$data$abundance |> is.na() |> sum()
model_name <- "f_condtion_r_peptide"
formula_randomPeptide <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1 | sampleName)",
   model_name = model_name)


mod <- prolfqua::build_model(
 D$data,
 formula_randomPeptide,
 model_name = model_name,
 subject_id = D$config$hierarchy_keys_depth())
aovtable <- mod$get_anova()

mod <- prolfqua::build_model(
 LFQData$new(D$data, D$config),
 formula_randomPeptide,
 model_name = model_name)
model_summary(mod)


D <- prolfqua::sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.1)
D$data$abundance |> is.na() |> sum()
D <- prolfqua::sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.1, seed =3)
D$data$abundance |> is.na() |> sum()
model_name <- "f_condtion_r_peptide"
formula_randomPeptide <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1 | sampleName)",
   model_name = model_name)


mod <- prolfqua::build_model(
 D$data,
 formula_randomPeptide,
 model_name = model_name,
 subject_id = D$config$hierarchy_keys_depth())
aovtable <- mod$get_anova()

mod <- prolfqua::build_model(
 LFQData$new(D$data, D$config),
 formula_randomPeptide,
 model_name = model_name)
model_summary(mod)

Build Firth logistic model for nested LFQData

Description

Encodes missingness as a binary response and fits the peptide-aware Firth logistic backend. Proteins with multiple child features are fitted with the lowest hierarchy key appended to the formula.

Usage

build_model_glm_peptide(lfqdata, modelstr)
build_model_glm_peptide(lfqdata, modelstr)

Arguments

lfqdata

nested LFQData object

modelstr

model formula string without the response variable (e.g. "~ group_")

Value

a ModelFirth object

Examples

istar <- sim_lfq_data_peptide_config(
  Nprot = 10, with_missing = TRUE, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
mod <- build_model_glm_peptide(lfqdata, "~ group_")
head(mod$get_coefficients())
istar <- sim_lfq_data_peptide_config(
  Nprot = 10, with_missing = TRUE, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
mod <- build_model_glm_peptide(lfqdata, "~ group_")
head(mod$get_coefficients())

Build Firth logistic model for aggregated LFQData

Description

Encodes missingness as a binary response and fits the Firth logistic backend used by the missingness model path in prolfquapp.

Usage

build_model_glm_protein(lfqdata, modelstr)
build_model_glm_protein(lfqdata, modelstr)

Arguments

lfqdata

aggregated LFQData object

modelstr

model formula string without the response variable (e.g. "~ group_")

Value

a ModelFirth object

Examples

istar <- sim_lfq_data_protein_config(
  Nprot = 10, with_missing = TRUE, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
mod <- build_model_glm_protein(lfqdata, "~ group_")
head(mod$get_coefficients())
istar <- sim_lfq_data_protein_config(
  Nprot = 10, with_missing = TRUE, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
mod <- build_model_glm_protein(lfqdata, "~ group_")
head(mod$get_coefficients())

Build protein models with LOD imputation for failed fits

Description

Fits per-protein models, then re-fits failed/singular proteins after imputing missing values with the limit of detection (LOD) and clamping. Covariance is borrowed from successful fits so that variance is not underestimated by the constant imputation.

Usage

build_model_impute(
  lfqdata,
  model_strategy,
  model_name = paste0(model_strategy$model_name, "Imputed"),
  lod = NULL,
  borrow_method = c("sigma", "vcov"),
  df_method = c("observed", "borrowed"),
  on_misalign = c("fallback", "fail")
)
build_model_impute(
  lfqdata,
  model_strategy,
  model_name = paste0(model_strategy$model_name, "Imputed"),
  lod = NULL,
  borrow_method = c("sigma", "vcov"),
  df_method = c("observed", "borrowed"),
  on_misalign = c("fallback", "fail")
)

Arguments

lfqdata

LFQData object (aggregated to protein level)

model_strategy

model strategy object (e.g. from strategy_lm)

model_name

model name (default appends "Imputed")

lod

numeric limit of detection; if NULL, auto-computed from data

borrow_method

"sigma" borrows scalar sigma and uses per-protein (X'X)^-1; "vcov" borrows element-wise median of full vcov matrices

df_method

"observed" uses max(n_observed - p, 1); "borrowed" uses median df from successful fits

on_misalign

passed to compute_borrowed_variance via impute_refit_singular. "fallback" (default) preserves the lm behaviour; "fail" skips the rescue when a full covariance matrix cannot be borrowed (used by the rfit backend, which has no sigma-method fallback).

Value

a object of class Model

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
strat <- strategy_lm(paste(lfqdata$response(), "~ group_"))
mod <- build_model_impute(lfqdata, strat)

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
strat <- strategy_lm(paste(lfqdata$response(), "~ group_"))
mod <- build_model_impute(lfqdata, strat)

Build limma model from LFQData

Description

Analogous to build_model but uses limma's matrix-based pipeline. Takes an LFQData object and a strategy from strategy_limma, pivots data to wide format, fits with lmFit, and returns a ModelLimma object.

Usage

build_model_limma(lfqdata, strategy, model_name = strategy$model_name)
build_model_limma(lfqdata, strategy, model_name = strategy$model_name)

Arguments

lfqdata

an LFQData object

strategy

output of strategy_limma

model_name

name of model (default from strategy)

Value

a ModelLimma object

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod_limma <- build_model_limma(lProt, strat)
mod_limma$get_coefficients()
mod_limma$get_anova()

istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod_limma <- build_model_limma(lProt, strat)
mod_limma$get_coefficients()
mod_limma$get_anova()

Build limma model with LOD imputation for failed proteins

Description

Analogous to build_model_impute but for limma's matrix-based pipeline. Fits all proteins with lmFit, identifies proteins with NA coefficients (typically from entire missing groups), imputes their missing values with the limit of detection (LOD), refits, and replaces the variance with a borrowed estimate from successful proteins.

Usage

build_model_limma_impute(
  lfqdata,
  strategy,
  model_name = paste0(strategy$model_name, "Imputed"),
  lod = NULL,
  df_method = c("observed", "borrowed")
)
build_model_limma_impute(
  lfqdata,
  strategy,
  model_name = paste0(strategy$model_name, "Imputed"),
  lod = NULL,
  df_method = c("observed", "borrowed")
)

Arguments

lfqdata

an LFQData object (aggregated to protein level)

strategy

output of strategy_limma

model_name

name of model (default: strategy name + "Imputed")

lod

numeric limit of detection; if NULL, auto-computed from data via MissingHelpers

df_method

how to set degrees of freedom for imputed proteins: "observed" (default) uses max(n_observed - p, 1) where n_observed counts only non-missing values; "borrowed" uses the median df from successful fits

Details

The LOD imputation gives plausible coefficients (fold change direction), while the borrowed sigma and corrected degrees of freedom ensure that inference is not artificially precise from the constant imputation.

Value

a ModelLimma object with a hybrid fit

Examples

istar <- sim_lfq_data_protein_config(Nprot = 50, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_impute(lfqdata, strat)
mod$get_coefficients()

istar <- sim_lfq_data_protein_config(Nprot = 50, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_impute(lfqdata, strat)
mod$get_coefficients()

Build limma model with vooma precision weights (proteomics)

Description

Estimates observation-level precision weights from a mean-variance trend (vooma) and fits a weighted least squares model via lmFit. For proteomics data that is already log2-transformed.

Usage

build_model_limma_voom(
  lfqdata,
  strategy,
  model_name = strategy$model_name,
  span = 0.5,
  plot = FALSE
)
build_model_limma_voom(
  lfqdata,
  strategy,
  model_name = strategy$model_name,
  span = 0.5,
  plot = FALSE
)

Arguments

lfqdata

an LFQData object (aggregated to protein level)

strategy

output of strategy_limma

model_name

name of model (default from strategy)

span

lowess smoother span for mean-variance trend (default 0.5)

plot

logical; if TRUE, plot the mean-variance trend

Details

When strategy$weights is set (e.g. to nr_children), these external weights enter the preliminary fit (so the trend estimation accounts for measurement precision) and are multiplied element-wise with the vooma weights for the final fit. See the voom integration notes in ../TODO/prolfqua/Archive/TODO_limma_voom_integration.md for the mathematical basis.

Value

a ModelLimma object

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_voom(lProt, strat)
mod$get_coefficients()

istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_voom(lProt, strat)
mod$get_coefficients()

Build limma-voom model with LOD imputation for failed proteins

Description

Combines vooma precision weights with LOD imputation for proteins that have entire missing groups (NA coefficients). Mirrors build_model_limma_impute but uses vooma weights.

Usage

build_model_limma_voom_impute(
  lfqdata,
  strategy,
  model_name = paste0(strategy$model_name, "Imputed"),
  lod = NULL,
  df_method = c("observed", "borrowed"),
  span = 0.5,
  plot = FALSE
)
build_model_limma_voom_impute(
  lfqdata,
  strategy,
  model_name = paste0(strategy$model_name, "Imputed"),
  lod = NULL,
  df_method = c("observed", "borrowed"),
  span = 0.5,
  plot = FALSE
)

Arguments

lfqdata

an LFQData object (aggregated to protein level)

strategy

output of strategy_limma

model_name

name of model (default: strategy name + "Imputed")

lod

numeric limit of detection; if NULL, auto-computed from data

df_method

how to set degrees of freedom for imputed proteins: "observed" (default) uses max(n_observed - p, 1); "borrowed" uses the median df from successful fits

span

lowess smoother span for vooma trend (default 0.5)

plot

logical; if TRUE, plot the mean-variance trend

Value

a ModelLimma object with a hybrid fit

Examples

istar <- sim_lfq_data_protein_config(Nprot = 50, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_voom_impute(lfqdata, strat)
mod$get_coefficients()

istar <- sim_lfq_data_protein_config(Nprot = 50, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma_voom_impute(lfqdata, strat)
mod$get_coefficients()

Build limpa vooma model from aggregated LFQData

Description

Takes protein-level LFQData produced by AggregateLimpa (which contains standard error and observation count columns) and fits a limpa vooma model via limpa::voomaLmFitWithImputation. The standard errors from DPC quantification are used as a bivariate predictor for the vooma precision weights, and the observation counts identify imputed entries for degree-of-freedom correction.

Usage

build_model_limpa(lfqdata, strategy, model_name = strategy$model_name)
build_model_limpa(lfqdata, strategy, model_name = strategy$model_name)

Arguments

lfqdata

LFQData object. Must have config$nr_children set (used to flag imputed observations via nr_children == 0). config$opt_se is optional: when set (e.g. from AggregateLimpa), the SE column is used as a vooma precision-weight predictor; otherwise vooma is fit without an external predictor.

strategy

StrategyLimpa object

model_name

model name (default from strategy)

Details

Returns a ModelLimma object (since the output is a standard MArrayLM), which can be passed directly to ContrastsLimma.

Value

ModelLimma object

Examples

if (requireNamespace("limpa", quietly = TRUE)) {
  istar <- sim_lfq_data_peptide_config(Nprot = 10)
  lfq <- LFQData$new(istar$data, istar$config)
  lfq <- lfq$get_Transformer()$log2()$lfq
  lfq <- AggregateLimpa$new(lfq, "protein")$aggregate()
  strat <- strategy_limpa("limpa ~ group_")
  mod <- build_model_limpa(lfq, strat)
  mod$model_name
}
if (requireNamespace("limpa", quietly = TRUE)) {
  istar <- sim_lfq_data_peptide_config(Nprot = 10)
  lfq <- LFQData$new(istar$data, istar$config)
  lfq <- lfq$get_Transformer()$log2()$lfq
  lfq <- AggregateLimpa$new(lfq, "protein")$aggregate()
  strat <- strategy_limpa("limpa ~ group_")
  mod <- build_model_limpa(lfq, strat)
  mod$model_name
}

center to reference

Description

takes the median of the lfqdareference per sample and subtracts it from a copy of lfqdata

Usage

center_to_reference_cfg(lfqdata, lfqdareference)
center_to_reference_cfg(lfqdata, lfqdareference)

Arguments

lfqdata

LFQData object containing the data to center

lfqdareference

LFQData object containing the reference subset

Value

The computed result.

Examples

# example code

bb <- sim_lfq_data_peptide_config(Nprot = 100)
x <- LFQData$new(bb$data, bb$config)
xc <- x$get_copy()
xc$set_data(xc$data_long() |> dplyr::filter(protein_Id == "0EfVhX~3967"))
xxd <- center_to_reference_cfg(x, xc)
xxd$response()
x$response()

# example code

bb <- sim_lfq_data_peptide_config(Nprot = 100)
x <- LFQData$new(bb$data, bb$config)
xc <- x$get_copy()
xc$set_data(xc$data_long() |> dplyr::filter(protein_Id == "0EfVhX~3967"))
xxd <- center_to_reference_cfg(x, xc)
xxd$response()
x$response()

Contrast Configuration

Description

Holds the column-role mapping and behaviour flags that describe how a specific contrast backend (lm, limma, SAINT, ...) lays out its result table. Mirrors AnalysisConfiguration for the modelling side: it carries names, not data.

Consumers that need to operate on a contrast table generically (filter, build ORA inputs, build rank lists, summarise per protein) should read column names through the accessors on this object rather than hard-coding them. This lets backends with different column conventions (e.g. SAINT's BFDR/log2_EFCs) plug into the same downstream code path as LM-style backends without renaming columns.

Defaults match the standard prolfqua contrast schema produced by Contrasts, ContrastsLimma, ContrastsModerated etc.

Value

An R6 class generator.

Public fields

subject_id: columns identifying the contrast subject (e.g. protein id)
model_name_col: column carrying the model name label
contrast_col: column carrying the contrast label
effect_col: column with the (signed) effect size; for LM this is diff, for SAINT log2_EFCs
score_col: column with the per-contrast test statistic
pvalue_col: column with the raw p-value, or "" for backends that do not produce one (e.g. SAINTexpress). NA_character_ is accepted by the constructor and stored as "".
fdr_col: column with the FDR / BFDR / adjusted p-value
avg_abundance_col: column with the per-feature mean abundance
supports_dea_qc: whether the backend supports the legacy differential-expression QC HTML report
needs_saint_annotation: whether the backend needs annotation reading in SAINT mode (bait / control columns)
significance_directional: if TRUE, the backend only considers positive-effect features biologically meaningful (e.g. SAINTexpress, where negative log2_EFCs are not interpreted as "down" hits). Affects filter_significant(): when TRUE the filter is one-sided on the effect column; when FALSE it is symmetric (|effect| > threshold).

Methods

Public methods

ContrastConfiguration$new()
ContrastConfiguration$has_pvalue()
ContrastConfiguration$clone()

Method `new()`

Construct a ContrastConfiguration. All arguments map directly to fields of the same name.

Usage

ContrastConfiguration$new(
  subject_id = character(),
  model_name_col = "modelName",
  contrast_col = "contrast",
  effect_col = "diff",
  score_col = "statistic",
  pvalue_col = "p.value",
  fdr_col = "FDR",
  avg_abundance_col = "avgAbd",
  supports_dea_qc = TRUE,
  needs_saint_annotation = FALSE,
  significance_directional = FALSE
)

Arguments

subject_id: columns identifying the contrast subject
model_name_col: model-name column
contrast_col: contrast-label column
effect_col: signed effect column
score_col: test-statistic column
pvalue_col: raw p-value column, or "" / NA_character_ when the backend produces none
fdr_col: FDR / adjusted-p column
avg_abundance_col: mean-abundance column
supports_dea_qc: supports DEA QC HTML
needs_saint_annotation: needs SAINT-mode annotation
significance_directional: one-sided filter on effect column

Method `has_pvalue()`

Does the backend produce a usable p-value column?

Usage

ContrastConfiguration$has_pvalue()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastConfiguration$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

cfg <- ContrastConfiguration$new(subject_id = "protein_Id")
cfg$effect_col
cfg$fdr_col
cfg$has_pvalue()
cfg$supports_dea_qc

# SAINT-flavoured config
saint <- ContrastConfiguration$new(
  subject_id = "protein_Id",
  contrast_col = "Bait",
  effect_col = "log2_EFCs",
  score_col = "SaintScore",
  pvalue_col = NA_character_,
  fdr_col = "BFDR",
  supports_dea_qc = FALSE,
  needs_saint_annotation = TRUE
)
saint$has_pvalue()
cfg <- ContrastConfiguration$new(subject_id = "protein_Id")
cfg$effect_col
cfg$fdr_col
cfg$has_pvalue()
cfg$supports_dea_qc

# SAINT-flavoured config
saint <- ContrastConfiguration$new(
  subject_id = "protein_Id",
  contrast_col = "Bait",
  effect_col = "log2_EFCs",
  score_col = "SaintScore",
  pvalue_col = NA_character_,
  fdr_col = "BFDR",
  supports_dea_qc = FALSE,
  needs_saint_annotation = TRUE
)
saint$has_pvalue()

Estimate contrasts using Wald Test

Description

Estimate contrasts using Wald Test

Details

The per-protein contrast computation uses compute_contrast and linfct_matrix_contrasts internally.

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> Contrast

Public fields

models: Model
contrasts: character with contrasts
contrastfun: function to compute contrasts
model_name: model name
subject_id: name of column containing e.g., protein Id's
p.adjust: function to adjust p-values (default prolfqua::adjust_p_values)
contrast_result: data frame containing results of contrast computation
global: use a global linear function (determined by get_linfct)

Methods

Public methods

Contrasts$new()
Contrasts$get_contrast_sides()
Contrasts$get_linfct()
Contrasts$get_contrasts()
Contrasts$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_Plotter()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize create Contrast

Usage

Contrasts$new(
  model,
  contrasts,
  p.adjust = prolfqua::adjust_p_values,
  global = FALSE,
  model_name = "WaldTest"
)

Arguments

model: a dataframe with a structure similar to that generated by build_model
contrasts: a character vector with contrast specification
p.adjust: function to adjust the p-values
global: development/internal argument (if FALSE determine linfct for each model.)
model_name: name of contrast method, default WaldTest

Method `get_contrast_sides()`

get both sides of contrasts

Usage

Contrasts$get_contrast_sides()

Method `get_linfct()`

get linear functions from contrasts

Usage

Contrasts$get_linfct(global = TRUE, avg = TRUE)

Arguments

global: logical TRUE - get the a linear functions for all models, FALSE - linear function for each model
avg: logical TRUE - get also linfct for averages

Method `get_contrasts()`

get table with contrast estimates

Usage

Contrasts$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)

Returns

data.frame with contrasts

Method `clone()`

The objects of this class are cloneable with this method.

Usage

Contrasts$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


# Fitting mixed effects model to peptide data
istar <- prolfqua::sim_lfq_data_peptide_config()

modelFunction <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1 | sample)")

config <- istar$config
config$hierarchy_keys_depth()

mod <- build_model(
  istar$data,
  modelFunction,
  subject_id = config$hierarchy_keys_depth()
)

ref_lfc <- data.frame(
  `(Intercept)` = c(0, 0, 0),
  group_B = c(0, 1, 0),
  group_Ctrl = c(-1, -1, 1),
  row.names = c("groupA_vs_Ctrl", "dil.e_vs_b", "dil.ctrl_vs_b")
)
prolfqua::model_summary(mod)
Contr <- c(
  "groupA_vs_Ctrl" = "group_A - group_Ctrl",
  "dil.e_vs_b" = "group_B - group_Ctrl",
  "dil.ctrl_vs_b" = "group_Ctrl - group_A"
)
contrastX <- prolfqua::Contrasts$new(mod, Contr)
y <- contrastX$get_linfct(avg = FALSE)
stopifnot(all(ref_lfc == y))
t <- contrastX$get_linfct(global = FALSE)

x <- contrastX$get_contrasts()
stopifnot(all(x$p.value < 1 & x$p.value > 0))
stopifnot(all(x$avgAbd > 0))
stopifnot(all(x$FDR > 0 & x$FDR < 1))

x <- contrastX$get_contrast_sides()
xd <- contrastX$column_description()
modelFunction <-
  strategy_lm("abundance  ~ group_")
mod <- build_model(
  istar$data,
  modelFunction,
  subject_id = config$hierarchy_keys_depth()
)
contrastX <- prolfqua::Contrasts$new(mod, Contr)
y <- contrastX$get_linfct(avg = FALSE)
stopifnot(all(ref_lfc == y))

# Fitting mixed effects model to peptide data
istar <- prolfqua::sim_lfq_data_peptide_config()

modelFunction <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1 | sample)")

config <- istar$config
config$hierarchy_keys_depth()

mod <- build_model(
  istar$data,
  modelFunction,
  subject_id = config$hierarchy_keys_depth()
)

ref_lfc <- data.frame(
  `(Intercept)` = c(0, 0, 0),
  group_B = c(0, 1, 0),
  group_Ctrl = c(-1, -1, 1),
  row.names = c("groupA_vs_Ctrl", "dil.e_vs_b", "dil.ctrl_vs_b")
)
prolfqua::model_summary(mod)
Contr <- c(
  "groupA_vs_Ctrl" = "group_A - group_Ctrl",
  "dil.e_vs_b" = "group_B - group_Ctrl",
  "dil.ctrl_vs_b" = "group_Ctrl - group_A"
)
contrastX <- prolfqua::Contrasts$new(mod, Contr)
y <- contrastX$get_linfct(avg = FALSE)
stopifnot(all(ref_lfc == y))
t <- contrastX$get_linfct(global = FALSE)

x <- contrastX$get_contrasts()
stopifnot(all(x$p.value < 1 & x$p.value > 0))
stopifnot(all(x$avgAbd > 0))
stopifnot(all(x$FDR > 0 & x$FDR < 1))

x <- contrastX$get_contrast_sides()
xd <- contrastX$column_description()
modelFunction <-
  strategy_lm("abundance  ~ group_")
mod <- build_model(
  istar$data,
  modelFunction,
  subject_id = config$hierarchy_keys_depth()
)
contrastX <- prolfqua::Contrasts$new(mod, Contr)
y <- contrastX$get_linfct(avg = FALSE)
stopifnot(all(ref_lfc == y))

Quasibinomial detection-count facade for nested input

Description

Quasibinomial detection-count facade for nested input

Details

Completes and encodes child-feature detection using the same preparation as ContrastsFirthNestedFacade, then collapses the binary rows into detected and undetected counts per parent and sample. The resulting quasibinomial model reports parent-level log odds ratios in diff; avgAbd is the average linear predictor on the log-odds scale.

Child features are treated as exchangeable binomial trials. The symmetric pseudo-count stabilizes complete separation but is not equivalent to Firth's bias-reducing penalty. Empirical-Bayes dispersion moderation uses ContrastsModerated; by default posterior dispersion is bounded below by one.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsBinomialNestedFacade

Methods

Public methods

ContrastsBinomialNestedFacade$new()
ContrastsBinomialNestedFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

Fit the nested detection-count model and its contrasts.

Usage

ContrastsBinomialNestedFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  prior_count = 0.1,
  binomial_bound = TRUE,
  ...
)

Arguments

lfqdata: nested LFQData
modelstr: right-hand-side model formula
contrasts: named contrast expressions
prior_count: non-negative symmetric pseudo-count
binomial_bound: bound posterior dispersion below by one
...: passed to strategy_binomial

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsBinomialNestedFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
facade <- ContrastsBinomialNestedFacade$new(lfqdata, "~ group_", contrasts)
head(facade$get_contrasts())
istar <- sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.5, seed = 3)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
facade <- ContrastsBinomialNestedFacade$new(lfqdata, "~ group_", contrasts)
head(facade$get_contrasts())

DEqMS contrast analysis facade

Description

DEqMS contrast analysis facade

Details

Encapsulates the pipeline: strategy_lm -> build_model -> Contrasts -> ContrastsModeratedDEqMS.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsDEqMSFacade

Methods

Public methods

ContrastsDEqMSFacade$new()
ContrastsDEqMSFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsDEqMSFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  weights = lfqdata$nr_children_col(),
  ...
)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
...: passed to strategy_lm

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsDEqMSFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsDEqMSFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsDEqMSFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

DEqMS contrast analysis with vooma weights facade

Description

DEqMS contrast analysis with vooma weights facade

Details

Combines vooma precision weights (mean-variance trend) with DEqMS count-dependent variance moderation. Vooma runs without external weights so it captures only the mean-variance relationship; the peptide count enters solely through DEqMS moderation, avoiding double-counting.

Pipeline: strategy_limma (weights = NULL) -> build_model_limma_voom -> ContrastsLimma (eBayes = FALSE) -> ContrastsModeratedDEqMS.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsDEqMSVoomFacade

Methods

Public methods

ContrastsDEqMSVoomFacade$new()
ContrastsDEqMSVoomFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsDEqMSVoomFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  span = 0.5,
  plot = FALSE,
  ...
)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
span: lowess smoother span for vooma trend (default 0.5)
plot: logical; if TRUE, plot the mean-variance trend
...: passed to strategy_limma (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsDEqMSVoomFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsDEqMSVoomFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
istar <- sim_lfq_data_protein_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsDEqMSVoomFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())

Estimate contrasts using Wald Test

Description

Estimate contrasts using Wald Test

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsFirth

Public fields

models: Model
contrasts: character with contrasts
model_name: name of model
subject_id: name of column containing e.g., protein Id's
p.adjust: function to adjust p-values (default prolfqua::adjust_p_values)
contrast_result: data frame containing results of contrast computation

Methods

Public methods

ContrastsFirth$new()
ContrastsFirth$get_contrast_sides()
ContrastsFirth$get_linfct()
ContrastsFirth$get_contrasts()
ContrastsFirth$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_Plotter()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize create Contrast

Usage

ContrastsFirth$new(
  model,
  contrasts,
  p.adjust = prolfqua::adjust_p_values,
  model_name = "firth"
)

Arguments

model: a dataframe with a structure similar to that generated by build_model
contrasts: a character vector with contrast specification
p.adjust: function to adjust the p-values
model_name: name of contrast method, default firth

Method `get_contrast_sides()`

get both sides of contrasts

Usage

ContrastsFirth$get_contrast_sides()

Method `get_linfct()`

get linear functions from contrasts

Usage

ContrastsFirth$get_linfct(avg = TRUE)

Arguments

avg: logical TRUE - get also linfct for averages

Returns

a list with the linfct-annotated model containers (models1 / models2); the shared input model is left untouched.

Method `get_contrasts()`

get table with contrast estimates

Usage

ContrastsFirth$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)

Returns

data.frame with contrasts

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsFirth$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


modi <- sim_build_models_logistf(model = "parallel3", weight_missing = 1)
contrasts <- c(Avs = "group_A - group_B", AvsCtrl = "group_A - group_Ctrl")

ctr <- ContrastsFirth$new(modi,contrasts)
ctr$get_contrast_sides()
ctr$get_linfct()
ctr$get_contrasts()

mod3 <- sim_build_models_logistf(model = "parallel3", weight_missing = 1, peptide=TRUE)

ctrpep <- ContrastsFirth$new(mod3,contrasts)
ctrpep$get_contrast_sides()

ctrpep$get_linfct()
ctrpep$get_contrasts()
pl <- ctrpep$get_Plotter()

modi <- sim_build_models_logistf(model = "parallel3", weight_missing = 1)
contrasts <- c(Avs = "group_A - group_B", AvsCtrl = "group_A - group_Ctrl")

ctr <- ContrastsFirth$new(modi,contrasts)
ctr$get_contrast_sides()
ctr$get_linfct()
ctr$get_contrasts()

mod3 <- sim_build_models_logistf(model = "parallel3", weight_missing = 1, peptide=TRUE)

ctrpep <- ContrastsFirth$new(mod3,contrasts)
ctrpep$get_contrast_sides()

ctrpep$get_linfct()
ctrpep$get_contrasts()
pl <- ctrpep$get_Plotter()

Firth logistic missingness contrast analysis facade

Description

Firth logistic missingness contrast analysis facade

Details

Encapsulates the pipeline: encode missingness -> build_model_glm_protein -> ContrastsFirth.

Requires aggregated (protein-level) LFQData. For nested peptide-level input use ContrastsFirthNestedFacade.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsFirthFacade

Methods

Public methods

ContrastsFirthFacade$new()
ContrastsFirthFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsFirthFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
...: ignored; accepted for dispatch compatibility with other facades

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsFirthFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 20, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsFirthFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 20, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsFirthFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Firth logistic missingness contrast analysis facade for nested input

Description

Firth logistic missingness contrast analysis facade for nested input

Details

Encapsulates the pipeline: encode missingness -> build_model_glm_peptide -> ContrastsFirth.

Takes nested (peptide-level) LFQData and returns protein-level fold-change estimates. For protein-level (aggregated) input use ContrastsFirthFacade instead.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsFirthNestedFacade

Methods

Public methods

ContrastsFirthNestedFacade$new()
ContrastsFirthNestedFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsFirthNestedFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: nested LFQData (peptide-level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
...: ignored; accepted for dispatch compatibility with other facades

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsFirthNestedFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsFirthNestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_peptide_config(Nprot = 20, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsFirthNestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Base class for all Contrasts classes

Description

Base class for all Contrasts classes

Value

An R6 class generator.

Public fields

config: ContrastConfiguration describing the backend's column-role mapping and behaviour flags. Subclasses should populate this in initialize() so generic consumers can resolve column names without hard-coding them.

Methods

Public methods

ContrastsInterface$get_config()
ContrastsInterface$get_contrast_sides()
ContrastsInterface$get_contrasts()
ContrastsInterface$get_Plotter()
ContrastsInterface$to_wide()
ContrastsInterface$get_missing()
ContrastsInterface$get_rank()
ContrastsInterface$get_ora()
ContrastsInterface$filter_significant()
ContrastsInterface$contrast_summary_table()
ContrastsInterface$extra_artifacts()
ContrastsInterface$column_description()
ContrastsInterface$clone()

Method `get_config()`

Return the ContrastConfiguration, constructing an LM-flavoured default on demand if a subclass has not set one.

Usage

ContrastsInterface$get_config()

Method `get_contrast_sides()`

get table with sides of the contrast

Usage

ContrastsInterface$get_contrast_sides()

Method `get_contrasts()`

get table with contrast results (similar to limma topTable function)

Usage

ContrastsInterface$get_contrasts()

Method `get_Plotter()`

return ContrastsPlotter for the contrast results

Usage

ContrastsInterface$get_Plotter(fc_threshold = 1, fdr_threshold = 0.1)

Arguments

fc_threshold: fold change threshold to show in plots
fdr_threshold: FDR threshold to show in plots

Returns

ContrastsPlotter

Method `to_wide()`

convert contrast results to wide format

Usage

ContrastsInterface$to_wide(columns = c("p.value", "FDR", "statistic"))

Arguments

columns: value columns to spread next to diff

Returns

data.frame

Method `get_missing()`

get protein × contrast pairs that could not be estimated. Returns a data.frame with hierarchy columns and contrast, or 0 rows.

Usage

ContrastsInterface$get_missing()

Method `get_rank()`

get rank-list input table for enrichment tools.

If the backend's ContrastConfiguration has a p-value column the default rank score is signed sign(effect) * -log10(p.value). Otherwise the default is the effect column itself.

Usage

ContrastsInterface$get_rank(score = NULL)

Arguments

score: column name to use as rank score, or NULL to pick from the config

Returns

data.frame with subject id, contrast, and score columns.

Method `get_ora()`

get ORA input table — features passing the FDR and absolute effect-size threshold in the requested direction.

Usage

ContrastsInterface$get_ora(up = TRUE, FDR_threshold = 0.05, diff_threshold = 1)

Arguments

up: if TRUE return positive effects, otherwise negative effects
FDR_threshold: false discovery rate threshold
diff_threshold: absolute effect-size threshold

Returns

filtered contrast data.frame for ORA input.

Method `filter_significant()`

Filter contrast rows that pass the FDR and effect-size threshold.

By default this is symmetric on the effect column (|effect| > diff_threshold). For backends whose ContrastConfiguration has significance_directional = TRUE (e.g. SAINTexpress, where only positive enrichment is biologically meaningful), the filter is one-sided (effect > diff_threshold).

Usage

ContrastsInterface$filter_significant(FDR_threshold = 0.05, diff_threshold = 1)

Arguments

FDR_threshold: false discovery rate threshold
diff_threshold: effect-size threshold

Returns

filtered contrast data.frame.

Method `contrast_summary_table()`

Per-subject summary table for downstream report grobs. Returns a small data.frame with canonical column names contrast, effect, score, fdr regardless of backend.

Usage

ContrastsInterface$contrast_summary_table(rounded = TRUE)

Arguments

rounded: if TRUE round numeric columns to 3 significant digits (matches the prolfquapp report convention)

Method `extra_artifacts()`

Backend-specific extra artifacts that a report should carry along with the contrast results (e.g. SAINT input tables). Default returns an empty named list.

Usage

ContrastsInterface$extra_artifacts()

Method `column_description()`

column description

Usage

ContrastsInterface$column_description()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsInterface$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

int <- ContrastsInterface$new()
testthat::expect_error(int$get_contrast_sides())
testthat::expect_error(int$get_contrasts())
testthat::expect_error(int$get_missing())
# get_Plotter / to_wide / get_rank / get_ora / filter_significant call get_contrasts()
# internally, which is not implemented on the bare interface, so they surface that error.
testthat::expect_error(int$get_Plotter())
testthat::expect_error(int$to_wide())
testthat::expect_error(int$get_rank())
testthat::expect_error(int$get_ora())
testthat::expect_error(int$filter_significant())
identical(int$extra_artifacts(), list())
int$column_description()
int <- ContrastsInterface$new()
testthat::expect_error(int$get_contrast_sides())
testthat::expect_error(int$get_contrasts())
testthat::expect_error(int$get_missing())
# get_Plotter / to_wide / get_rank / get_ora / filter_significant call get_contrasts()
# internally, which is not implemented on the bare interface, so they surface that error.
testthat::expect_error(int$get_Plotter())
testthat::expect_error(int$to_wide())
testthat::expect_error(int$get_rank())
testthat::expect_error(int$get_ora())
testthat::expect_error(int$filter_significant())
identical(int$extra_artifacts(), list())
int$column_description()

Limma-based contrasts (direct limma pipeline)

Description

Limma-based contrasts (direct limma pipeline)

Details

Uses limma's contrasts.fit + eBayes pipeline directly, rather than fitting per-protein lm models and then moderating. Inherits from ContrastsInterface with the same API as Contrasts.

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsLimma

Public fields

model: ModelLimma object
contrasts: named character vector of contrasts
model_name: model name
subject_id: columns with subject_id (proteinID)
p.adjust: function to adjust p-values
contrast_result: cached contrast results
eBayes: logical, apply limma eBayes moderation (default TRUE). Set to FALSE to return raw/unmoderated statistics, e.g. for downstream DEqMS moderation via ContrastsModeratedDEqMS.

Methods

Public methods

ContrastsLimma$new()
ContrastsLimma$get_contrast_sides()
ContrastsLimma$get_linfct()
ContrastsLimma$get_contrasts()
ContrastsLimma$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_Plotter()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize ContrastsLimma

Usage

ContrastsLimma$new(
  model,
  contrasts,
  p.adjust = prolfqua::adjust_p_values,
  model_name = NULL,
  eBayes = TRUE
)

Arguments

model: a ModelLimma object
contrasts: named character vector of contrasts
p.adjust: function to adjust p-values
model_name: name of the contrast method
eBayes: logical, apply limma eBayes moderation (default TRUE). Set to FALSE to return raw/unmoderated statistics suitable for wrapping with ContrastsModeratedDEqMS.

Method `get_contrast_sides()`

get both sides of contrasts

Usage

ContrastsLimma$get_contrast_sides()

Method `get_linfct()`

get linear functions from contrasts

Usage

ContrastsLimma$get_linfct(global = TRUE, avg = TRUE)

Arguments

global: ignored (for API compatibility)
avg: logical, also compute avgAbd linfct

Method `get_contrasts()`

get table with contrast estimates via limma pipeline

Usage

ContrastsLimma$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)

Returns

data.frame with contrasts

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimma$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod_limma <- build_model_limma(lProt, strat)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contr_limma <- ContrastsLimma$new(mod_limma, Contr)
res <- contr_limma$get_contrasts()
head(res)
stopifnot(all(c("diff", "FDR", "p.value", "statistic") %in% colnames(res)))

# Compare with prolfqua's own pipeline
modelFunction <- strategy_lm("transformedIntensity ~ group_")
mod <- build_model(lProt, modelFunction)
contr_prolfqua <- Contrasts$new(mod, Contr)
res_prolfqua <- contr_prolfqua$get_contrasts()

# fold changes should be very similar
merged <- dplyr::inner_join(
  dplyr::select(res, protein_Id, diff_limma = diff),
  dplyr::select(res_prolfqua, protein_Id, diff_prolfqua = diff),
  by = "protein_Id")
stopifnot(cor(merged$diff_limma, merged$diff_prolfqua, use = "complete.obs") > 0.99)

# Plotter works
pl <- contr_limma$get_Plotter()

# to_wide works
wide <- contr_limma$to_wide()
head(wide)

# merge_contrasts_results works
Contr2 <- c("dil.b_vs_a" = "group_A - group_Ctrl")
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr2)$contrast
merged_res <- merge_contrasts_results(contr_limma, imp)

istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")

strat <- strategy_limma("transformedIntensity ~ group_")
mod_limma <- build_model_limma(lProt, strat)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contr_limma <- ContrastsLimma$new(mod_limma, Contr)
res <- contr_limma$get_contrasts()
head(res)
stopifnot(all(c("diff", "FDR", "p.value", "statistic") %in% colnames(res)))

# Compare with prolfqua's own pipeline
modelFunction <- strategy_lm("transformedIntensity ~ group_")
mod <- build_model(lProt, modelFunction)
contr_prolfqua <- Contrasts$new(mod, Contr)
res_prolfqua <- contr_prolfqua$get_contrasts()

# fold changes should be very similar
merged <- dplyr::inner_join(
  dplyr::select(res, protein_Id, diff_limma = diff),
  dplyr::select(res_prolfqua, protein_Id, diff_prolfqua = diff),
  by = "protein_Id")
stopifnot(cor(merged$diff_limma, merged$diff_prolfqua, use = "complete.obs") > 0.99)

# Plotter works
pl <- contr_limma$get_Plotter()

# to_wide works
wide <- contr_limma$to_wide()
head(wide)

# merge_contrasts_results works
Contr2 <- c("dil.b_vs_a" = "group_A - group_Ctrl")
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr2)$contrast
merged_res <- merge_contrasts_results(contr_limma, imp)

Limma contrast analysis facade

Description

Limma contrast analysis facade

Details

Encapsulates the pipeline: strategy_limma -> build_model_limma -> ContrastsLimma.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimmaFacade

Methods

Public methods

ContrastsLimmaFacade$new()
ContrastsLimmaFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimmaFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  weights = lfqdata$nr_children_col(),
  ...
)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
...: passed to strategy_limma (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimmaFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Limma contrast analysis with LOD imputation facade

Description

Limma contrast analysis with LOD imputation facade

Details

Encapsulates the pipeline: strategy_limma -> build_model_limma_impute -> ContrastsLimma.

Proteins whose limma fit produces NA coefficients (typically from entire missing groups) are recovered by imputing missing values with the limit of detection (LOD) and refitting. The variance is borrowed from successful fits and degrees of freedom are corrected so that inference is not artificially precise from the constant imputation.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimmaImputeFacade

Methods

Public methods

ContrastsLimmaImputeFacade$new()
ContrastsLimmaImputeFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimmaImputeFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  lod = NULL,
  df_method = c("observed", "borrowed"),
  weights = lfqdata$nr_children_col(),
  ...
)

Arguments

lfqdata: LFQData object (aggregated to protein level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
lod: numeric limit of detection; if NULL, auto-computed from data
df_method: "observed" uses max(n_observed - p, 1); "borrowed" uses median df from successful fits
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
...: passed to strategy_limma (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimmaImputeFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Limma-voom contrast analysis facade

Description

Limma-voom contrast analysis facade

Details

Encapsulates the pipeline: strategy_limma -> build_model_limma_voom -> ContrastsLimma.

Uses vooma-style precision weights derived from a mean-variance trend, optionally combined with external weights (e.g. peptide/precursor counts).

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimmaVoomFacade

Methods

Public methods

ContrastsLimmaVoomFacade$new()
ContrastsLimmaVoomFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimmaVoomFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  weights = lfqdata$nr_children_col(),
  span = 0.5,
  plot = FALSE,
  ...
)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
span: lowess smoother span for vooma trend (default 0.5)
plot: logical; if TRUE, plot the mean-variance trend
...: passed to strategy_limma (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimmaVoomFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaVoomFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaVoomFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Limma-voom contrast analysis with LOD imputation facade

Description

Limma-voom contrast analysis with LOD imputation facade

Details

Encapsulates the pipeline: strategy_limma -> build_model_limma_voom_impute -> ContrastsLimma.

Combines vooma precision weights with LOD imputation for proteins whose fit produces NA coefficients (typically from entire missing groups).

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimmaVoomImputeFacade

Methods

Public methods

ContrastsLimmaVoomImputeFacade$new()
ContrastsLimmaVoomImputeFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimmaVoomImputeFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  lod = NULL,
  df_method = c("observed", "borrowed"),
  weights = lfqdata$nr_children_col(),
  span = 0.5,
  plot = FALSE,
  ...
)

Arguments

lfqdata: LFQData object (aggregated to protein level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
lod: numeric limit of detection; if NULL, auto-computed from data
df_method: "observed" uses max(n_observed - p, 1); "borrowed" uses median df from successful fits
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
span: lowess smoother span for vooma trend (default 0.5)
plot: logical; if TRUE, plot the mean-variance trend
...: passed to strategy_limma (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimmaVoomImputeFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaVoomImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimmaVoomImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Limpa contrast analysis facade

Description

Limpa contrast analysis facade

Details

Encapsulates the pipeline: strategy_limpa -> build_model_limpa -> ContrastsLimma.

Operates as a "same"-level facade: it consumes whatever aggregated LFQData prolfqua's normal aggregation pipeline produced and fits limpa at that hierarchy level. If config$opt_se is set (e.g. when the input came from AggregateLimpa), the per-observation standard error is used as a vooma precision-weight predictor; otherwise plain vooma is fit. The config$nr_children column is required and is used to flag imputed observations (nr_children == 0) for vooma's imputation-aware DF correction.

For nested (peptide/precursor) input that should be rolled up to proteins via limpa's DPC quantification, use ContrastsLimpaNestedFacade instead — that facade owns the AggregateLimpa pre-step.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimpaFacade

Methods

Public methods

ContrastsLimpaFacade$new()
ContrastsLimpaFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimpaFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  plot = FALSE,
  span = NULL,
  ...
)

Arguments

lfqdata: aggregated LFQData. If config$opt_se is set (e.g. from AggregateLimpa), the SE column is used as a vooma precision-weight predictor; otherwise vooma is fit without an external predictor.
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
plot: logical; if TRUE, plot the vooma mean-variance trend
span: lowess smoother span (NULL = auto)
...: passed to strategy_limpa (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimpaFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
agg <- AggregateLimpa$new(lfqdata, "protein")
lfq_agg <- agg$aggregate()
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimpaFacade$new(lfq_agg, "~ group_", contrasts)
head(fa$get_contrasts())
istar <- prolfqua::sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
agg <- AggregateLimpa$new(lfqdata, "protein")
lfq_agg <- agg$aggregate()
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLimpaFacade$new(lfq_agg, "~ group_", contrasts)
head(fa$get_contrasts())

Limpa contrast analysis facade for nested input

Description

Limpa contrast analysis facade for nested input

Details

Encapsulates the full precursor -> protein pipeline: AggregateLimpa -> strategy_limpa -> build_model_limpa -> ContrastsLimma.

Takes nested (precursor/peptide-level) LFQData, runs limpa's DPC-based aggregation internally to produce protein-level expression with posterior standard errors, then fits a vooma model with imputation-aware precision weights.

For protein-level input that was already aggregated upstream via AggregateLimpa, use ContrastsLimpaFacade instead.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLimpaNestedFacade

Methods

Public methods

ContrastsLimpaNestedFacade$new()
ContrastsLimpaNestedFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLimpaNestedFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  prefix = "protein",
  dpc_slope = 0.8,
  plot = FALSE,
  span = NULL,
  ...
)

Arguments

lfqdata: nested LFQData (precursor/peptide-level, log2-transformed)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
prefix: prefix for the aggregated hierarchy level (default "protein")
dpc_slope: DPC slope parameter passed to AggregateLimpa (default 0.8)
plot: logical; if TRUE, plot the vooma mean-variance trend
span: lowess smoother span (NULL = auto)
...: passed to strategy_limpa (e.g. trend, robust)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLimpaNestedFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (requireNamespace("limpa", quietly = TRUE)) {
  istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10)
  lfqdata <- LFQData$new(istar$data, istar$config)
  lfqdata <- lfqdata$get_Transformer()$log2()$lfq
  contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
  fa <- ContrastsLimpaNestedFacade$new(lfqdata, "~ group_", contrasts)
  head(fa$get_contrasts())
}
if (requireNamespace("limpa", quietly = TRUE)) {
  istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10)
  lfqdata <- LFQData$new(istar$data, istar$config)
  lfqdata <- lfqdata$get_Transformer()$log2()$lfq
  contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
  fa <- ContrastsLimpaNestedFacade$new(lfqdata, "~ group_", contrasts)
  head(fa$get_contrasts())
}

Lmer (mixed-model) contrast analysis facade for nested input

Description

Lmer (mixed-model) contrast analysis facade for nested input

Details

Encapsulates the pipeline: strategy_lmer -> build_model -> Contrasts -> ContrastsModerated.

Takes child-hierarchy LFQData (e.g. peptide-level measurements nested within proteins) and returns protein-level fold-change estimates.

modelstr may be either a fixed-effects-only formula (e.g. "~ group_") - in which case the facade appends (1 | <deepest child hierarchy key>) + (1 | <sample name>) derived from the LFQData - or a full mixed-model formula already containing random effects, which is passed through unchanged.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLmerNestedFacade

Methods

Public methods

ContrastsLmerNestedFacade$new()
ContrastsLmerNestedFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLmerNestedFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: nested LFQData (subject_id strict subset of hierarchy_keys)
modelstr: model formula RHS as string. Fixed effects only (e.g. "~ group_") - random effects are derived from lfqdata as (1 | <deepest child key>) + (1 | <sample name>). A formula already containing random-effect bars is passed through unchanged.
contrasts: named character vector of contrasts
...: passed to strategy_lmer

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLmerNestedFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
# Random effects derived from lfqdata; equivalent to passing the full
# mixed-model formula explicitly.
fa <- ContrastsLmerNestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
# Random effects derived from lfqdata; equivalent to passing the full
# mixed-model formula explicitly.
fa <- ContrastsLmerNestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

LM contrast analysis facade

Description

LM contrast analysis facade

Details

Encapsulates the pipeline: strategy_lm -> build_model -> Contrasts -> ContrastsModerated.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLMFacade

Methods

Public methods

ContrastsLMFacade$new()
ContrastsLMFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLMFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  weights = lfqdata$nr_children_col(),
  ...
)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
...: passed to strategy_lm

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLMFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLMFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLMFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

LM contrast analysis with LOD imputation facade

Description

LM contrast analysis with LOD imputation facade

Details

Encapsulates the pipeline: strategy_lm -> build_model (with impute = TRUE) -> Contrasts -> ContrastsModerated.

Proteins whose initial lm fit fails or produces NA coefficients are re-fitted after imputing missing values with the limit of detection (LOD). The covariance matrix is borrowed from successful fits so that the variance is not underestimated by the constant imputation.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsLMImputeFacade

Methods

Public methods

ContrastsLMImputeFacade$new()
ContrastsLMImputeFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsLMImputeFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  lod = NULL,
  borrow_method = c("sigma", "vcov"),
  df_method = c("observed", "borrowed"),
  weights = lfqdata$nr_children_col(),
  ...
)

Arguments

lfqdata: LFQData object (aggregated to protein level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
lod: numeric limit of detection; if NULL, auto-computed from data
borrow_method: "sigma" borrows scalar sigma and uses per-protein (X'X)^-1; "vcov" borrows element-wise median of full vcov matrices
df_method: "observed" uses max(n_observed - p, 1); "borrowed" uses median df from successful fits
weights: column name for per-observation weights (default: lfqdata$nr_children_col()). Pass NULL for unweighted.
...: passed to strategy_lm

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsLMImputeFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLMImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsLMImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Limma moderated contrasts

Description

Limma moderated contrasts

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsModerated

Public fields

Contrast: Class implementing the Contrast interface
model_name: name of model
subject_id: columns with subject_id (proteinID)
p.adjust: function to adjust p-values
variance_floor: optional lower bound for posterior variances

Methods

Public methods

ContrastsModerated$new()
ContrastsModerated$get_contrast_sides()
ContrastsModerated$get_linfct()
ContrastsModerated$get_contrasts()
ContrastsModerated$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_Plotter()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize

Usage

ContrastsModerated$new(
  Contrast,
  model_name = paste0(Contrast$model_name, "_moderated"),
  p.adjust = prolfqua::adjust_p_values,
  variance_floor = NULL
)

Arguments

Contrast: class implementing the ContrastInterface
model_name: name of the model
p.adjust: function to adjust p-values - default BH
variance_floor: optional positive lower bound for posterior variances

Method `get_contrast_sides()`

get both sides of contrasts

Usage

ContrastsModerated$get_contrast_sides()

Method `get_linfct()`

get linear functions from contrasts

Usage

ContrastsModerated$get_linfct(global = TRUE)

Arguments

global: logical TRUE - get the a linear functions for all models, FALSE - linear function for each model

Method `get_contrasts()`

applies limma moderation

Usage

ContrastsModerated$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)
global: use a global linear function (determined by get_linfct)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsModerated$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
protIntensity <- istar$data
config <- istar$config


lProt <- LFQData$new(protIntensity, config)
lProt$rename_response("transformedIntensity")
modelFunction <-
  strategy_lm("transformedIntensity  ~ group_")
mod <- build_model(
 lProt,
 modelFunction)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contrast <- prolfqua::Contrasts$new(mod,
 Contr)
contrast <- ContrastsModerated$new(contrast)
bb <- contrast$get_contrasts()
# LOD-imputed refit for proteins the plain lm cannot estimate
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr)$contrast

contrast$get_contrasts() |> dim()
(xx <- imp$get_contrasts())   |> dim()
merged <- merge_contrasts_results(contrast, imp)

merged$more$get_contrasts() |> dim()
stopifnot(all(dim(merged$merged$get_contrasts()) == c(50, 16)))
stopifnot(all(dim(merged$same$get_contrasts()) == c(49, 16)))

cs <- contrast$get_contrast_sides()
cslf <- contrast$get_linfct()
ctr <- contrast$get_contrasts()
ctrwide <- contrast$to_wide()
cp <- contrast$get_Plotter()

print(cp$histogram()$p.value, vp=NULL)
print(cp$histogram()$FDR, vp = NULL)

cp$volcano()
cp$ma_plot()



istar <- sim_lfq_data_protein_config(Nprot = 50)
protIntensity <- istar$data
config <- istar$config


lProt <- LFQData$new(protIntensity, config)
lProt$rename_response("transformedIntensity")
modelFunction <-
  strategy_lm("transformedIntensity  ~ group_")
mod <- build_model(
 lProt,
 modelFunction)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contrast <- prolfqua::Contrasts$new(mod,
 Contr)
contrast <- ContrastsModerated$new(contrast)
bb <- contrast$get_contrasts()
# LOD-imputed refit for proteins the plain lm cannot estimate
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr)$contrast

contrast$get_contrasts() |> dim()
(xx <- imp$get_contrasts())   |> dim()
merged <- merge_contrasts_results(contrast, imp)

merged$more$get_contrasts() |> dim()
stopifnot(all(dim(merged$merged$get_contrasts()) == c(50, 16)))
stopifnot(all(dim(merged$same$get_contrasts()) == c(49, 16)))

cs <- contrast$get_contrast_sides()
cslf <- contrast$get_linfct()
ctr <- contrast$get_contrasts()
ctrwide <- contrast$to_wide()
cp <- contrast$get_Plotter()

print(cp$histogram()$p.value, vp=NULL)
print(cp$histogram()$FDR, vp = NULL)

cp$volcano()
cp$ma_plot()

DEqMS count-dependent moderated contrasts

Description

DEqMS count-dependent moderated contrasts

Details

Decorator that wraps any Contrasts object and applies count-dependent empirical Bayes variance shrinkage. Similar to ContrastsModerated but the prior variance depends on the number of quantified peptides/PSMs per protein: proteins with many peptides get less shrinkage, proteins with few peptides get more.

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsModeratedDEqMS

Public fields

Contrast: Class implementing the Contrast interface
count_df: data.frame with subject_id + count column
count_column: name of the count column in count_df
loess_span: span parameter for LOESS fit
model_name: name of model
subject_id: columns with subject_id (proteinID)
p.adjust: function to adjust p-values

Methods

Public methods

ContrastsModeratedDEqMS$new()
ContrastsModeratedDEqMS$get_contrast_sides()
ContrastsModeratedDEqMS$get_linfct()
ContrastsModeratedDEqMS$get_contrasts()
ContrastsModeratedDEqMS$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_Plotter()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize

Usage

ContrastsModeratedDEqMS$new(
  Contrast,
  count_df,
  count_column,
  loess_span = 0.75,
  model_name = paste0(Contrast$model_name, "_DEqMS"),
  p.adjust = prolfqua::adjust_p_values
)

Arguments

Contrast: class implementing the ContrastInterface
count_df: data.frame with subject_id columns and a count column
count_column: name of the count column in count_df
loess_span: span for LOESS variance fit (default 0.75)
model_name: name of the model
p.adjust: function to adjust p-values - default BH

Method `get_contrast_sides()`

get both sides of contrasts

Usage

ContrastsModeratedDEqMS$get_contrast_sides()

Method `get_linfct()`

get linear functions from contrasts

Usage

ContrastsModeratedDEqMS$get_linfct(global = TRUE)

Arguments

global: logical TRUE - get linear functions for all models

Method `get_contrasts()`

applies DEqMS-style count-dependent moderation

Usage

ContrastsModeratedDEqMS$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsModeratedDEqMS$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
protIntensity <- istar$data
config <- istar$config

lProt <- LFQData$new(protIntensity, config)
lProt$rename_response("transformedIntensity")
modelFunction <-
  strategy_lm("transformedIntensity ~ group_")
mod <- build_model(
 lProt,
 modelFunction)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contrast <- prolfqua::Contrasts$new(mod, Contr)

# Build count_df from config
count_df <- dplyr::select(protIntensity,
  dplyr::all_of(c(config$hierarchy_keys_depth(), "nrPeptides"))) |>
  dplyr::distinct()

deqms <- ContrastsModeratedDEqMS$new(contrast,
  count_df = count_df,
  count_column = "nrPeptides")

bb <- deqms$get_contrasts()
stopifnot(all(c("diff", "p.value", "FDR", "sigma") %in% colnames(bb)))

# Merge with LOD-imputed refits for proteins the plain lm cannot estimate
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr)$contrast
merged <- merge_contrasts_results(deqms, imp)

cs <- deqms$get_contrast_sides()
cslf <- deqms$get_linfct()
ctrwide <- deqms$to_wide()
cp <- deqms$get_Plotter()
cp$volcano()

istar <- sim_lfq_data_protein_config(Nprot = 50)
protIntensity <- istar$data
config <- istar$config

lProt <- LFQData$new(protIntensity, config)
lProt$rename_response("transformedIntensity")
modelFunction <-
  strategy_lm("transformedIntensity ~ group_")
mod <- build_model(
 lProt,
 modelFunction)

Contr <- c("dil.b_vs_a" = "group_A - group_Ctrl")
contrast <- prolfqua::Contrasts$new(mod, Contr)

# Build count_df from config
count_df <- dplyr::select(protIntensity,
  dplyr::all_of(c(config$hierarchy_keys_depth(), "nrPeptides"))) |>
  dplyr::distinct()

deqms <- ContrastsModeratedDEqMS$new(contrast,
  count_df = count_df,
  count_column = "nrPeptides")

bb <- deqms$get_contrasts()
stopifnot(all(c("diff", "p.value", "FDR", "sigma") %in% colnames(bb)))

# Merge with LOD-imputed refits for proteins the plain lm cannot estimate
imp <- ContrastsLMImputeFacade$new(lProt, "~ group_", Contr)$contrast
merged <- merge_contrasts_results(deqms, imp)

cs <- deqms$get_contrast_sides()
cslf <- deqms$get_linfct()
ctrwide <- deqms$to_wide()
cp <- deqms$get_Plotter()
cp$volcano()

plot contrasts

Description

plot contrasts

Value

An R6 class generator.

Public fields

model_name: of column with model name
subject_id: hierarchy key columns
diff: column with fold change differences
contrast: column with contrasts names, default "contrast"
group: crosstalk group name for linked brushing, default "BB"
volcano_spec: volcano plot specification
score_spec: score plot specification
histogram_spec: plot specification
fcthresh: fold change threshold
avg.abundance: name of column containing avg abundance values.

Methods

Public methods

ContrastsPlotter$new()
ContrastsPlotter$histogram()
ContrastsPlotter$histogram_diff()
ContrastsPlotter$volcano()
ContrastsPlotter$volcano_plotly()
ContrastsPlotter$ma_plot()
ContrastsPlotter$ma_plotly()
ContrastsPlotter$score_plot()
ContrastsPlotter$barplot_threshold()
ContrastsPlotter$clone()

Method `new()`

create ContrastsPlotter

Usage

ContrastsPlotter$new(
  contrast_df,
  subject_id,
  volcano = list(list(score = "FDR", thresh = 0.1)),
  histogram = list(list(score = "p.value", xlim = c(0, 1, 0.05)), list(score = "FDR",
    xlim = c(0, 1, 0.05))),
  score = list(list(score = "statistic", thresh = NULL)),
  fcthresh = 1,
  modelName = "modelName",
  diff = "diff",
  contrast = "contrast",
  avg.abundance = "avgAbd",
  group = "BB"
)

Arguments

contrast_df: frame with contrast data
subject_id: columns containing subject Identifier
volcano: which score to plot and which ablines to add.
histogram: which scores to plot and which range (x) should be shown.
score: score parameters
fcthresh: default 1 (log2 FC threshold)
modelName: name of the colour column. Defaults to "modelName"; when left at the default and the contrast carries an estimate_type column, that column is used for colouring instead (so observed vs imputed/fallback rows stay distinguishable).
diff: fold change (difference) diff column
contrast: contrast column
avg.abundance: name of column with average abundance
group: crosstalk group name for linked brushing, default "BB"

Method `histogram()`

plot histogram of selected scores (e.g. p-value, FDR, t-statistics)

Usage

ContrastsPlotter$histogram()

Method `histogram_diff()`

plot histogram of differences (diff) fold change

Usage

ContrastsPlotter$histogram_diff(binwidth = 0.05)

Arguments

binwidth: with of bin in histogram

Method `volcano()`

volcano plots (fold change vs FDR)

Usage

ContrastsPlotter$volcano(
  colour = self$model_name,
  legend = TRUE,
  scales = c("fixed", "free", "free_x", "free_y"),
  min_score = NULL
)

Arguments

colour: column name with color information default modelName
legend: default TRUE
scales: default fixed facet_wrap, scales argument
min_score: optional lower bound for p-values or FDR values. If NULL, only exact zero and negative values are replaced with the smallest positive observed value in the same score column.

Method `volcano_plotly()`

plotly volcano plots

Usage

ContrastsPlotter$volcano_plotly(
  colour = self$model_name,
  legend = TRUE,
  scales = c("fixed", "free", "free_x", "free_y"),
  min_score = NULL
)

Arguments

colour: column in contrast matrix with colour coding
legend: default TRUE
scales: default fixed facet_wrap, scales argument
min_score: optional lower bound for p-values or FDR values. If NULL, only exact zero and negative values are replaced with the smallest positive observed value in the same score column.

Returns

list of ggplots

Method `ma_plot()`

ma plot

MA plot displays the effect size estimate as a function of the mean protein intensity across groups. Each dot represents an observed protein. Red horizontal lines represent the fold-change threshold.

Sometimes measured effects sizes (differences between samples groups) are biased by the signal intensity (here protein abundance). Such systematic effects can be explored using MA-plots.

Usage

ContrastsPlotter$ma_plot(
  fc = self$fcthresh,
  colour = self$model_name,
  legend = TRUE,
  rank = TRUE
)

Arguments

fc: fold change abline
colour: column in contrast matrix with colour coding
legend: enable legend default TRUE
rank: default FALSE, if TRUE then rank of avgAbd is used.

Returns

ggplot

Method `ma_plotly()`

ma plotly

Usage

ContrastsPlotter$ma_plotly(
  fc = self$fcthresh,
  colour = self$model_name,
  legend = TRUE,
  rank = FALSE
)

Arguments

fc: horizontal lines
colour: column in contrast matrix with colour coding.
legend: enable legend default TRUE
rank: default FALSE, if TRUE then rank of avgAbd is used.

Returns

list of ggplots

Method `score_plot()`

plot a score against the log2 fc e.g. t-statistic

Usage

ContrastsPlotter$score_plot(scorespec, colour = self$model_name, legend = TRUE)

Arguments

scorespec: list(score="statistics", fcthres = 2, thresh = 5)
colour: column with colour coding
legend: enable legend default TRUE

Returns

list of ggplots

Method `barplot_threshold()`

shows the number of significant proteins per contrasts

Usage

ContrastsPlotter$barplot_threshold()

Returns

list which contains ggplots and summary tables

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsPlotter$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_peptide_config(Nprot = 20)
model_name <- "Model"
modelFunction <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1|sample)",
   model_name = model_name)
pepIntensity <- istar$data
config <- istar$config

mod <- build_model(
  pepIntensity,
  modelFunction,
  model_name = model_name,
  subject_id = config$hierarchy_keys_depth())

 Contr <- c("group_A_vs_Ctrl" = "group_A - group_Ctrl",
 "group_B_vs_Ctrl" = "group_B - group_Ctrl"
 )
contrast <- prolfqua::Contrasts$new(mod,
  Contr)
contr <- contrast$get_contrasts()
cp <- ContrastsPlotter$new(contr,
  contrast$subject_id,
  volcano = list(list(score = "FDR", thresh = 0.1)),
  histogram = list(list(score = "p.value", xlim = c(0,1,0.05)),
                 list(score = "FDR", xlim = c(0,1,0.05))),
  score =list(list(score = "statistic",  thresh = 5))
  )

stopifnot("plotly" %in% class(cp$volcano_plotly()$FDR))
stopifnot("ggplot" %in% class(cp$score_plot(legend=FALSE)$statistic))
p <- cp$histogram()
stopifnot("ggplot" %in% class(p$FDR))
stopifnot("ggplot" %in% class(p$p.value))
res <- cp$volcano()
stopifnot("ggplot" %in% class(res$FDR))
res_plotly <- cp$volcano_plotly()
stopifnot("plotly" %in% class(res_plotly$FDR))
stopifnot("ggplot" %in%  class(cp$ma_plot()))
stopifnot("plotly" %in%   class(cp$ma_plotly(rank=TRUE)))
res  <- cp$barplot_threshold()
stopifnot("ggplot" %in% class(res$FDR$plot))
stopifnot("ggplot" %in%  class(cp$histogram_diff()))

istar <- sim_lfq_data_peptide_config(Nprot = 20)
model_name <- "Model"
modelFunction <-
  strategy_lmer("abundance  ~ group_ + (1 | peptide_Id) + (1|sample)",
   model_name = model_name)
pepIntensity <- istar$data
config <- istar$config

mod <- build_model(
  pepIntensity,
  modelFunction,
  model_name = model_name,
  subject_id = config$hierarchy_keys_depth())

 Contr <- c("group_A_vs_Ctrl" = "group_A - group_Ctrl",
 "group_B_vs_Ctrl" = "group_B - group_Ctrl"
 )
contrast <- prolfqua::Contrasts$new(mod,
  Contr)
contr <- contrast$get_contrasts()
cp <- ContrastsPlotter$new(contr,
  contrast$subject_id,
  volcano = list(list(score = "FDR", thresh = 0.1)),
  histogram = list(list(score = "p.value", xlim = c(0,1,0.05)),
                 list(score = "FDR", xlim = c(0,1,0.05))),
  score =list(list(score = "statistic",  thresh = 5))
  )

stopifnot("plotly" %in% class(cp$volcano_plotly()$FDR))
stopifnot("ggplot" %in% class(cp$score_plot(legend=FALSE)$statistic))
p <- cp$histogram()
stopifnot("ggplot" %in% class(p$FDR))
stopifnot("ggplot" %in% class(p$p.value))
res <- cp$volcano()
stopifnot("ggplot" %in% class(res$FDR))
res_plotly <- cp$volcano_plotly()
stopifnot("plotly" %in% class(res_plotly$FDR))
stopifnot("ggplot" %in%  class(cp$ma_plot()))
stopifnot("plotly" %in%   class(cp$ma_plotly(rank=TRUE)))
res  <- cp$barplot_threshold()
stopifnot("ggplot" %in% class(res$FDR$plot))
stopifnot("ggplot" %in%  class(cp$histogram_diff()))

Rfit rank-based regression contrast analysis facade

Description

Rfit rank-based regression contrast analysis facade

Details

Encapsulates the pipeline: strategy_rfit -> build_model -> Contrasts -> ContrastsModerated, using rfit rank-based linear regression. Output is the standard Wald contrast schema; the empirical-Bayes moderation shrinks the rank-based scale across proteins.

Unlike ContrastsLMFacade this backend takes no observation weights (rfit has no weights argument), so nr_children weighting is not applied.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsRfitFacade

Methods

Public methods

ContrastsRfitFacade$new()
ContrastsRfitFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsRfitFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
...: passed to strategy_rfit

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsRfitFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRfitFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRfitFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

Rfit rank-regression contrast analysis with LOD imputation facade

Description

Rfit rank-regression contrast analysis with LOD imputation facade

Details

The rank-based analogue of ContrastsLMImputeFacade. Encapsulates the pipeline: strategy_rfit -> build_model_impute -> Contrasts -> ContrastsModerated, using rfit.

Proteins whose initial rfit fit fails, is singular, or has an incomplete design are re-fitted after imputing missing values with the limit of detection (LOD). Uncertainty is borrowed from successful donor fits so it is not underestimated by the constant imputation.

Differences from ContrastsLMImputeFacade:

Covariance borrowing is vcov-only (element-wise median of the donors' full named covariance matrices). rfit exposes no lm-style cov.unscaled, so the scalar-sigma borrowing mode is not available and is not exposed.
Borrowing is fail-hard (on_misalign = "fail"): if the donor covariance matrices cannot be aligned by coefficient name, the rescue is skipped and those proteins remain in get_missing() rather than carrying an lm-specific approximation.
No observation weights (rfit has no weights argument), matching plain rfit.

modelName is the facade key "rfit_impute"; rescued rows are flagged in the estimate_type column as "lod_imputed", the same convention used by ContrastsLMImputeFacade.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsRfitImputeFacade

Methods

Public methods

ContrastsRfitImputeFacade$new()
ContrastsRfitImputeFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsRfitImputeFacade$new(
  lfqdata,
  modelstr,
  contrasts,
  lod = NULL,
  df_method = c("observed", "borrowed"),
  ...
)

Arguments

lfqdata: LFQData object (aggregated to protein level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
lod: numeric limit of detection; if NULL, auto-computed from data
df_method: "observed" uses max(n_observed - p, 1); "borrowed" uses median df from successful fits
...: passed to strategy_rfit

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsRfitImputeFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRfitImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRfitImputeFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

RLM contrast analysis facade

Description

RLM contrast analysis facade

Details

Encapsulates the pipeline: strategy_rlm -> build_model -> Contrasts -> ContrastsModerated.

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsRLMFacade

Methods

Public methods

ContrastsRLMFacade$new()
ContrastsRLMFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_Plotter()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_contrasts()
prolfqua::ContrastsFacadeBase$get_missing()
prolfqua::ContrastsFacadeBase$to_wide()

Method `new()`

initialize

Usage

ContrastsRLMFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: LFQData object
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
...: passed to strategy_rlm

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsRLMFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRLMFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_protein_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsRLMFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

ROPECA reproducibility-optimization method

Description

ROPECA reproducibility-optimization method

Details

ROPECA optimizes the reproducibility of statistical testing on peptide-level and aggregates the peptide-level changes to determine differential protein-level expression.

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsROPECA

Public fields

Contrast: Contrast
contrast_result: contrast result
model_name: model name
subject_id: columns with protein ID's
p.adjust: method to use for p.value adjustment

Methods

Public methods

ContrastsROPECA$new()
ContrastsROPECA$get_contrast_sides()
ContrastsROPECA$get_linfct()
ContrastsROPECA$get_contrasts()
ContrastsROPECA$get_Plotter()
ContrastsROPECA$to_wide()
ContrastsROPECA$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()

Method `new()`

initialize

Usage

ContrastsROPECA$new(
  Contrast,
  model_name = "ROPECA",
  p.adjust = prolfqua::adjust_p_values
)

Arguments

Contrast: e.g. instance of Contrasts class, or ContrastsModerated
model_name: default ROPECA
p.adjust: function to use for p.value adjustment

Method `get_contrast_sides()`

show names of contrasts

Usage

ContrastsROPECA$get_contrast_sides()

Returns

data.frame

Method `get_linfct()`

get linear function used to determine contrasts

Usage

ContrastsROPECA$get_linfct()

Returns

data.frame

Method `get_contrasts()`

get contrasts

Usage

ContrastsROPECA$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)
global: use a global linear function (determined by get_linfct)

Returns

data.frame

Method `get_Plotter()`

get ContrastsPlotter

Usage

ContrastsROPECA$get_Plotter(fdr_threshold = 0.1, fc_threshold = 2)

Arguments

fdr_threshold: FDR threshold
fc_threshold: FC threshold

Returns

ContrastsPlotter

Method `to_wide()`

convert to wide format

Usage

ContrastsROPECA$to_wide(
  columns = c("beta.based.significance", "FDR.beta.based.significance")
)

Arguments

columns: value column default beta.based.significance

Returns

data.frame

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsROPECA$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- prolfqua::sim_lfq_data_peptide_config(Nprot=20)
istar_data <- istar$data
modelFunction <-
  strategy_lm("abundance  ~ group_")
pepIntensity <- istar_data
config <- istar$config$clone(deep = TRUE)
config$hierarchy_depth <- 2
config$hierarchy_keys_depth()

mod <- build_model(
 pepIntensity,
 modelFunction,
 subject_id = config$hierarchy_keys_depth())

 Contr <- c("AvsCtrl" = "group_A - group_Ctrl")


 contr <- prolfqua::Contrasts$new(mod, Contr)
 dim(contr$get_contrasts())
 contrM <- prolfqua::ContrastsModerated$new(contr)
 dim(contrM$get_contrasts())
 contrast <- prolfqua::ContrastsROPECA$new(contrM)
 contrast$get_contrasts()
 contrast <- prolfqua::ContrastsROPECA$new(contr)
 tmp <- contrast$get_contrasts()
 dim(tmp)
 pl <- contrast$get_Plotter()
 contrast$to_wide()
 contrast$get_linfct()
 contrast$get_contrast_sides()
 pl$histogram()
 pl$ma_plot()

istar <- prolfqua::sim_lfq_data_peptide_config(Nprot=20)
istar_data <- istar$data
modelFunction <-
  strategy_lm("abundance  ~ group_")
pepIntensity <- istar_data
config <- istar$config$clone(deep = TRUE)
config$hierarchy_depth <- 2
config$hierarchy_keys_depth()

mod <- build_model(
 pepIntensity,
 modelFunction,
 subject_id = config$hierarchy_keys_depth())

 Contr <- c("AvsCtrl" = "group_A - group_Ctrl")


 contr <- prolfqua::Contrasts$new(mod, Contr)
 dim(contr$get_contrasts())
 contrM <- prolfqua::ContrastsModerated$new(contr)
 dim(contrM$get_contrasts())
 contrast <- prolfqua::ContrastsROPECA$new(contrM)
 contrast$get_contrasts()
 contrast <- prolfqua::ContrastsROPECA$new(contr)
 tmp <- contrast$get_contrasts()
 dim(tmp)
 pl <- contrast$get_Plotter()
 contrast$to_wide()
 contrast$get_linfct()
 contrast$get_contrast_sides()
 pl$histogram()
 pl$ma_plot()

ROPECA contrast analysis facade for nested input

Description

ROPECA contrast analysis facade for nested input

Details

Encapsulates the pipeline: strategy_lm -> build_model -> Contrasts -> ContrastsROPECA.

ROPECA fits per-peptide linear models and aggregates peptide-level p-values to the protein level. Input must be nested LFQData (peptide subject_id strictly below the protein hierarchy).

Value

An R6 class generator.

Super classes

prolfqua::ContrastsInterface -> prolfqua::ContrastsFacadeBase -> ContrastsROPECANestedFacade

Methods

Public methods

ContrastsROPECANestedFacade$new()
ContrastsROPECANestedFacade$get_contrasts()
ContrastsROPECANestedFacade$get_Plotter()
ContrastsROPECANestedFacade$to_wide()
ContrastsROPECANestedFacade$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_contrast_sides()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsFacadeBase$get_config()
prolfqua::ContrastsFacadeBase$get_missing()

Method `new()`

initialize

Usage

ContrastsROPECANestedFacade$new(lfqdata, modelstr, contrasts, ...)

Arguments

lfqdata: nested LFQData (peptide-level)
modelstr: model formula string (e.g. "~ group_")
contrasts: named character vector of contrasts
...: passed to strategy_lm

Method `get_contrasts()`

Get contrast results with standardized column names.

ROPECA's beta.based.significance is mapped to p.value and FDR.beta.based.significance to FDR.

Columns not directly produced by ROPECA are derived heuristically:

std.error = diff / statistic (algebraic: t = estimate / SE)
sigma = mad.estimate (MAD of peptide fold changes)
df = n_not_na (number of contributing peptides)
conf.low/conf.high via diff +/- qt(0.975, df) * |std.error|

Usage

ContrastsROPECANestedFacade$get_contrasts()

Method `get_Plotter()`

get ContrastsPlotter (uses standardized column names)

Usage

ContrastsROPECANestedFacade$get_Plotter(fc_threshold = 2, fdr_threshold = 0.1)

Arguments

fc_threshold: fold change threshold
fdr_threshold: FDR threshold

Method `to_wide()`

convert results to wide format

Usage

ContrastsROPECANestedFacade$to_wide(columns = c("p.value", "FDR", "statistic"))

Arguments

columns: value columns to pivot, default c("p.value", "FDR", "statistic")

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsROPECANestedFacade$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsROPECANestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()
istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
contrasts <- c("A_vs_Ctrl" = "group_A - group_Ctrl")
fa <- ContrastsROPECANestedFacade$new(lfqdata, "~ group_", contrasts)
head(fa$get_contrasts())
fa$to_wide()

holds results when contrasts are added.

Description

holds results when contrasts are added.

Value

An R6 class generator.

Super class

prolfqua::ContrastsInterface -> ContrastsTable

Public fields

contrast_result: contrast results
model_name: model name
subject_id: default protein_Id

Methods

Public methods

ContrastsTable$new()
ContrastsTable$get_contrast_sides()
ContrastsTable$get_linfct()
ContrastsTable$get_contrasts()
ContrastsTable$get_Plotter()
ContrastsTable$clone()

Inherited methods

prolfqua::ContrastsInterface$column_description()
prolfqua::ContrastsInterface$contrast_summary_table()
prolfqua::ContrastsInterface$extra_artifacts()
prolfqua::ContrastsInterface$filter_significant()
prolfqua::ContrastsInterface$get_config()
prolfqua::ContrastsInterface$get_missing()
prolfqua::ContrastsInterface$get_ora()
prolfqua::ContrastsInterface$get_rank()
prolfqua::ContrastsInterface$to_wide()

Method `new()`

initialize

Usage

ContrastsTable$new(
  contrastsdf,
  subject_id = "protein_Id",
  model_name = "ContrastTable"
)

Arguments

contrastsdf: data.frame
subject_id: default protein_Id
model_name: default ContrastTable

Method `get_contrast_sides()`

return sides of contrast

Usage

ContrastsTable$get_contrast_sides()

Returns

data.frame

Method `get_linfct()`

not implemented

Usage

ContrastsTable$get_linfct()

Method `get_contrasts()`

get contrasts

Usage

ContrastsTable$get_contrasts(all = FALSE)

Arguments

all: should all columns be returned (default FALSE)
global: use a global linear function (determined by get_linfct)

Method `get_Plotter()`

get ContrastsPlotter

Usage

ContrastsTable$get_Plotter(fc_threshold = 1, fdr_threshold = 0.1)

Arguments

fc_threshold: fold change threshold
fdr_threshold: fdr threshold

Returns

ContrastsPlotter

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ContrastsTable$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- prolfqua::sim_lfq_data_protein_config(Nprot = 30)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
Contrasts <- c("aC" = "group_A - group_Ctrl",
"bC" = "group_B - group_Ctrl")
fa <- ContrastsLMFacade$new(lfqdata, "~ group_", Contrasts)
ctr <- fa$get_contrasts()
fa$contrast$subject_id
xcx <- ContrastsTable$new(ctr, subject_id = fa$contrast$subject_id, model_name = "TableTest")
xcx$get_contrasts()
xcx$get_Plotter()$volcano()
stopifnot(is.null(xcx$get_contrast_sides()))
stopifnot(is.null(xcx$get_linfct()))
stopifnot(ncol(xcx$to_wide()) == 9)

istar <- prolfqua::sim_lfq_data_protein_config(Nprot = 30)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$rename_response("transformedIntensity")
Contrasts <- c("aC" = "group_A - group_Ctrl",
"bC" = "group_B - group_Ctrl")
fa <- ContrastsLMFacade$new(lfqdata, "~ group_", Contrasts)
ctr <- fa$get_contrasts()
fa$contrast$subject_id
xcx <- ContrastsTable$new(ctr, subject_id = fa$contrast$subject_id, model_name = "TableTest")
xcx$get_contrasts()
xcx$get_Plotter()$volcano()
stopifnot(is.null(xcx$get_contrast_sides()))
stopifnot(is.null(xcx$get_linfct()))
stopifnot(ncol(xcx$to_wide()) == 9)

Residual degrees of freedom for an augmented rfit fit

Description

Residual degrees of freedom for an augmented rfit fit

Usage

## S3 method for class 'rfit_prolfqua'
df.residual(object, ...)
## S3 method for class 'rfit_prolfqua'
df.residual(object, ...)

Arguments

object

an 'rfit_prolfqua' fit

...

ignored

Value

number of observations minus number of coefficients.

Examples

fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
df.residual(fit)
fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
df.residual(fit)

Registry of available contrast facade classes

Description

Read-only snapshot of the built-in prolfqua facade registry, taken at package load time. It is derived from the single source of truth (.seed_facade_registry()); use register_facade to add entries from downstream packages, lookup_facade to resolve a single entry, and list_facades() to enumerate the live registry (which reflects downstream registrations).

Usage

FACADE_REGISTRY
FACADE_REGISTRY

Format

An object of class facade_registry (inherits from list) of length 18.

Details

Each entry has fields class, needs, package, and needs_saint_annotation.

Value

A named list of built-in facade registry entries.

Examples

names(FACADE_REGISTRY)
FACADE_REGISTRY$limma$class
names(FACADE_REGISTRY)
FACADE_REGISTRY$limma$class

find file stored in package

Description

find file stored in package

Usage

find_package_file(packagename, file)
find_package_file(packagename, file)

Arguments

packagename

name of the R package

file

relative path to the file within the package

Value

The computed result.

Examples

find_package_file("prolfqua", "CITATION")

find_package_file("prolfqua", "CITATION")

Combined generate_contrasts

Description

Combined generate_contrasts

Usage

generate_contrasts(primary_levels, secondary_levels, interactions = TRUE)
generate_contrasts(primary_levels, secondary_levels, interactions = TRUE)

Arguments

primary_levels

character vector of primary factor levels

secondary_levels

character vector of secondary factor levels

interactions

logical, if TRUE include interaction contrasts

Value

Contrast definitions or contrast results.

Examples

primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
generate_contrasts(primary_levels, secondary_levels)
generate_contrasts(secondary_levels, primary_levels)
primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
generate_contrasts(primary_levels, secondary_levels)
generate_contrasts(secondary_levels, primary_levels)

Single-factor contrasts (pairwise comparisons)

Description

Single-factor contrasts (pairwise comparisons)

Usage

generate_contrasts_for_factor(levels)
generate_contrasts_for_factor(levels)

Arguments

levels

character vector of factor levels

Value

Contrast definitions or contrast results.

Examples

# example code
primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
x2 <- level_specific_contrasts(primary_levels, secondary_levels)
x3 <- level_specific_contrasts(secondary_levels, primary_levels)
generate_contrasts_for_factor(names(x2))
# example code
primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
x2 <- level_specific_contrasts(primary_levels, secondary_levels)
x3 <- level_specific_contrasts(secondary_levels, primary_levels)
generate_contrasts_for_factor(names(x2))

Extracts uniprot ID

Description

Extracts uniprot ID

Usage

get_uniprot_id_from_fasta_header(df, idcolumn = "protein_Id")
get_uniprot_id_from_fasta_header(df, idcolumn = "protein_Id")

Arguments

df

data.frame

idcolumn

character, column name to extract uniprot IDs from

Value

data.frame

Examples


df <- data.frame(
  protein_Id = c("sp|P12345|PROT1_HUMAN", "sp|Q67890|PROT2_HUMAN",
                 "tr|A0A0A0|PROT3_HUMAN", "NOFASTA_ID"),
  value = 1:4
)
result <- get_uniprot_id_from_fasta_header(df, idcolumn = "protein_Id")
stopifnot("UniprotID" %in% colnames(result))
stopifnot(sum(!is.na(result$UniprotID)) == 3)

df <- data.frame(
  protein_Id = c("sp|P12345|PROT1_HUMAN", "sp|Q67890|PROT2_HUMAN",
                 "tr|A0A0A0|PROT3_HUMAN", "NOFASTA_ID"),
  value = 1:4
)
result <- get_uniprot_id_from_fasta_header(df, idcolumn = "protein_Id")
stopifnot("UniprotID" %in% colnames(result))
stopifnot(sum(!is.na(result$UniprotID)) == 3)

group label function

Description

group label function

Usage

group_label(primary, secondary)
group_label(primary, secondary)

Arguments

primary

character, primary factor level

secondary

character, secondary factor level

Value

A character vector with combined group labels.

Examples

group_label("a", "b")
group_label("a", "b")

Count distinct elements for each level of hierarchy per sample

Description

Count distinct elements for each level of hierarchy per sample

Details

Provides wide, long, and plot views of hierarchy counts.

Value

An R6 class generator.

Public fields

.summary: summarised data frame
.configuration: AnalysisConfiguration object

Methods

Public methods

HierarchyCountsSample$new()
HierarchyCountsSample$wide()
HierarchyCountsSample$long()
HierarchyCountsSample$plot()
HierarchyCountsSample$clone()

Method `new()`

Create a new HierarchyCountsSample

Usage

HierarchyCountsSample$new(pdata, configuration, nr_children = 1)

Arguments

pdata: data frame
configuration: AnalysisConfiguration
nr_children: minimum number of children to include

Method `wide()`

Return wide-format summary

Usage

HierarchyCountsSample$wide()

Method `long()`

Return long-format summary

Usage

HierarchyCountsSample$long()

Method `plot()`

Return barplot of hierarchy counts

Usage

HierarchyCountsSample$plot()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

HierarchyCountsSample$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

bb <- sim_lfq_data_protein_config()
counts <- HierarchyCountsSample$new(bb$data, bb$config)
head(counts$wide())
bb <- sim_lfq_data_protein_config()
counts <- HierarchyCountsSample$new(bb$data, bb$config)
head(counts$wide())

Fill missing responses with predictions from per-subject linear models

Description

Completes the data to every subject x every sample and fills each missing response cell with predict() of that subject's fitted lm, on the response scale. Observed values are left unchanged. The imputed values carry no noise.

Usage

impute_from_model(model, lfqdata)
impute_from_model(model, lfqdata)

Arguments

model

a Model with lm fits, as returned by build_model or build_model_impute

lfqdata

the LFQData the model was fitted on (same response and subject_id)

Details

Each subject gets one route, checked in this order:

complete: no missing cell.
lod_refit: the subject was refitted at the LOD by build_model_impute (model_df$imputed); that fit's predictions fill the missing cells.
fitted: the observed-data fit has no NA coefficient and knows every factor level of the missing cells; its predictions fill them.
none: no usable fit; the missing cells stay NA.

Value

list with lfqdata, a new LFQData holding every subject x sample row with missing responses filled, and summary, a tibble with the subject_id columns and n_observed, n_imputed and route per subject

Examples

istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
strat <- strategy_lm(paste(lfqdata$response(), "~ group_"))
mod <- build_model_impute(lfqdata, strat)
res <- impute_from_model(mod, lfqdata)
table(res$summary$route)
n_missing <- function(x) sum(is.na(x$data_long()$abundance))
stopifnot(n_missing(res$lfqdata) <= n_missing(lfqdata))
istar <- sim_lfq_data_protein_config(Nprot = 30, weight_missing = 0.5)
lfqdata <- LFQData$new(istar$data, istar$config)
strat <- strategy_lm(paste(lfqdata$response(), "~ group_"))
mod <- build_model_impute(lfqdata, strat)
res <- impute_from_model(mod, lfqdata)
table(res$summary$route)
n_missing <- function(x) sum(is.na(x$data_long()$abundance))
stopifnot(n_missing(res$lfqdata) <= n_missing(lfqdata))

Interaction contrasts (difference of differences)

Description

Interaction contrasts (difference of differences)

Usage

interaction_contrasts(primary_levels, secondary_levels)
interaction_contrasts(primary_levels, secondary_levels)

Arguments

primary_levels

character vector of primary factor levels

secondary_levels

character vector of secondary factor levels

Value

Contrast definitions or contrast results.

Examples

primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
interaction_contrasts(primary_levels, secondary_levels)
interaction_contrasts(secondary_levels, primary_levels)
primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
interaction_contrasts(primary_levels, secondary_levels)
interaction_contrasts(secondary_levels, primary_levels)

Internal Functions by category

Description

Internal Functions by category

Value

A documentation topic.

Detect contaminant identifiers

Description

As is_decoy, but for contaminant entries (keratin, trypsin, BSA, ...). Built-in anchored default prefixes unioned with an optional configured pattern; empty / NULL uses the defaults only.

Usage

is_contaminant(ids, pattern = NULL)
is_contaminant(ids, pattern = NULL)

Arguments

ids

character vector of (prefixed) identifiers

pattern

optional additional contaminant regex, unioned with defaults

Value

logical vector, TRUE where an id looks like a contaminant

Examples

is_contaminant(c("zz|Cont00001|X", "sp|P2|X", "CON__ALBU"))
is_contaminant(c("zz|Cont00001|X", "sp|P2|X", "CON__ALBU"))

Detect decoy / reverse-database identifiers

Description

Flags identifiers that look like decoy (reversed) database entries. Built-in anchored default prefixes are always considered, unioned with an optional configured pattern. An empty string or NULL pattern uses the defaults only.

Usage

is_decoy(ids, pattern = NULL)
is_decoy(ids, pattern = NULL)

Arguments

ids

character vector of (prefixed) identifiers

pattern

optional additional decoy regex, unioned with the defaults

Value

logical vector, TRUE where an id looks like a decoy

Examples

is_decoy(c("REV_sp|P1|X", "sp|P2|X", "decoy_3", "normalProtein"))
is_decoy(c("shuffled_1", "P2"), pattern = "^shuffled_")
is_decoy(c("REV_sp|P1|X", "sp|P2|X", "decoy_3", "normalProtein"))
is_decoy(c("shuffled_1", "P2"), pattern = "^shuffled_")

Level-specific contrasts (per secondary level)

Description

Level-specific contrasts (per secondary level)

Usage

level_specific_contrasts(primary_levels, secondary_levels)
level_specific_contrasts(primary_levels, secondary_levels)

Arguments

primary_levels

character vector of primary factor levels

secondary_levels

character vector of secondary factor levels

Value

Contrast definitions or contrast results.

Examples

# example code

primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
x2 <- level_specific_contrasts(primary_levels, secondary_levels)
x3 <- level_specific_contrasts(secondary_levels, primary_levels)
# example code

primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T0", "T150", "T300")
x2 <- level_specific_contrasts(primary_levels, secondary_levels)
x3 <- level_specific_contrasts(secondary_levels, primary_levels)

LFQData R6 class

Description

LFQData R6 class

Value

An R6 class generator.

Missing Data Assumptions

The filtering and imputation methods in this package assume that missing values are Missing Completely At Random (MCAR) or Missing At Random (MAR). Abundance-dependent missingness (MNAR), which is common in DDA proteomics, is not modelled. Users should be aware that MNAR can bias fold-change estimates and inflate false discovery rates.

Public fields

prefix: e.g. "peptide_", "protein_", "compound_"

Methods

Public methods

LFQData$new()
LFQData$set_data()
LFQData$get_config()
LFQData$set_config_value()
LFQData$get_copy()
LFQData$get_sample()
LFQData$get_subset()
LFQData$subject_id()
LFQData$is_transformed()
LFQData$remove_small_intensities()
LFQData$filter_proteins_by_peptide_count()
LFQData$remove_decoys()
LFQData$decoy_proportion()
LFQData$complete_cases()
LFQData$data_wide()
LFQData$factors()
LFQData$hierarchy()
LFQData$response()
LFQData$hierarchy_keys()
LFQData$relevant_hierarchy_keys()
LFQData$factor_keys()
LFQData$relevant_factor_keys()
LFQData$sample_name()
LFQData$file_name()
LFQData$nr_children_col()
LFQData$isotope_label()
LFQData$data_long()
LFQData$rename_response()
LFQData$hierarchy_counts()
LFQData$summarize_hierarchy()
LFQData$get_Plotter()
LFQData$get_Summariser()
LFQData$get_Stats()
LFQData$get_Transformer()
LFQData$get_Aggregator()
LFQData$clone()

Method `new()`

initialize

Usage

LFQData$new(data, config, prefix = "ms_", setup = FALSE)

Arguments

data: data.frame
config: configuration
prefix: will be use as output prefix
setup: is data setup needed, default = FALSE, if TRUE, calls setup_analysis on data first.

Method `set_data()`

set data (replaces the internal data frame)

Usage

LFQData$set_data(new_data)

Arguments

new_data: data.frame

Returns

self (invisible)

Method `get_config()`

return the AnalysisConfiguration object

Usage

LFQData$get_config()

Returns

AnalysisConfiguration

Method `set_config_value()`

set a config field value

Usage

LFQData$set_config_value(field, value)

Arguments

field: character — field name
value: the value to set

Method `get_copy()`

get deep copy

Usage

LFQData$get_copy()

Method `get_sample()`

samples subset of data

Usage

LFQData$get_sample(size = 100, seed = NULL)

Arguments

size: size of subset default 100
seed: set seed

Method `get_subset()`

get subset of data

Usage

LFQData$get_subset(x)

Arguments

x: data frame with columns containing subject_id

Method `subject_id()`

get subject ID columns

Usage

LFQData$subject_id()

Method `is_transformed()`

is data transformed

Usage

LFQData$is_transformed(is_transformed)

Arguments

is_transformed: logical

Returns

logical

Method `remove_small_intensities()`

some software is reporting NA's as 0, you must remove it from your data

Usage

LFQData$remove_small_intensities(threshold = 4)

Arguments

threshold: default 4.

Returns

self

Method `filter_proteins_by_peptide_count()`

remove proteins with less than X peptides

Usage

LFQData$filter_proteins_by_peptide_count()

Returns

self

Method `remove_decoys()`

remove decoy / reverse-database entries (rows whose top-level hierarchy id, e.g. protein_Id, is a decoy). Detection uses the configured 'pattern_decoys' unioned with the built-in defaults (see is_decoy); decoys are always detectable even when no pattern is configured. Returns a new decoy-free LFQData (self is not modified).

Usage

LFQData$remove_decoys()

Returns

LFQData without decoys

Method `decoy_proportion()`

proportion of modelling-level keys (subject_id, e.g. protein or peptide by 'hierarchy_depth') that are decoys — an empirical-FDR / target-decoy QC signal. Decoy status derives from the top-level hierarchy id (protein_Id).

Usage

LFQData$decoy_proportion()

Returns

numeric in [0, 1]

Method `complete_cases()`

some software is reporting NA's as 0, you must remove it from your data

Usage

LFQData$complete_cases()

Arguments

threshold: default 4.

Returns

self

Method `data_wide()`

converts the data to wide

Usage

LFQData$data_wide(as.matrix = FALSE, value = NULL)

Arguments

as.matrix: return as data.frame or matrix
value: see possible lfqdata$get_config()$value_vars()

Returns

list with data, annotation, and configuration

Method `factors()`

Annotation table

Usage

LFQData$factors()

Returns

data.frame

Method `hierarchy()`

Hierarchy table

Usage

LFQData$hierarchy()

Method `response()`

name of response variable

Usage

LFQData$response()

Returns

data.frame

Method `hierarchy_keys()`

return all hierarchy column names

Usage

LFQData$hierarchy_keys()

Method `relevant_hierarchy_keys()`

return hierarchy column names at current depth (alias for subject_id)

Usage

LFQData$relevant_hierarchy_keys()

Method `factor_keys()`

return all factor column names

Usage

LFQData$factor_keys()

Method `relevant_factor_keys()`

return factor column names at current depth

Usage

LFQData$relevant_factor_keys()

Method `sample_name()`

return sample name column

Usage

LFQData$sample_name()

Method `file_name()`

return file name column

Usage

LFQData$file_name()

Method `nr_children_col()`

return name of nr_children column

Usage

LFQData$nr_children_col()

Method `isotope_label()`

return isotope label column name

Usage

LFQData$isotope_label()

Method `data_long()`

return the tidy (long-format) data frame

Usage

LFQData$data_long(na.omit = FALSE)

Arguments

na.omit: if TRUE, remove rows with NA in response column

Method `rename_response()`

new name of response variable

Usage

LFQData$rename_response(newname = "Intensity")

Arguments

newname: default Intensity

Method `hierarchy_counts()`

number of elements at each level

Usage

LFQData$hierarchy_counts()

Method `summarize_hierarchy()`

e.g. number of peptides per protein etc

Usage

LFQData$summarize_hierarchy()

Returns

data.frame

Method `get_Plotter()`

get Plotter

Usage

LFQData$get_Plotter()

Returns

LFQDataPlotter

Method `get_Summariser()`

get Summariser

Usage

LFQData$get_Summariser()

Returns

LFQDataSummarizer

Method `get_Stats()`

Get LFQDataStats. For more details see LFQDataStats.

Usage

LFQData$get_Stats(stats = c("everything", "interaction", "all"))

Arguments

stats: default interaction, computes statistics within interaction.

Returns

LFQDataStats

Method `get_Transformer()`

get Stats

Usage

LFQData$get_Transformer()

Returns

LFQDataTransformer

Method `get_Aggregator()`

get Aggregator

Usage

LFQData$get_Aggregator(method = "medpolish", ...)

Arguments

method: aggregation method: "medpolish", "rlm", or "topN"
...: passed to aggregator constructor (e.g. prefix, N, func)

Returns

AggregateMedpolish, AggregateRlm, or AggregateTopN

Method `clone()`

The objects of this class are cloneable with this method.

Usage

LFQData$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$filter_proteins_by_peptide_count()
tmp <- lfqdata$data_wide()
testthat::expect_equal(nrow(tmp$data) , nrow(tmp$rowdata))
testthat::expect_equal(ncol(tmp$data) , nrow(tmp$annotation) + ncol(tmp$rowdata))

stopifnot("data.frame" %in% class(tmp$data))
tmp <- lfqdata$data_wide(as.matrix = TRUE)
stopifnot("matrix" %in% class(tmp$data))
stopifnot(lfqdata$is_transformed()==FALSE)
lfqdata$summarize_hierarchy()

lfqdata$response()
lfqdata$rename_response("peptide.intensity")
lfqdata$response()
stopifnot("LFQData" %in% class(lfqdata$get_copy()))
stopifnot("LFQDataTransformer" %in% class(lfqdata$get_Transformer()))
stopifnot("LFQDataStats" %in% class(lfqdata$get_Stats()))
stopifnot("LFQDataSummariser" %in% class(lfqdata$get_Summariser()))
stopifnot("LFQDataPlotter" %in% class(lfqdata$get_Plotter()))
stopifnot("AggregateMedpolish" %in% class(lfqdata$get_Aggregator("medpolish")))

tmp <-lfqdata$get_sample(5, seed = 4)
stopifnot(nrow(tmp$hierarchy()) == 5)

istar <- sim_lfq_data_peptide_config()
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata$filter_proteins_by_peptide_count()
tmp <- lfqdata$data_wide()
testthat::expect_equal(nrow(tmp$data) , nrow(tmp$rowdata))
testthat::expect_equal(ncol(tmp$data) , nrow(tmp$annotation) + ncol(tmp$rowdata))

stopifnot("data.frame" %in% class(tmp$data))
tmp <- lfqdata$data_wide(as.matrix = TRUE)
stopifnot("matrix" %in% class(tmp$data))
stopifnot(lfqdata$is_transformed()==FALSE)
lfqdata$summarize_hierarchy()

lfqdata$response()
lfqdata$rename_response("peptide.intensity")
lfqdata$response()
stopifnot("LFQData" %in% class(lfqdata$get_copy()))
stopifnot("LFQDataTransformer" %in% class(lfqdata$get_Transformer()))
stopifnot("LFQDataStats" %in% class(lfqdata$get_Stats()))
stopifnot("LFQDataSummariser" %in% class(lfqdata$get_Summariser()))
stopifnot("LFQDataPlotter" %in% class(lfqdata$get_Plotter()))
stopifnot("AggregateMedpolish" %in% class(lfqdata$get_Aggregator("medpolish")))

tmp <-lfqdata$get_sample(5, seed = 4)
stopifnot(nrow(tmp$hierarchy()) == 5)

LFQDataPlotter —- Create various visualization of the LFQdata

Description

LFQDataPlotter —- Create various visualization of the LFQdata

Value

An R6 class generator.

Public fields

lfq: LFQData object
prefix: prefix to figure names when writing, e.g. protein_

Methods

Public methods

LFQDataPlotter$new()
LFQDataPlotter$raster()
LFQDataPlotter$heatmap()
LFQDataPlotter$heatmap_cor()
LFQDataPlotter$pca()
LFQDataPlotter$pca_plotly()
LFQDataPlotter$boxplots()
LFQDataPlotter$missigness_histogram()
LFQDataPlotter$na_heatmap()
LFQDataPlotter$intensity_distribution_density()
LFQDataPlotter$intensity_distribution_violin()
LFQDataPlotter$pairs_smooth()
LFQDataPlotter$sample_correlation()
LFQDataPlotter$upset_missing()
LFQDataPlotter$write_boxplots()
LFQDataPlotter$write_pdf()
LFQDataPlotter$clone()

Method `new()`

create LFQDataPlotter

Usage

LFQDataPlotter$new(lfqdata, prefix = "ms_")

Arguments

lfqdata: LFQData
prefix: will be prepended to outputs written

Method `raster()`

plot intensities in raster

Usage

LFQDataPlotter$raster(
  arrange = c("mean", "var"),
  not_na = FALSE,
  rownames = FALSE,
  max_rownames_chars = 60,
  max_sample_label_chars = 20
)

Arguments

arrange: arrange by either mean or var
not_na: TRUE arrange by number of NA's, FALSE by arrange by intensity
rownames: show rownames (default FALSE - do not show.)
max_rownames_chars: maximum displayed row label length
max_sample_label_chars: maximum displayed sample label length. Labels keep their suffix because sample prefixes are often shared.

Returns

ComplexHeatmap::Heatmap

Method `heatmap()`

heatmap of intensities - columns are samples, rows are proteins or peptides.

The abundances of each protein (row) are z-scored. Afterward, the mean abundance for each protein is zero, and the standard variation is one. z-scoring allows to compare (cluster) the proteins according to the difference in the expression in the samples. Without the z-scoring, the proteins would group according to their abundance, e.g., high abundant proteins would be one cluster.

Only the top_n most variable features are shown. Row clustering uses stats::hclust, which errors above 65536 features, so the rows are ranked by their variability statistic (CV for untransformed data, sd for transformed data; see LFQDataStats) and the most variable are kept. This keeps the heatmap feasible and legible for peptide-list / entrapment searches with tens of thousands of features.

Usage

LFQDataPlotter$heatmap(
  na_fraction = 1,
  rownames = FALSE,
  max_rownames_chars = 60,
  max_sample_label_chars = 20,
  top_n = 1000
)

Arguments

na_fraction: maximum fraction of missing values per row. The default keeps every row that can be meaningfully z-scored.
rownames: show rownames (default FALSE - do not show.)
max_rownames_chars: maximum displayed row label length
max_sample_label_chars: maximum displayed sample label length. Labels keep their suffix because sample prefixes are often shared.
top_n: keep the top_n most variable features (default 1000); NULL or Inf keeps all features.

Returns

ComplexHeatmap::Heatmap

Method `heatmap_cor()`

heatmap of sample correlations.

The Spearman correlation among all samples is computed. Then the euclidean distance is used to compute the distances.

Usage

LFQDataPlotter$heatmap_cor(max_sample_label_chars = 20)

Arguments

max_sample_label_chars: maximum displayed sample label length. Labels keep their suffix because sample prefixes are often shared.

Returns

ComplexHeatmap::Heatmap

Method `pca()`

PCA plot

A PCA is applied and the first and second principal component are shown. Features with missing values are removed. To keep all features, impute first, e.g. with AggregateLimpa in impute_only mode.

Usage

LFQDataPlotter$pca(PC = c(1, 2), add_txt = TRUE, nudge = 0.1)

Arguments

PC: default c(1,2) - first and second principal component
add_txt: show sample names
nudge: default 0.1 nudge point labels

Returns

ggplot

Method `pca_plotly()`

pca plot

Usage

LFQDataPlotter$pca_plotly(PC = c(1, 2), add_txt = FALSE)

Arguments

PC: default c(1,2) - first and second principal component
add_txt: show sample names

Returns

plotly

Method `boxplots()`

boxplots for all proteins

Usage

LFQDataPlotter$boxplots(facet = TRUE)

Arguments

facet: enable facet wrap if hierarchy_depth less then hierarchy length.

Returns

tibble with column boxplots containing ggplot objects

Method `missigness_histogram()`

histogram of intensities given number of missing in conditions

Usage

LFQDataPlotter$missigness_histogram()

Returns

ggplot

Method `na_heatmap()`

heatmap of features with missing values

Usage

LFQDataPlotter$na_heatmap()

Returns

ComplexHeatmap::Heatmap

Method `intensity_distribution_density()`

density distribution of intensities

Usage

LFQDataPlotter$intensity_distribution_density(
  legend = NA,
  max_legend_samples = 16
)

Arguments

legend: show legend TRUE, FALSE do not show, NA selects automatically based on sample count.
max_legend_samples: maximum number of samples for automatic legend display.

Returns

ggplot

Method `intensity_distribution_violin()`

Violinplot showing distribution of intensities in all samples

Usage

LFQDataPlotter$intensity_distribution_violin()

Returns

ggplot

Method `pairs_smooth()`

pairsplot of intensities

Usage

LFQDataPlotter$pairs_smooth(max = 10)

Arguments

max: maximal number of samples to show

Returns

NULL

Method `sample_correlation()`

plot of sample correlations

Usage

LFQDataPlotter$sample_correlation()

Returns

NULL

Method `upset_missing()`

upset plot based on presence absence information

Usage

LFQDataPlotter$upset_missing()

Returns

plot

Method `write_boxplots()`

write boxplots to file

Usage

LFQDataPlotter$write_boxplots(path_qc, filename = NULL, width = 6, height = 6)

Arguments

path_qc: path to write to
filename: file to write into
width: fig width
height: fig height

Method `write_pdf()`

write figure to pdf

Usage

LFQDataPlotter$write_pdf(fig, path_qc, fig_name, width = 7, height = 7)

Arguments

fig: ggplot or ComplexHeatmap::Heatmap
path_qc: path to write to
fig_name: name of figure (no extension)
width: figure width
height: figure height

Returns

path the file was written to

Method `clone()`

The objects of this class are cloneable with this method.

Usage

LFQDataPlotter$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_peptide_config()

lfqdata <- LFQData$new(
 istar$data,
 istar$config)
lfqplotter <- lfqdata$get_Plotter()

stopifnot(methods::is(lfqplotter$heatmap(), "Heatmap"))
stopifnot(methods::is(lfqplotter$heatmap_cor(), "Heatmap"))
stopifnot("ggplot" %in% class(lfqplotter$pca()))
stopifnot("plotly" %in%  class(lfqplotter$pca_plotly()))
tmp <- lfqplotter$boxplots()
stopifnot("ggplot" %in%  class(tmp$boxplot[[1]]))
stopifnot("ggplot" %in% class(lfqplotter$missigness_histogram()))

stopifnot(methods::is(lfqplotter$na_heatmap(), "Heatmap"))
class(lfqplotter$intensity_distribution_density())
class(lfqplotter$intensity_distribution_violin())
stopifnot(is.null(lfqplotter$pairs_smooth()))
stopifnot(class(lfqplotter$sample_correlation()) == "list")
stopifnot(methods::is(lfqplotter$raster(), "Heatmap"))
stopifnot("upset" == class(lfqplotter$upset_missing()))
wide <- lfqdata$data_wide(as.matrix = TRUE)
stopifnot(class(prolfqua::plot_sample_correlation(wide$data)) == "list")


istar <- sim_lfq_data_peptide_config()

lfqdata <- LFQData$new(
 istar$data,
 istar$config)
lfqplotter <- lfqdata$get_Plotter()

stopifnot(methods::is(lfqplotter$heatmap(), "Heatmap"))
stopifnot(methods::is(lfqplotter$heatmap_cor(), "Heatmap"))
stopifnot("ggplot" %in% class(lfqplotter$pca()))
stopifnot("plotly" %in%  class(lfqplotter$pca_plotly()))
tmp <- lfqplotter$boxplots()
stopifnot("ggplot" %in%  class(tmp$boxplot[[1]]))
stopifnot("ggplot" %in% class(lfqplotter$missigness_histogram()))

stopifnot(methods::is(lfqplotter$na_heatmap(), "Heatmap"))
class(lfqplotter$intensity_distribution_density())
class(lfqplotter$intensity_distribution_violin())
stopifnot(is.null(lfqplotter$pairs_smooth()))
stopifnot(class(lfqplotter$sample_correlation()) == "list")
stopifnot(methods::is(lfqplotter$raster(), "Heatmap"))
stopifnot("upset" == class(lfqplotter$upset_missing()))
wide <- lfqdata$data_wide(as.matrix = TRUE)
stopifnot(class(prolfqua::plot_sample_correlation(wide$data)) == "list")

Decorates LFQData with methods to compute statistics of interactions

Description

Decorates LFQData with methods to compute statistics of interactions

Details

compute stdv, mean and CV per peptide or protein and condition.

Value

An R6 class generator.

Public fields

lfq: LFQData
stat: either CV or sd (if is_transformed)
statsdf: frame with statistics.

Methods

Public methods

LFQDataStats$new()
LFQDataStats$stats()
LFQDataStats$stats_wide()
LFQDataStats$stats_quantiles()
LFQDataStats$density()
LFQDataStats$density_median()
LFQDataStats$violin()
LFQDataStats$violin_median()
LFQDataStats$stdv_vs_mean()
LFQDataStats$power_t_test_quantiles()
LFQDataStats$power_t_test()
LFQDataStats$clone()

Method `new()`

create analyse variances and CV

Usage

LFQDataStats$new(lfqdata, stats = c("everything", "interaction", "all"))

Arguments

lfqdata: LFQData object
stats: if 'interaction', compute within-group stats; if 'all', compute overall CV; if 'pooled', use pooled variance with grouping information

Method `stats()`

access data.frame with statistics

Usage

LFQDataStats$stats()

Returns

data.frame with computed statistics

Method `stats_wide()`

access data.frame with statistics in wide format

Usage

LFQDataStats$stats_wide()

Returns

data.frame with computed statistics in wide format

Method `stats_quantiles()`

Determine CV or sd for the quantiles

Usage

LFQDataStats$stats_quantiles(probs = c(0.1, 0.25, 0.5, 0.75, 0.9))

Arguments

probs: for which quantile to determine CV or sd

Method `density()`

plots density or ecdf

Usage

LFQDataStats$density(ggstat = c("density", "ecdf"))

Arguments

ggstat: either density or ecdf

Returns

ggplot

Method `density_median()`

plot density or ecdf of CV or sd for the 50

Usage

LFQDataStats$density_median(ggstat = c("density", "ecdf"))

Arguments

ggstat: either density of ecdf

Returns

ggplot

Method `violin()`

plot violinplot of CV or sd

Usage

LFQDataStats$violin()

Arguments

ggstat: either density of ecdf

Returns

ggplot

Method `violin_median()`

plot violinplot of CV or sd for the 50

Usage

LFQDataStats$violin_median()

Returns

ggplot

Method `stdv_vs_mean()`

plot sd vs mean

Usage

LFQDataStats$stdv_vs_mean(size = 200)

Arguments

size: number of points to sample (default 200)

Returns

ggplot

Method `power_t_test_quantiles()`

compute sample size for entire dataset

Usage

LFQDataStats$power_t_test_quantiles(
  probs = c(0.1, 0.25, 0.5, 0.75, 0.9),
  delta = c(0.59, 1, 2),
  power = 0.8,
  sig.level = 0.05
)

Arguments

probs: quantiles of sd for which sample size should be computed
delta: effect size
power: power of test
sig.level: significance level.

Method `power_t_test()`

compute sample for each protein

Usage

LFQDataStats$power_t_test(delta = c(0.59, 1, 2), power = 0.8, sig.level = 0.05)

Arguments

delta: effect size
power: power of test
sig.level: significance level.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

LFQDataStats$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


# study variance of not normalized data
#source("c:/Users/wewol/prog/prolfqua/R/LFQData.R")
runallfuncs <- function(x){

  stopifnot("data.frame" %in% class(x$stats()))
  stopifnot("data.frame" %in% class(x$stats_wide()))
  stopifnot(c("long", "wide") %in% names(x$stats_quantiles()))
  stopifnot("ggplot" %in% class(x$density()))
  stopifnot("ggplot" %in% class(x$density_median()))
  stopifnot("ggplot" %in% class(x$density("ecdf")))
  stopifnot("ggplot" %in% class(x$density_median("ecdf")))
  stopifnot("ggplot" %in% class(x$violin()))
  stopifnot("ggplot" %in% class(x$violin_median()))
  stopifnot("ggplot" %in% class(x$stdv_vs_mean(size = 400)))
  if(!x$lfq$is_transformed()){
    stopifnot(is.null(x$power_t_test()))
    stopifnot(is.null(x$power_t_test_quantiles()))
  }
}
bb <- prolfqua::sim_lfq_data_peptide_config()



lfqdata <- LFQData$new(bb$data, bb$config)
lfqstats <- lfqdata$get_Stats()
stopifnot(ncol(lfqstats$stats_wide()) == 30)
lfqstats$violin()
runallfuncs(lfqstats)
x <- lfqstats

#study variance of normalized data


lfqdata <- LFQData$new(bb$data, bb$config)
lfqdata$is_transformed(TRUE)
lfqstats <- lfqdata$get_Stats()
stopifnot(ncol(lfqstats$stats_wide()) == 26)
runallfuncs(lfqstats)

#Slightly different dataset


# estimates statistics for all samples
lfqstats <- lfqdata$get_Stats(stats = "all")
stopifnot(ncol(lfqstats$stats_wide()) == 8)
runallfuncs(lfqstats)
lfqstats <- lfqdata$get_Stats(stats = "interaction")
stopifnot(ncol(lfqstats$stats_wide()) == 20)
runallfuncs(lfqstats)

# Group size 1
bb <- prolfqua::sim_lfq_data_peptide_config(N=1)
lfq1 <- LFQData$new(bb$data, bb$config)
table_factors_size(lfq1$data_long(), lfq1$file_name(),
  lfq1$sample_name(), lfq1$factor_keys(), lfq1$relevant_factor_keys())
lfqdata <- LFQData$new(bb$data, bb$config)
lfqstats <- lfqdata$get_Stats()

# stopifnot(ncol(lfqstats$stats_wide()) == 30)
lfqstats$violin()
runallfuncs(lfqstats)
# study variance of not normalized data
#source("c:/Users/wewol/prog/prolfqua/R/LFQData.R")
runallfuncs <- function(x){

  stopifnot("data.frame" %in% class(x$stats()))
  stopifnot("data.frame" %in% class(x$stats_wide()))
  stopifnot(c("long", "wide") %in% names(x$stats_quantiles()))
  stopifnot("ggplot" %in% class(x$density()))
  stopifnot("ggplot" %in% class(x$density_median()))
  stopifnot("ggplot" %in% class(x$density("ecdf")))
  stopifnot("ggplot" %in% class(x$density_median("ecdf")))
  stopifnot("ggplot" %in% class(x$violin()))
  stopifnot("ggplot" %in% class(x$violin_median()))
  stopifnot("ggplot" %in% class(x$stdv_vs_mean(size = 400)))
  if(!x$lfq$is_transformed()){
    stopifnot(is.null(x$power_t_test()))
    stopifnot(is.null(x$power_t_test_quantiles()))
  }
}
bb <- prolfqua::sim_lfq_data_peptide_config()



lfqdata <- LFQData$new(bb$data, bb$config)
lfqstats <- lfqdata$get_Stats()
stopifnot(ncol(lfqstats$stats_wide()) == 30)
lfqstats$violin()
runallfuncs(lfqstats)
x <- lfqstats

#study variance of normalized data


lfqdata <- LFQData$new(bb$data, bb$config)
lfqdata$is_transformed(TRUE)
lfqstats <- lfqdata$get_Stats()
stopifnot(ncol(lfqstats$stats_wide()) == 26)
runallfuncs(lfqstats)

#Slightly different dataset


# estimates statistics for all samples
lfqstats <- lfqdata$get_Stats(stats = "all")
stopifnot(ncol(lfqstats$stats_wide()) == 8)
runallfuncs(lfqstats)
lfqstats <- lfqdata$get_Stats(stats = "interaction")
stopifnot(ncol(lfqstats$stats_wide()) == 20)
runallfuncs(lfqstats)

# Group size 1
bb <- prolfqua::sim_lfq_data_peptide_config(N=1)
lfq1 <- LFQData$new(bb$data, bb$config)
table_factors_size(lfq1$data_long(), lfq1$file_name(),
  lfq1$sample_name(), lfq1$factor_keys(), lfq1$relevant_factor_keys())
lfqdata <- LFQData$new(bb$data, bb$config)
lfqstats <- lfqdata$get_Stats()

# stopifnot(ncol(lfqstats$stats_wide()) == 30)
lfqstats$violin()
runallfuncs(lfqstats)

Summarize LFQData

Description

Summarize LFQData

Value

An R6 class generator.

Public fields

lfq: LFQData

Methods

Public methods

LFQDataSummariser$new()
LFQDataSummariser$hierarchy_counts()
LFQDataSummariser$hierarchy_counts_sample()
LFQDataSummariser$plot_hierarchy_counts_sample()
LFQDataSummariser$interaction_missing_stats()
LFQDataSummariser$upset_interaction_missing_stats()
LFQDataSummariser$missingness_per_group()
LFQDataSummariser$missingness_per_group_cumsum()
LFQDataSummariser$plot_missingness_per_group()
LFQDataSummariser$plot_missingness_per_group_cumsum()
LFQDataSummariser$percentage_abundance()
LFQDataSummariser$clone()

Method `new()`

initialize

Usage

LFQDataSummariser$new(lfqdata)

Arguments

lfqdata: LFQData

Method `hierarchy_counts()`

summarize hierarchy

Usage

LFQDataSummariser$hierarchy_counts()

Method `hierarchy_counts_sample()`

number of elements at each level in every sample

Usage

LFQDataSummariser$hierarchy_counts_sample(
  value = c("wide", "long"),
  nr_children = 1
)

Arguments

value: wide - wide format, long - long format, plot - ggplot
nr_children: get summary for 1,2 or more number of children

Method `plot_hierarchy_counts_sample()`

barplot showing number of elements at each level in every sample

Usage

LFQDataSummariser$plot_hierarchy_counts_sample(nr_children = 1)

Arguments

nr_children: get summary for 1,2 or more number of children
value: wide - wide format, long - long format, plot - ggplot

Method `interaction_missing_stats()`

missing per condition and protein

Usage

LFQDataSummariser$interaction_missing_stats()

Method `upset_interaction_missing_stats()`

upset plot with missing information per protein and condition

Usage

LFQDataSummariser$upset_interaction_missing_stats(tr = 2)

Arguments

tr: if less than tr observations in condition then missing

Method `missingness_per_group()`

missing stats per condition

Usage

LFQDataSummariser$missingness_per_group()

Method `missingness_per_group_cumsum()`

missing stats per condition as cumulative sum

Usage

LFQDataSummariser$missingness_per_group_cumsum()

Method `plot_missingness_per_group()`

barplot with number of features with 1,2, etc missing in condition

Usage

LFQDataSummariser$plot_missingness_per_group()

Returns

ggplot

Method `plot_missingness_per_group_cumsum()`

barplot with cumulative sum of features with 1,2, etc missing in condition

Usage

LFQDataSummariser$plot_missingness_per_group_cumsum()

Returns

ggplot

Method `percentage_abundance()`

Does roll up to highest hierarchy and Computes the percent abundance of proteins overall and within each group

Usage

LFQDataSummariser$percentage_abundance()

Returns

data frame

Method `clone()`

The objects of this class are cloneable with this method.

Usage

LFQDataSummariser$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config()

data <- istar$data
lfqdata <- LFQData$new(data, istar$config)
sum <- lfqdata$get_Summariser()
sum
sum$hierarchy_counts()
sum$hierarchy_counts_sample("wide")
sum$hierarchy_counts_sample("long")
sum$plot_hierarchy_counts_sample()
sum$plot_hierarchy_counts_sample()
tmp <- sum$interaction_missing_stats()

sum$missingness_per_group()
sum$missingness_per_group_cumsum()
sum$plot_missingness_per_group()
sum$plot_missingness_per_group_cumsum()
sum$upset_interaction_missing_stats()
sum$percentage_abundance()
istar <- prolfqua::sim_lfq_data_peptide_config()

data <- istar$data
lfqdata <- LFQData$new(data, istar$config)
sum <- lfqdata$get_Summariser()
sum
sum$hierarchy_counts()
sum$hierarchy_counts_sample("wide")
sum$hierarchy_counts_sample("long")
sum$plot_hierarchy_counts_sample()
sum$plot_hierarchy_counts_sample()
tmp <- sum$interaction_missing_stats()

sum$missingness_per_group()
sum$missingness_per_group_cumsum()
sum$plot_missingness_per_group()
sum$plot_missingness_per_group_cumsum()
sum$upset_interaction_missing_stats()
sum$percentage_abundance()

converts LFQData object to SummarizedExperiment

Description

For compatibility with Bioconductor

Usage

LFQDataToSummarizedExperiment(lfqdata)
LFQDataToSummarizedExperiment(lfqdata)

Arguments

lfqdata

LFQData object

Value

SummarizedExperiment (bioconductor)

Examples


istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data
lfqdata <- LFQData$new(data, istar$config)
lfqdata$data_wide()
if(require("SummarizedExperiment")){
   tmp <- LFQDataToSummarizedExperiment(lfqdata)
}

istar <- prolfqua::sim_lfq_data_peptide_config()
data <- istar$data
lfqdata <- LFQData$new(data, istar$config)
lfqdata$data_wide()
if(require("SummarizedExperiment")){
   tmp <- LFQDataToSummarizedExperiment(lfqdata)
}

Decorate LFQData with Methods for transforming Intensities

Description

Decorate LFQData with Methods for transforming Intensities

Value

An R6 class generator.

Public fields

lfq: LFQData object

Methods

Public methods

LFQDataTransformer$new()
LFQDataTransformer$log2()
LFQDataTransformer$get_scales()
LFQDataTransformer$robscale()
LFQDataTransformer$robscale_subset()
LFQDataTransformer$center_to_reference()
LFQDataTransformer$intensity_array()
LFQDataTransformer$intensity_matrix()
LFQDataTransformer$clone()

Method `new()`

initialize

Usage

LFQDataTransformer$new(lfqdata)

Arguments

lfqdata: LFQData object to transform

Method `log2()`

log2 transform data

Usage

LFQDataTransformer$log2(force = FALSE)

Arguments

force: if 'FALSE', data already log2 transformed will not be transformed a second time. 'TRUE' forces log transformation

Returns

LFQDataTransformer

Method `get_scales()`

get mean and variance and standard deviation in each sample

Usage

LFQDataTransformer$get_scales()

Returns

list with means and mads

Method `robscale()`

robust scale data

Usage

LFQDataTransformer$robscale(
  preserve_mean = TRUE,
  colname = "transformedIntensity"
)

Arguments

preserve_mean: should original mean value be preserved TRUE, if FALSE then center at zero
colname: new name of transformed column

Returns

LFQDataTransformer (self)

Method `robscale_subset()`

log2 transform and robust scale data based on subset

Usage

LFQDataTransformer$robscale_subset(
  lfqsubset,
  preserve_mean = TRUE,
  colname = "transformed_abundance"
)

Arguments

lfqsubset: LFQData subset to use for normalization
preserve_mean: should original mean value be preserved TRUE, if FALSE then center at zero
colname: - how to name the transformed intensities, default transformedIntensity

Returns

LFQDataTransformer (self)

Method `center_to_reference()`

center data to a reference subset

Usage

LFQDataTransformer$center_to_reference(lfqsubset)

Arguments

lfqsubset: LFQData subset to use as reference

Returns

LFQDataTransformer (self)

Method `intensity_array()`

Transforms intensities

Usage

LFQDataTransformer$intensity_array(.func = log2, force = FALSE)

Arguments

.func: transformation function working with arrays e.g. log2, log10, asinh etc.
force: transformation on already transformed data.

Returns

LFQDataTransformer (self)

Method `intensity_matrix()`

pass a function which works with matrices, e.g., vsn::justvsn

Usage

LFQDataTransformer$intensity_matrix(.func = robust_scale, force = FALSE)

Arguments

.func: any function taking a matrix and returning a matrix (columns sample, rows feature, e.g. 'base::scale()'), default 'robust_scale'
force: transformation on data already transformed

Returns

LFQDataTransformer (self)

Method `clone()`

The objects of this class are cloneable with this method.

Usage

LFQDataTransformer$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_peptide_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)

lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()

x <- lfqTrans$intensity_array(log2)

x$lfq$is_transformed()
x <- x$intensity_matrix(robust_scale)
plotter <- x$lfq$get_Plotter()
plotter$intensity_distribution_density()

# transform by asinh root and scale

lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
x <- lfqTrans$intensity_array(asinh)
mads1 <- mean(x$get_scales()$mads)
x <- lfqTrans$intensity_matrix(robust_scale, force = TRUE)
mads2 <- mean(x$get_scales()$mads)
stopifnot(abs(mads1 - mads2) < 1e-8)


stopifnot(abs(mean(x$get_scales()$medians)) < 1e-8)
lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
lfqTrans$log2()
before <- lfqTrans$get_scales()
lfqTrans$robscale()
after <- lfqTrans$get_scales()
stopifnot(abs(mean(before$medians) - mean(after$medians)) < 1e-8)
stopifnot(abs(mean(before$mads) - mean(after$mads)) < 1e-8)

# normalize data using vsn
lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
lfqTransCheck <- lfqcopy$get_Transformer()

lfqTransCheck$log2()
lfqTransCheck$get_scales()
lfqTransCheck$lfq$get_Plotter()$intensity_distribution_density()

if(require("vsn")){
 res <- lfqTrans$intensity_matrix( .func = vsn::justvsn)
 res$lfq$get_Plotter()$intensity_distribution_density()
 res$get_scales()
}
if(require("preprocessCore")){
quant <- function(y){
 ynorm <- preprocessCore::normalize.quantiles(y)
 rownames(ynorm) <- rownames(y)
 colnames(ynorm) <- colnames(y)
 return(ynorm)
}
 res <- lfqTrans$intensity_matrix( .func = quant)
 res$lfq$get_Plotter()$intensity_distribution_density()
}


#subset scaling

istar2 <- sim_lfq_data_peptide_config()
lfqdata2 <- LFQData$new(istar2$data, istar2$config)
lfqdata2 <- lfqdata2$get_Transformer()$intensity_array(log2)$lfq
head(lfqdata2$hierarchy())
internal <- lfqdata2$get_subset(head(lfqdata2$hierarchy()))
internal$hierarchy()
tr <- lfqdata2$get_Transformer()
tr$center_to_reference(internal)
pl <- tr$lfq$get_Plotter()
pl$intensity_distribution_density()
lfqdata2$get_Plotter()$intensity_distribution_density()
robscale <- lfqdata2$get_Transformer()$robscale_subset(internal)$lfq
robscale$get_Plotter()$intensity_distribution_density()

istar <- sim_lfq_data_peptide_config(Nprot = 50)
lfqdata <- LFQData$new(istar$data, istar$config)

lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()

x <- lfqTrans$intensity_array(log2)

x$lfq$is_transformed()
x <- x$intensity_matrix(robust_scale)
plotter <- x$lfq$get_Plotter()
plotter$intensity_distribution_density()

# transform by asinh root and scale

lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
x <- lfqTrans$intensity_array(asinh)
mads1 <- mean(x$get_scales()$mads)
x <- lfqTrans$intensity_matrix(robust_scale, force = TRUE)
mads2 <- mean(x$get_scales()$mads)
stopifnot(abs(mads1 - mads2) < 1e-8)


stopifnot(abs(mean(x$get_scales()$medians)) < 1e-8)
lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
lfqTrans$log2()
before <- lfqTrans$get_scales()
lfqTrans$robscale()
after <- lfqTrans$get_scales()
stopifnot(abs(mean(before$medians) - mean(after$medians)) < 1e-8)
stopifnot(abs(mean(before$mads) - mean(after$mads)) < 1e-8)

# normalize data using vsn
lfqcopy <- lfqdata$get_copy()
lfqTrans <- lfqcopy$get_Transformer()
lfqTransCheck <- lfqcopy$get_Transformer()

lfqTransCheck$log2()
lfqTransCheck$get_scales()
lfqTransCheck$lfq$get_Plotter()$intensity_distribution_density()

if(require("vsn")){
 res <- lfqTrans$intensity_matrix( .func = vsn::justvsn)
 res$lfq$get_Plotter()$intensity_distribution_density()
 res$get_scales()
}
if(require("preprocessCore")){
quant <- function(y){
 ynorm <- preprocessCore::normalize.quantiles(y)
 rownames(ynorm) <- rownames(y)
 colnames(ynorm) <- colnames(y)
 return(ynorm)
}
 res <- lfqTrans$intensity_matrix( .func = quant)
 res$lfq$get_Plotter()$intensity_distribution_density()
}


#subset scaling

istar2 <- sim_lfq_data_peptide_config()
lfqdata2 <- LFQData$new(istar2$data, istar2$config)
lfqdata2 <- lfqdata2$get_Transformer()$intensity_array(log2)$lfq
head(lfqdata2$hierarchy())
internal <- lfqdata2$get_subset(head(lfqdata2$hierarchy()))
internal$hierarchy()
tr <- lfqdata2$get_Transformer()
tr$center_to_reference(internal)
pl <- tr$lfq$get_Plotter()
pl$intensity_distribution_density()
lfqdata2$get_Plotter()$intensity_distribution_density()
robscale <- lfqdata2$get_Transformer()$robscale_subset(internal)$lfq
robscale$get_Plotter()$intensity_distribution_density()

List currently registered contrast facades

Description

List currently registered contrast facades

Usage

list_facades()
list_facades()

Value

A list mirroring FACADE_REGISTRY but reflecting any entries added via register_facade from downstream packages.

Examples

head(names(list_facades()))
head(names(list_facades()))

read minimal yaml to reconstruct configuration

Description

read minimal yaml to reconstruct configuration

Usage

list_to_AnalysisConfiguration(dd)
list_to_AnalysisConfiguration(dd)

Arguments

dd

list with table and parameter elements as produced by R6_extract_values

Value

The computed result.

Examples


DEAconfig <- AnalysisConfiguration$new()
DEAconfig$file_name <- "Replicate.Name"
DEAconfig$hierarchy[["protein_Id"]] <- "Protein.Name"
DEAconfig$hierarchy[["peptide_Id"]] <- "Peptide.Sequence"
DEAconfig$hierarchy[["precursor_Id"]] <- c("Peptide.Sequence", "Precursor.Charge")
DEAconfig$hierarchy[["fragment_Id"]] <- c(
  "Peptide.Sequence", "Precursor.Charge",
  "Fragment.Ion", "Product.Charge"
)
DEAconfig$ident_q_value <- "annotation_QValue"
DEAconfig$set_response("Area")
DEAconfig$isotope_label <- "Isotope.Label"
configList <- prolfqua::R6_extract_values(DEAconfig)
stopifnot(class(configList) == "list")
config <- list_to_AnalysisConfiguration(configList)
all.equal(prolfqua::R6_extract_values(config), configList)
DEAconfig <- AnalysisConfiguration$new()
DEAconfig$file_name <- "Replicate.Name"
DEAconfig$hierarchy[["protein_Id"]] <- "Protein.Name"
DEAconfig$hierarchy[["peptide_Id"]] <- "Peptide.Sequence"
DEAconfig$hierarchy[["precursor_Id"]] <- c("Peptide.Sequence", "Precursor.Charge")
DEAconfig$hierarchy[["fragment_Id"]] <- c(
  "Peptide.Sequence", "Precursor.Charge",
  "Fragment.Ion", "Product.Charge"
)
DEAconfig$ident_q_value <- "annotation_QValue"
DEAconfig$set_response("Area")
DEAconfig$isotope_label <- "Isotope.Label"
configList <- prolfqua::R6_extract_values(DEAconfig)
stopifnot(class(configList) == "list")
config <- list_to_AnalysisConfiguration(configList)
all.equal(prolfqua::R6_extract_values(config), configList)

Rebuild a ContrastConfiguration from a list

Description

Round-trip partner of R6_extract_values, which keeps the fields but drops the methods. Use it to restore the column-role mapping from a serialized result artifact (e.g. 'SummarizedExperiment' metadata or an AnnData 'uns' entry) so consumers can call has_pvalue() and friends again.

Usage

list_to_ContrastConfiguration(dd)
list_to_ContrastConfiguration(dd)

Arguments

dd

named list of ContrastConfiguration fields

Value

A ContrastConfiguration.

Examples

cfg <- ContrastConfiguration$new(
  subject_id = "protein_Id",
  contrast_col = "Bait",
  pvalue_col = NA_character_
)
values <- R6_extract_values(cfg)
restored <- list_to_ContrastConfiguration(values)
stopifnot(all.equal(R6_extract_values(restored), values))
restored$has_pvalue()
cfg <- ContrastConfiguration$new(
  subject_id = "protein_Id",
  contrast_col = "Bait",
  pvalue_col = NA_character_
)
values <- R6_extract_values(cfg)
restored <- list_to_ContrastConfiguration(values)
stopifnot(all.equal(R6_extract_values(restored), values))
restored$has_pvalue()

Look up a contrast facade by short name

Description

Look up a contrast facade by short name

Usage

lookup_facade(name)
lookup_facade(name)

Arguments

name

short method name (e.g. "lm", "saint").

Value

Registry entry list with at least class, needs, package, needs_saint_annotation; or NULL if no entry exists.

Examples

lookup_facade("lm")$class
is.null(lookup_facade("does-not-exist"))
lookup_facade("lm")$class
is.null(lookup_facade("does-not-exist"))

Likelihood ratio test

Description

Likelihood ratio test

Usage

LR_test(
  complete_models,
  model_name,
  complete_models_int,
  model_name_int,
  subject_id = "protein_Id",
  path = NULL
)
LR_test(
  complete_models,
  model_name,
  complete_models_int,
  model_name_int,
  subject_id = "protein_Id",
  path = NULL
)

Arguments

complete_models

table with models (see build model)

model_name

name of model

complete_models_int

reduced model

model_name_int

name of reduced model

subject_id

subject id typically Accession or protein_Id

path

default NULL, set to a directory if you need to write diagnostic plots.

Value

The computed result.

Examples

data_2Factor <- prolfqua::sim_lfq_data_2factor_config(
 Nprot = 200,
 with_missing = TRUE,
 weight_missing = 2)

pMerged <- LFQData$new(data_2Factor$data, data_2Factor$config)

pMerged$response()
pMerged$factors()

formula_condition_and_Batches <-
  prolfqua::strategy_lm("abundance ~ Treatment + Background")
modCB <- prolfqua::build_model(
  pMerged,
  formula_condition_and_Batches)

formula_condition <-
  prolfqua::strategy_lm("abundance ~ Treatment")
modC <- prolfqua::build_model(
  pMerged,
  formula_condition)

tmp <- LR_test(modCB$model_df, "modCB", modC$model_df, "modB")
hist(tmp$likelihood_ratio_test.pValue)

data_2Factor <- prolfqua::sim_lfq_data_2factor_config(
 Nprot = 200,
 with_missing = TRUE,
 weight_missing = 2)

pMerged <- LFQData$new(data_2Factor$data, data_2Factor$config)

pMerged$response()
pMerged$factors()

formula_condition_and_Batches <-
  prolfqua::strategy_lm("abundance ~ Treatment + Background")
modCB <- prolfqua::build_model(
  pMerged,
  formula_condition_and_Batches)

formula_condition <-
  prolfqua::strategy_lm("abundance ~ Treatment")
modC <- prolfqua::build_model(
  pMerged,
  formula_condition)

tmp <- LR_test(modCB$model_df, "modCB", modC$model_df, "modB")
hist(tmp$likelihood_ratio_test.pValue)

main effects contrasts

Description

main effects contrasts

Usage

main_effect_contrasts(primary_levels, secondary_levels)
main_effect_contrasts(primary_levels, secondary_levels)

Arguments

primary_levels

character vector of primary factor levels

secondary_levels

character vector of secondary factor levels

Value

Contrast definitions or contrast results.

Examples

primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T150", "T0", "T300")
main_effect_contrasts(primary_levels, secondary_levels)
main_effect_contrasts(secondary_levels, primary_levels)
primary_levels <- c("MI", "MINOCAM")
secondary_levels <- c("T150", "T0", "T300")
main_effect_contrasts(primary_levels, secondary_levels)
main_effect_contrasts(secondary_levels, primary_levels)

matrix or data.frame to tibble (taken from tidyquant)

Description

matrix or data.frame to tibble (taken from tidyquant)

Usage

matrix_to_tibble(x, preserve_row_names = "row.names", ...)
matrix_to_tibble(x, preserve_row_names = "row.names", ...)

Arguments

x

a matrix

preserve_row_names

give name to rownames column, if NULL discard rownames

...

further parameters passed to as_tibble

Value

The computed result.

Examples

x <- matrix(rnorm(20), ncol=4)
rownames(x) <- LETTERS[seq_len(nrow(x))]
matrix_to_tibble(x)
!(is.matrix(x) || is.data.frame(x))

x <- matrix(rnorm(20), ncol=4)
rownames(x) <- LETTERS[seq_len(nrow(x))]
matrix_to_tibble(x)
!(is.matrix(x) || is.data.frame(x))

Merge contrast results coming from two different model.

Description

Typically used to fill proteins missing from Contrasts with estimates from another model, e.g. the LOD-imputed refits of ContrastsLMImputeFacade

Usage

merge_contrasts_results(prefer, add, model_name = "mergedModel")
merge_contrasts_results(prefer, add, model_name = "mergedModel")

Arguments

prefer

contrasts to use preferentially

add

contrasts to add from if missing in prefer

model_name

name of the merged model default "mergedModel"

Value

Contrast definitions or contrast results.

Examples

prefer_df <- data.frame(
  protein_Id = c("P1", "P2"),
  contrast = "A_vs_B",
  modelName = "prefer",
  diff = c(1, 2),
  p.value = c(0.01, 0.2),
  FDR = c(0.02, 0.2),
  statistic = c(3, NA)
)
add_df <- data.frame(
  protein_Id = c("P1", "P2"),
  contrast = "A_vs_B",
  modelName = "add",
  diff = c(1.1, 2.1),
  p.value = c(0.02, 0.03),
  FDR = c(0.03, 0.04),
  statistic = c(2.8, 2.5)
)
prefer <- ContrastsTable$new(prefer_df, subject_id = "protein_Id", model_name = "prefer")
add <- ContrastsTable$new(add_df, subject_id = "protein_Id", model_name = "add")
merged <- merge_contrasts_results(prefer, add)
merged$merged$get_contrasts()
prefer_df <- data.frame(
  protein_Id = c("P1", "P2"),
  contrast = "A_vs_B",
  modelName = "prefer",
  diff = c(1, 2),
  p.value = c(0.01, 0.2),
  FDR = c(0.02, 0.2),
  statistic = c(3, NA)
)
add_df <- data.frame(
  protein_Id = c("P1", "P2"),
  contrast = "A_vs_B",
  modelName = "add",
  diff = c(1.1, 2.1),
  p.value = c(0.02, 0.03),
  FDR = c(0.03, 0.04),
  statistic = c(2.8, 2.5)
)
prefer <- ContrastsTable$new(prefer_df, subject_id = "protein_Id", model_name = "prefer")
add <- ContrastsTable$new(add_df, subject_id = "protein_Id", model_name = "add")
merged <- merge_contrasts_results(prefer, add)
merged$merged$get_contrasts()

Group statistics and limit of detection

Description

Group statistics and limit of detection

Details

Computes per-group statistics of the abundances and, from them, the limit of detection (LOD) used by build_model_impute to impute groups without observations.

Value

An R6 class generator.

Public fields

data: data
config: config
prob: quantile of groups with one observed value to estimate LOD
stats: data.frame with group statistics

Methods

Public methods

MissingHelpers$new()
MissingHelpers$get_stats()
MissingHelpers$get_lod()
MissingHelpers$clone()

Method `new()`

initialize

Usage

MissingHelpers$new(data, config, prob = 0.5)

Arguments

data: data
config: config
prob: default 0.5, median of groups with one observed value

Method `get_stats()`

get data.frame with statistics

Usage

MissingHelpers$get_stats()

Returns

data.frame

Method `get_lod()`

determine limit of detection computes quantile of abundances in groups with a single observation; without such groups, of the partly observed groups with the fewest observations, and when every group is complete or empty, the lowest group mean

Usage

MissingHelpers$get_lod()

Returns

numeric LOD

Method `clone()`

The objects of this class are cloneable with this method.

Usage

MissingHelpers$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

dd <- prolfqua::sim_lfq_data_protein_config(Nprot = 100,weight_missing = 2)
mh <- prolfqua::MissingHelpers$new(dd$data, dd$config, prob = 0.8)
xx <- mh$get_stats()
xx <- mh$get_lod()

dd <- prolfqua::sim_lfq_data_2factor_config(Nprot = 100,weight_missing = 0.1)
mh <- prolfqua::MissingHelpers$new(dd$data, dd$config, prob = 0.8)
mh$get_stats()$interaction |> table()
mh$get_lod()

dd <- prolfqua::sim_lfq_data_protein_config(Nprot = 100,weight_missing = 2)
mh <- prolfqua::MissingHelpers$new(dd$data, dd$config, prob = 0.8)
xx <- mh$get_stats()
xx <- mh$get_lod()

dd <- prolfqua::sim_lfq_data_2factor_config(Nprot = 100,weight_missing = 0.1)
mh <- prolfqua::MissingHelpers$new(dd$data, dd$config, prob = 0.8)
mh$get_stats()$interaction |> table()
mh$get_lod()

R6 class representing modelling result

Description

R6 class representing modelling result

Value

An R6 class generator.

Super class

prolfqua::ModelInterface -> Model

Public fields

model_df: data.frame with modelling data and model.
model_name: name of model
subject_id: e.g. protein_Id
model_strategy: function to create the models
p.adjust: function to adjust p-values

Methods

Public methods

Model$new()
Model$get_coefficients()
Model$get_anova()
Model$clone()

Inherited methods

prolfqua::ModelInterface$anova_histogram()
prolfqua::ModelInterface$coef_histogram()
prolfqua::ModelInterface$coef_pairs()
prolfqua::ModelInterface$coef_volcano()

Method `new()`

initialize

Usage

Model$new(
  model_df,
  model_strategy,
  model_name,
  subject_id = "protein_Id",
  p.adjust = prolfqua::adjust_p_values
)

Arguments

model_df: dataframe with modelling results
model_strategy: model_strategy see strategy_lmer
model_name: name of model
subject_id: subject column name
p.adjust: method to adjust p-values

Method `get_coefficients()`

return model coefficient table

Usage

Model$get_coefficients()

Method `get_anova()`

return anova table

Usage

Model$get_anova()

Method `clone()`

The objects of this class are cloneable with this method.

Usage

Model$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples




istar <- sim_lfq_data_peptide_config(Nprot = 20)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
model_name <- "f_condtion_r_peptide"
formula_randomPeptide <-
  strategy_lmer(paste0(lfqdata$response(), " ~ group_ + (1 | peptide_Id)"),
   model_name = model_name)
mod <- prolfqua::build_model(
 lfqdata,
 formula_randomPeptide,
 model_name = model_name)

mod$model_df
aovtable  <- mod$get_anova()
mod$get_coefficients()
mod$coef_histogram()
mod$coef_volcano()
mod$coef_pairs()
mod$anova_histogram()

istar <- sim_lfq_data_peptide_config(Nprot = 20)
lfqdata <- LFQData$new(istar$data, istar$config)
lfqdata <- lfqdata$get_Transformer()$log2()$lfq
model_name <- "f_condtion_r_peptide"
formula_randomPeptide <-
  strategy_lmer(paste0(lfqdata$response(), " ~ group_ + (1 | peptide_Id)"),
   model_name = model_name)
mod <- prolfqua::build_model(
 lfqdata,
 formula_randomPeptide,
 model_name = model_name)

mod$model_df
aovtable  <- mod$get_anova()
mod$get_coefficients()
mod$coef_histogram()
mod$coef_volcano()
mod$coef_pairs()
mod$anova_histogram()

R6 class representing modelling result

Description

R6 class representing modelling result

Value

An R6 class generator.

Super class

prolfqua::ModelInterface -> ModelFirth

Public fields

models: data.frame with modelling data and model.
model_name: name of model
subject_id: e.g. protein_Id
p.adjust: function to adjust p-values

Methods

Public methods

ModelFirth$new()
ModelFirth$get_coefficients()
ModelFirth$get_anova()
ModelFirth$anova_histogram()
ModelFirth$clone()

Inherited methods

prolfqua::ModelInterface$coef_histogram()
prolfqua::ModelInterface$coef_pairs()
prolfqua::ModelInterface$coef_volcano()

Method `new()`

initialize

Usage

ModelFirth$new(
  models,
  model_name = "modelFirth",
  subject_id = "protein_Id",
  p.adjust = prolfqua::adjust_p_values
)

Arguments

models: dataframe with modelling results
model_name: name of model
subject_id: subject column name
p.adjust: method to adjust p-values

Method `get_coefficients()`

return model coefficient table

Usage

ModelFirth$get_coefficients()

Method `get_anova()`

return anova table

Usage

ModelFirth$get_anova()

Method `anova_histogram()`

histogram of ANOVA results

Usage

ModelFirth$anova_histogram(what = c("p.value", "FDR"))

Arguments

what: show either "Pr..F." or "FDR.Pr..F."

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ModelFirth$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples





istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- prolfqua::encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
tmp <- LFQData$new(istar$data, istar$config)
formula <- paste0(tmp$get_config()$bin_resp , "~ group_")
mod <- build_model_logistf(tmp, formula)
tmp <- mod$get_coefficients()
mod$coef_histogram()
mod$coef_pairs()
mod$get_anova()
mod$coef_volcano()
mod$anova_histogram()

istar <- prolfqua::sim_lfq_data_protein_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- prolfqua::encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
tmp <- LFQData$new(istar$data, istar$config)
formula <- paste0(tmp$get_config()$bin_resp , "~ group_")
mod <- build_model_logistf(tmp, formula)
tmp <- mod$get_coefficients()
stopifnot(nrow(tmp) == 30)
mod$coef_histogram()
mod$coef_pairs()
mod$get_anova()
mod$coef_volcano()
mod$anova_histogram()
istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- prolfqua::encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
tmp <- LFQData$new(istar$data, istar$config)
formula <- paste0(tmp$get_config()$bin_resp , "~ group_")
mod <- build_model_logistf(tmp, formula)
tmp <- mod$get_coefficients()
mod$coef_histogram()
mod$coef_pairs()
mod$get_anova()
mod$coef_volcano()
mod$anova_histogram()

istar <- prolfqua::sim_lfq_data_protein_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- prolfqua::encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
tmp <- LFQData$new(istar$data, istar$config)
formula <- paste0(tmp$get_config()$bin_resp , "~ group_")
mod <- build_model_logistf(tmp, formula)
tmp <- mod$get_coefficients()
stopifnot(nrow(tmp) == 30)
mod$coef_histogram()
mod$coef_pairs()
mod$get_anova()
mod$coef_volcano()
mod$anova_histogram()

R6 interface class representing modelling result

Description

R6 interface class representing modelling result

Value

An R6 class generator.

Methods

Public methods

ModelInterface$get_coefficients()
ModelInterface$get_anova()
ModelInterface$coef_histogram()
ModelInterface$coef_volcano()
ModelInterface$coef_pairs()
ModelInterface$anova_histogram()
ModelInterface$clone()

Method `get_coefficients()`

return model coefficients

Usage

ModelInterface$get_coefficients()

Returns

data.frame

Method `get_anova()`

perform ANOVA analysis

Usage

ModelInterface$get_anova()

Returns

data.frame

Method `coef_histogram()`

histogram of coefficient p-values

Usage

ModelInterface$coef_histogram()

Returns

list with plot (ggplot) and file name

Method `coef_volcano()`

volcano plot of non-intercept coefficients

Usage

ModelInterface$coef_volcano()

Returns

list with plot (ggplot) and file name

Method `coef_pairs()`

coefficient estimates in wide format, one column per coefficient, for a pairs plot

Usage

ModelInterface$coef_pairs()

Returns

list with plot (data.frame) and file name

Method `anova_histogram()`

histogram of ANOVA p-values or FDR

Usage

ModelInterface$anova_histogram(what = c("p.value", "FDR"))

Arguments

what: show either "p.value" or "FDR"

Returns

list with plot (ggplot) and file name

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ModelInterface$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


mi <- ModelInterface$new()

testthat::expect_error(mi$get_coefficients())
testthat::expect_error(mi$get_anova())
testthat::expect_error(mi$coef_histogram())
testthat::expect_error(mi$coef_volcano())
testthat::expect_error(mi$coef_pairs())
testthat::expect_error(mi$anova_histogram())


mi <- ModelInterface$new()

testthat::expect_error(mi$get_coefficients())
testthat::expect_error(mi$get_anova())
testthat::expect_error(mi$coef_histogram())
testthat::expect_error(mi$coef_volcano())
testthat::expect_error(mi$coef_pairs())
testthat::expect_error(mi$anova_histogram())

R6 class representing a limma modelling result

Description

R6 class representing a limma modelling result

Details

Same API as Model: get_anova(), get_coefficients(), coef_histogram(), coef_volcano(), anova_histogram().

Value

An R6 class generator.

Super class

prolfqua::ModelInterface -> ModelLimma

Public fields

fit: limma MArrayLM object from lmFit
subject_id: protein ID column name(s)
model_name: model name
rowdata: protein ID mapping from to_wide()$rowdata
trend: passed to eBayes
robust: passed to eBayes
dummy_model: one fitted lm for linfct extraction
p.adjust: function to adjust p-values
imputed_proteins: character vector of protein ids (matching rownames(fit$coefficients)) that were rescued via LOD imputation. ContrastsLimma$get_contrasts() reads this to tag refit rows with a "_imputed" modelName suffix so the rescue is visible downstream. Empty for builders that do not impute.

Methods

Public methods

ModelLimma$new()
ModelLimma$get_coefficients()
ModelLimma$get_anova()
ModelLimma$clone()

Inherited methods

prolfqua::ModelInterface$anova_histogram()
prolfqua::ModelInterface$coef_histogram()
prolfqua::ModelInterface$coef_pairs()
prolfqua::ModelInterface$coef_volcano()

Method `new()`

initialize ModelLimma

Usage

ModelLimma$new(
  fit,
  subject_id,
  model_name,
  rowdata,
  trend = FALSE,
  robust = FALSE,
  dummy_model = NULL,
  p.adjust = prolfqua::adjust_p_values,
  imputed_proteins = character(0)
)

Arguments

fit: limma MArrayLM from lmFit
subject_id: protein ID column name(s)
model_name: model name
rowdata: protein ID mapping
trend: passed to eBayes
robust: passed to eBayes
dummy_model: one fitted lm for linfct extraction
p.adjust: function to adjust p-values
imputed_proteins: character vector of LOD-rescued protein ids

Method `get_coefficients()`

return model coefficient table in long format

Usage

ModelLimma$get_coefficients()

Returns

data.frame

Method `get_anova()`

return anova table (F-test per protein across all non-intercept coefficients)

Usage

ModelLimma$get_anova()

Returns

data.frame

Method `clone()`

The objects of this class are cloneable with this method.

Usage

ModelLimma$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples


istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")
strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma(lProt, strat)

coeffs <- mod$get_coefficients()
head(coeffs)
anova_tbl <- mod$get_anova()
head(anova_tbl)
mod$coef_histogram()
mod$coef_volcano()
mod$anova_histogram()

istar <- sim_lfq_data_protein_config(Nprot = 50)
lProt <- LFQData$new(istar$data, istar$config)
lProt$rename_response("transformedIntensity")
strat <- strategy_limma("transformedIntensity ~ group_")
mod <- build_model_limma(lProt, strat)

coeffs <- mod$get_coefficients()
head(coeffs)
anova_tbl <- mod$get_anova()
head(anova_tbl)
mod$coef_histogram()
mod$coef_volcano()
mod$anova_histogram()

Max nr_children across samples per hierarchy unit

Description

Aggregates nr_children per sample (via nr_obs_sample), then takes the max per hierarchy unit.

Usage

nr_children_experiment(
  data,
  response,
  hierarchy_keys_depth,
  file_name,
  nr_children_col,
  name_nr_child = "nr_child_exp"
)
nr_children_experiment(
  data,
  response,
  hierarchy_keys_depth,
  file_name,
  nr_children_col,
  name_nr_child = "nr_child_exp"
)

Arguments

data

data.frame

response

character — intensity column name

hierarchy_keys_depth

character vector — hierarchy columns at current depth

file_name

character — file name column

nr_children_col

character — nr_children column name

name_nr_child

character — output column name

Value

The computed result.

Examples

dd <- prolfqua::sim_lfq_data_peptide_config()
lfq <- LFQData$new(dd$data, dd$config)
xd <- nr_children_experiment(lfq$data_long(), lfq$response(),
  lfq$relevant_hierarchy_keys(), lfq$file_name(), lfq$nr_children_col())
stopifnot(min(xd$nr_child_exp) == 1)

dd <- prolfqua::sim_lfq_data_peptide_config()
lfq <- LFQData$new(dd$data, dd$config)
xd <- nr_children_experiment(lfq$data_long(), lfq$response(),
  lfq$relevant_hierarchy_keys(), lfq$file_name(), lfq$nr_children_col())
stopifnot(min(xd$nr_child_exp) == 1)

Count distinct child features per hierarchy unit

Description

Counts the number of distinct child-level entries (e.g. peptides per protein).

Usage

nr_features_experiment(
  data,
  hierarchy_keys,
  hierarchy_keys_depth,
  name_nr_child = "nr_child_exp"
)
nr_features_experiment(
  data,
  hierarchy_keys,
  hierarchy_keys_depth,
  name_nr_child = "nr_child_exp"
)

Arguments

data

data.frame

hierarchy_keys

character vector — all hierarchy column names

hierarchy_keys_depth

character vector — hierarchy columns at current depth

name_nr_child

character — output column name

Value

The computed result.

Examples

dd <- prolfqua::sim_lfq_data_peptide_config()
lfq <- LFQData$new(dd$data, dd$config)
xd <- nr_features_experiment(lfq$data_long(), lfq$hierarchy_keys(),
  lfq$relevant_hierarchy_keys())
stopifnot(min(xd$nr_child_exp) == 1)

dd <- prolfqua::sim_lfq_data_peptide_config()
lfq <- LFQData$new(dd$data, dd$config)
xd <- nr_features_experiment(lfq$data_long(), lfq$hierarchy_keys(),
  lfq$relevant_hierarchy_keys())
stopifnot(min(xd$nr_child_exp) == 1)

Aggregates e.g. protein abundances from peptide abundances

Description

Aggregates e.g. protein abundances from peptide abundances

Usage

nr_obs_sample(
  data,
  response,
  hierarchy_keys_depth,
  file_name,
  nr_children_col,
  new_child = nr_children_col
)
nr_obs_sample(
  data,
  response,
  hierarchy_keys_depth,
  file_name,
  nr_children_col,
  new_child = nr_children_col
)

Arguments

data

data.frame

response

character — intensity column name

hierarchy_keys_depth

character vector — hierarchy columns at current depth

file_name

character — file name column

nr_children_col

character — nr_children column name

new_child

character — output column name

Value

The computed result.

Examples

dd <- prolfqua::sim_lfq_data_peptide_config()
dd$data <- na.omit(dd$data)

xd <- nr_obs_sample(na.omit(dd$data), dd$config$get_response(),
  dd$config$hierarchy_keys_depth(), dd$config$file_name, dd$config$nr_children)
xd$nr_children |> table()
# xd |> pivot_wider(id_cols = protein_Id, names_from = sample, values_from = nr_children)

dp <- prolfqua::sim_lfq_data_protein_config()
xp <- nr_obs_sample(dp$data, dp$config$get_response(),
  dp$config$hierarchy_keys_depth(), dp$config$file_name, dp$config$nr_children)
# xp
# xp |> pivot_wider(id_cols = protein_Id, names_from = sample, values_from = nrPeptides)
xp$nrPeptides |> table()

dd <- prolfqua::sim_lfq_data_peptide_config()
dd$data <- na.omit(dd$data)

xd <- nr_obs_sample(na.omit(dd$data), dd$config$get_response(),
  dd$config$hierarchy_keys_depth(), dd$config$file_name, dd$config$nr_children)
xd$nr_children |> table()
# xd |> pivot_wider(id_cols = protein_Id, names_from = sample, values_from = nr_children)

dp <- prolfqua::sim_lfq_data_protein_config()
xp <- nr_obs_sample(dp$data, dp$config$get_response(),
  dp$config$hierarchy_keys_depth(), dp$config$file_name, dp$config$nr_children)
# xp
# xp |> pivot_wider(id_cols = protein_Id, names_from = sample, values_from = nrPeptides)
xp$nrPeptides |> table()

Package Data

Description

Package Data

Value

A documentation topic.

pivot model contrasts matrix to wide format produced by 'contrasts_linfct' and ...

Description

pivot model contrasts matrix to wide format produced by 'contrasts_linfct' and ...

Usage

pivot_model_contrasts_to_wide(
  model_interaction_contrasts,
  subject_id = "protein_Id",
  columns = c("estimate", "p.value", "p.value.adjusted"),
  contrast = "lhs"
)
pivot_model_contrasts_to_wide(
  model_interaction_contrasts,
  subject_id = "protein_Id",
  columns = c("estimate", "p.value", "p.value.adjusted"),
  contrast = "lhs"
)

Arguments

model_interaction_contrasts

data.frame with contrast results in long format

subject_id

column name(s) identifying subjects (e.g. protein_Id)

columns

character vector of value columns to pivot wide

contrast

column name containing contrast labels

Value

Contrast definitions or contrast results.

Examples


# this function is used by the contrast classes to implement the to wide method
contrast_df <- data.frame(
  protein_Id = c("P1", "P1", "P2", "P2"),
  lhs = c("A_vs_B", "C_vs_D", "A_vs_B", "C_vs_D"),
  estimate = c(1.0, 0.5, -0.2, 0.7),
  p.value = c(0.01, 0.2, 0.4, 0.05),
  p.value.adjusted = c(0.02, 0.25, 0.4, 0.08)
)
pivot_model_contrasts_to_wide(contrast_df)

# this function is used by the contrast classes to implement the to wide method
contrast_df <- data.frame(
  protein_Id = c("P1", "P1", "P2", "P2"),
  lhs = c("A_vs_B", "C_vs_D", "A_vs_B", "C_vs_D"),
  estimate = c(1.0, 0.5, -0.2, 0.7),
  p.value = c(0.01, 0.2, 0.4, 0.05),
  p.value.adjusted = c(0.02, 0.25, 0.4, 0.08)
)
pivot_model_contrasts_to_wide(contrast_df)

load data from prolfqua

Description

load data from prolfqua

Usage

prolfqua_data(datastr, package = "prolfqua")
prolfqua_data(datastr, package = "prolfqua")

Arguments

datastr

name of dataset

package

default prolfqua

Value

The computed result.

Examples

ionstar <- prolfqua_data("data_ionstar")
names(ionstar)
ionstar <- prolfqua_data("data_ionstar")
names(ionstar)

Extract all value slots in an R6 object

Description

Extract all value slots in an R6 object

Usage

R6_extract_values(r6class)
R6_extract_values(r6class)

Arguments

r6class

r6 class

Value

The computed result.

Examples

bb <- sim_lfq_data_peptide_config()
values <- R6_extract_values(bb$config)
names(values)
bb <- sim_lfq_data_peptide_config()
values <- R6_extract_values(bb$config)
names(values)

Register a contrast facade class

Description

Adds an entry to the prolfqua facade registry so that lookup_facade (and downstream consumers such as prolfquapp::R6_DEAnalyse$build_facade()) can resolve it by short name. Intended for downstream packages that ship additional modelling backends (e.g. prolfquasaint) and want to plug into the same dispatch path as the built-in facades.

Usage

register_facade(
  name,
  class,
  needs,
  package = "prolfqua",
  needs_saint_annotation = FALSE,
  ...
)
register_facade(
  name,
  class,
  needs,
  package = "prolfqua",
  needs_saint_annotation = FALSE,
  ...
)

Arguments

name

short method name, e.g. "saint".

class

character; R6 facade class name to instantiate.

needs

one of "same" (facade emits contrasts at the same hierarchy level as its input e.g. protein -> protein FC, peptide -> peptide FC) or "nested" (facade takes child-level input and emits parent-level contrasts, e.g. peptide -> protein FC).

package

package the facade class lives in. Defaults to "prolfqua".

needs_saint_annotation

TRUE if the backend requires annotation reading in SAINT mode (bait / control columns). Default FALSE.

...

additional fields stored on the registry entry, available to consumers via lookup_facade.

Value

The registered entry (invisibly).

Examples

lookup_facade("lm")$class
# downstream packages call this from .onLoad():
# prolfqua::register_facade("saint", class = "ContrastsSAINTFacade",
#                           needs = "same", package = "prolfquasaint",
#                           needs_saint_annotation = TRUE)
lookup_facade("lm")$class
# downstream packages call this from .onLoad():
# prolfqua::register_facade("saint", class = "ContrastsSAINTFacade",
#                           needs = "same", package = "prolfquasaint",
#                           needs_saint_annotation = TRUE)

Setup a tidy table compatible with a `AnalysisConfiguration`

Description

Extracts columns relevant for a configuration from a data frame and create new columns e.g. sampleName column etc.

Usage

setup_analysis(data, configuration, debug = FALSE)
setup_analysis(data, configuration, debug = FALSE)

Arguments

data

data.frame

configuration

AnalysisConfiguration

debug

if FALSE (default) and any hierarchy-key/sample combination has more than one observation, stop with an informative error. If TRUE, warn and return the diagnostic count table (inspect rows where n > 1) instead.

Value

A tibble with the columns relevant for configuration (sample, hierarchy, factor and response columns). Stops if any hierarchy-key/sample combination is observed more than once (unless debug = TRUE).

Examples


# Example with a normValue column (e.g., creatinine)
set.seed(1234)
data <- sim_lfq_data(Nprot = 10, PEPTIDE = TRUE, N = 4)
data$nr_children <- 1
data$isotopeLabel <- "light"
data$qValue <- 0

# Add creatinine values per sample (not per protein/peptide)
sample_creatinine <- data |> dplyr::select(sample) |> dplyr::distinct() |>
  dplyr::mutate(creatinine = rnorm(dplyr::n(), mean = 100, sd = 10))
data <- dplyr::inner_join(data, sample_creatinine, by = "sample")

config <- AnalysisConfiguration$new()
config$file_name = "sample"
config$factors["group_"] = "group"
config$hierarchy[["protein_Id"]] = c("proteinID", "idtype2")
config$hierarchy[["peptide_Id"]] = "peptideID"
config$set_response("abundance")
config$norm_value = "creatinine"

adata <- setup_analysis(data, config)

# Example with a normValue column (e.g., creatinine)
set.seed(1234)
data <- sim_lfq_data(Nprot = 10, PEPTIDE = TRUE, N = 4)
data$nr_children <- 1
data$isotopeLabel <- "light"
data$qValue <- 0

# Add creatinine values per sample (not per protein/peptide)
sample_creatinine <- data |> dplyr::select(sample) |> dplyr::distinct() |>
  dplyr::mutate(creatinine = rnorm(dplyr::n(), mean = 100, sd = 10))
data <- dplyr::inner_join(data, sample_creatinine, by = "sample")

config <- AnalysisConfiguration$new()
config$file_name = "sample"
config$factors["group_"] = "group"
config$hierarchy[["protein_Id"]] = c("proteinID", "idtype2")
config$hierarchy[["peptide_Id"]] = "peptideID"
config$set_response("abundance")
config$norm_value = "creatinine"

adata <- setup_analysis(data, config)

Scale estimate for an augmented rfit fit

Description

Returns the rank-based scale 'tauhat'. This stands in for the residual standard deviation of an 'lm'; it does not affect the Wald std.error or p-value (those come from 'vcov()'), but it is reported and used as the per- protein variance estimate if results are passed to limma/eBayes moderation.

Usage

## S3 method for class 'rfit_prolfqua'
sigma(object, ...)
## S3 method for class 'rfit_prolfqua'
sigma(object, ...)

Arguments

object

an 'rfit_prolfqua' fit

...

ignored

Value

the rank-based scale estimate 'tauhat'.

Examples

fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
sigma(fit)
fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
sigma(fit)

simulate protein level data with two groups

Description

simulate protein level data with two groups

Usage

sim_lfq_data(
  Nprot = 20,
  N = 4,
  fc = list(A = c(D = -2, U = 2, N = 0), B = c(D = 1, U = -4)),
  prop = list(A = c(D = 10, U = 10), B = c(D = 5, U = 20)),
  mean_prot = 20,
  sdlog = log(1.2),
  probability_of_success = 0.3,
  PEPTIDE = FALSE
)
sim_lfq_data(
  Nprot = 20,
  N = 4,
  fc = list(A = c(D = -2, U = 2, N = 0), B = c(D = 1, U = -4)),
  prop = list(A = c(D = 10, U = 10), B = c(D = 5, U = 20)),
  mean_prot = 20,
  sdlog = log(1.2),
  probability_of_success = 0.3,
  PEPTIDE = FALSE
)

Arguments

Nprot

number of proteins

N

group size

fc

D - down regulation N - matrix, U - regulation

prop

proportion of down (D), up (U) and not regulated (N)

mean_prot

mean protein abundance on the natural scale

sdlog

standard deviation of the log-normal abundance distribution

probability_of_success

probability of success for the geometric distribution of peptide counts

PEPTIDE

if TRUE, simulate peptide-level data; if FALSE, protein-level

Value

A data frame containing simulated LFQ data.

Examples


res <- sim_lfq_data(Nprot = 10)
res <- sim_lfq_data(Nprot = 10, PEPTIDE = TRUE)
res <- sim_lfq_data(Nprot = 10, N=4)

res <- sim_lfq_data(Nprot = 10)
res <- sim_lfq_data(Nprot = 10, PEPTIDE = TRUE)
res <- sim_lfq_data(Nprot = 10, N=4)

Simulate data, protein, with config with 2 factors Treatment and Background

Description

Simulate data, protein, with config with 2 factors Treatment and Background

Usage

sim_lfq_data_2factor_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  PEPTIDE = FALSE,
  seed = 1234,
  TWO = TRUE
)
sim_lfq_data_2factor_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  PEPTIDE = FALSE,
  seed = 1234,
  TWO = TRUE
)

Arguments

Nprot

number of proteins to simulate

with_missing

add missing values, default TRUE

weight_missing

controls proportion of missing values; greater weight means more missingness

PEPTIDE

if TRUE, simulate peptide-level data; if FALSE, protein-level

seed

seed for reproducibility, if NULL no seed is set.

TWO

use two factors for modelling

Value

A list containing simulated data and an analysis configuration.

Examples

x <- sim_lfq_data_2factor_config(PEPTIDE= FALSE)
dim(x$data)
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))
x <- sim_lfq_data_2factor_config(PEPTIDE = TRUE)

head(x$data)
x <- sim_lfq_data_2factor_config(PEPTIDE = TRUE, TWO = TRUE)
names(x)
nrow(x$data) > 10
x$data$Treatment |> table()
x <- sim_lfq_data_2factor_config(PEPTIDE= FALSE)
dim(x$data)
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))
x <- sim_lfq_data_2factor_config(PEPTIDE = TRUE)

head(x$data)
x <- sim_lfq_data_2factor_config(PEPTIDE = TRUE, TWO = TRUE)
names(x)
nrow(x$data) > 10
x$data$Treatment |> table()

Simulate data, protein and peptide, with config

Description

Simulate data, protein and peptide, with config

Usage

sim_lfq_data_peptide_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  seed = 1234,
  N = 4
)
sim_lfq_data_peptide_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  seed = 1234,
  N = 4
)

Arguments

Nprot

number of proteins to simulate

with_missing

add missing values, default TRUE

weight_missing

controls proportion of missing values; greater weight means more missingness

seed

seed for reproducibility, if NULL no seed is set.

N

number of replicates per group

Value

A list containing simulated data and an analysis configuration.

Examples


x <- sim_lfq_data_peptide_config()
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))
x <- sim_lfq_data_peptide_config()
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))

Simulate data, protein, with config

Description

Simulate data, protein, with config

Usage

sim_lfq_data_protein_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  seed = 1234,
  paired = FALSE
)
sim_lfq_data_protein_config(
  Nprot = 10,
  with_missing = TRUE,
  weight_missing = 0.2,
  seed = 1234,
  paired = FALSE
)

Arguments

Nprot

number of proteins to simulate

with_missing

add missing values, default TRUE

weight_missing

controls proportion of missing values; greater weight means more missingness

seed

seed for reproducibility, if NULL no seed is set.

paired

if TRUE, add a paired subject factor to the design

Value

A list containing simulated data and an analysis configuration.

Examples


x <- sim_lfq_data_protein_config()
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))
x <- sim_lfq_data_protein_config(with_missing = FALSE)

stopifnot(sum(is.na(x$data$abundance)) == 0)
xp <- sim_lfq_data_protein_config(with_missing = FALSE, paired = TRUE)
stopifnot(length(xp$config$factors) == 2)
stopifnot(nrow(xp$data) == nrow(x$data))
x <- sim_lfq_data_protein_config()
stopifnot("data.frame" %in% class(x$data))
stopifnot("AnalysisConfiguration" %in% class(x$config))
x <- sim_lfq_data_protein_config(with_missing = FALSE)

stopifnot(sum(is.na(x$data$abundance)) == 0)
xp <- sim_lfq_data_protein_config(with_missing = FALSE, paired = TRUE)
stopifnot(length(xp$config$factors) == 2)
stopifnot(nrow(xp$data) == nrow(x$data))

Create a quasibinomial detection-count strategy

Description

Create a quasibinomial detection-count strategy

Usage

strategy_binomial(modelstr, prior_count = 0.1, model_name = "binomial_nested")
strategy_binomial(modelstr, prior_count = 0.1, model_name = "binomial_nested")

Arguments

modelstr

right-hand-side model formula

prior_count

non-negative symmetric pseudo-count

model_name

model identity

Value

A StrategyBinomial object.

Examples

strategy <- strategy_binomial("~ group_", prior_count = 0.1)
strategy$formula
strategy <- strategy_binomial("~ group_", prior_count = 0.1)
strategy$formula

Create limma modelling strategy

Description

Wrapper that returns a StrategyLimma R6 object.

Usage

strategy_limma(
  modelstr,
  model_name = "limma",
  trend = FALSE,
  robust = FALSE,
  weights = NULL
)
strategy_limma(
  modelstr,
  model_name = "limma",
  trend = FALSE,
  robust = FALSE,
  weights = NULL
)

Arguments

modelstr

model formula as string (e.g. "abundance ~ group_")

model_name

name of model

trend

logical, passed to eBayes

robust

logical, passed to eBayes

weights

either a character string (column name in annotation for per-sample weights) or a numeric matrix (proteins x samples) passed to lmFit. Default NULL (no weights).

Value

The computed result.

Examples

strat <- strategy_limma("abundance ~ group_")
strat$formula
strat$model_name
strat <- strategy_limma("abundance ~ group_")
strat$formula
strat$model_name

Create limpa modelling strategy

Description

Wrapper that returns a StrategyLimpa R6 object.

Usage

strategy_limpa(
  modelstr,
  model_name = "limpa",
  trend = FALSE,
  robust = FALSE,
  plot = FALSE,
  span = NULL
)
strategy_limpa(
  modelstr,
  model_name = "limpa",
  trend = FALSE,
  robust = FALSE,
  plot = FALSE,
  span = NULL
)

Arguments

modelstr

model formula as string (e.g. "abundance ~ group_")

model_name

name of model

trend

logical, passed to eBayes

robust

logical, passed to eBayes

plot

logical, plot the vooma mean-variance trend

span

lowess smoother span (NULL = auto)

Value

The computed result.

Examples

strat <- strategy_limpa("abundance ~ group_")
strat$formula
strat$model_name
strat <- strategy_limpa("abundance ~ group_")
strat$formula
strat$model_name

Create Firth's logistic regression strategy

Description

Convenience wrapper that creates a StrategyLogistf object.

Convenience wrapper that creates a StrategyRfit object.

Convenience wrapper that creates a StrategyLmer object.

Convenience wrapper that creates a StrategyLM object.

Convenience wrapper that creates a StrategyRLM object.

Usage

strategy_logistf(modelstr, model_name = "logistf")

strategy_rfit(modelstr, model_name = "rfit")

strategy_lmer(modelstr, model_name = "Model")

strategy_lm(modelstr, model_name = "Model", weights = NULL)

strategy_rlm(modelstr, model_name = "Model")
strategy_logistf(modelstr, model_name = "logistf")

strategy_rfit(modelstr, model_name = "rfit")

strategy_lmer(modelstr, model_name = "Model")

strategy_lm(modelstr, model_name = "Model", weights = NULL)

strategy_rlm(modelstr, model_name = "Model")

Arguments

modelstr

model formula

model_name

name of model

weights

optional character string naming a column in the data containing per-observation weights, passed to lm. Default NULL (unweighted).

Value

a StrategyLogistf object

a StrategyRfit object

a StrategyLmer object

a StrategyLM object

a StrategyRLM object

Examples

tmp <- strategy_logistf("bin_resp ~ condition", model_name = "parallel design")
tmp$formula

istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
istar <- LFQData$new(istar$data, istar$config)
df <- istar$summarize_hierarchy()
df2 <- df[df[[ncol(df)]] > 1, ]
istar2 <- istar$get_subset(df2)
istar2$data_long() |>
  dplyr::group_by(protein_Id) |>
  tidyr::nest() -> nestProtein
modelFunction <- strategy_logistf("bin_resp ~ group_ + peptide_Id",
  model_name = "random_example")
modelFunction$model_fun(nestProtein$data[[1]])
modelFunction$model_fun(nestProtein$data[[4]])
tmp <- strategy_rfit("Intensity ~ condition", model_name = "parallel design")
tmp$formula
modelFunction <- strategy_lmer("abundanceC ~ group_ + (1|peptide_Id)",
  model_name = "random_example")
modelFunction$formula
tmp <- strategy_lm("Intensity ~ condition", model_name = "parallel design")
tmp$formula
tmp$weights
tmp <- strategy_rlm("Intensity ~ condition", model_name = "parallel design")
tmp$formula
tmp <- strategy_logistf("bin_resp ~ condition", model_name = "parallel design")
tmp$formula

istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = TRUE,
  weight_missing = 0.5, seed = 3)
istar$data <- encode_bin_resp(LFQData$new(istar$data, istar$config))
istar$config$bin_resp <- "bin_resp"
istar <- LFQData$new(istar$data, istar$config)
df <- istar$summarize_hierarchy()
df2 <- df[df[[ncol(df)]] > 1, ]
istar2 <- istar$get_subset(df2)
istar2$data_long() |>
  dplyr::group_by(protein_Id) |>
  tidyr::nest() -> nestProtein
modelFunction <- strategy_logistf("bin_resp ~ group_ + peptide_Id",
  model_name = "random_example")
modelFunction$model_fun(nestProtein$data[[1]])
modelFunction$model_fun(nestProtein$data[[4]])
tmp <- strategy_rfit("Intensity ~ condition", model_name = "parallel design")
tmp$formula
modelFunction <- strategy_lmer("abundanceC ~ group_ + (1|peptide_Id)",
  model_name = "random_example")
modelFunction$formula
tmp <- strategy_lm("Intensity ~ condition", model_name = "parallel design")
tmp$formula
tmp$weights
tmp <- strategy_rlm("Intensity ~ condition", model_name = "parallel design")
tmp$formula

Quasibinomial detection-count strategy

Description

Quasibinomial detection-count strategy

Details

Fits detected and undetected child-feature counts for each parent feature using glm with a quasibinomial family. The symmetric pseudo-count stabilizes fits under complete separation; it is not equivalent to Firth's bias-reducing penalty.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyBinomial

Public fields

prior_count: symmetric pseudo-count added to both outcomes

Methods

Public methods

StrategyBinomial$new()
StrategyBinomial$clone()

Inherited methods

prolfqua::StrategyBase$contrast_fun()
prolfqua::StrategyBase$df_residual()
prolfqua::StrategyBase$isSingular()
prolfqua::StrategyBase$model_fun()
prolfqua::StrategyBase$sigma()

Method `new()`

Create a quasibinomial count strategy. sigma() is the Pearson residual scale used by vcov().

Usage

StrategyBinomial$new(
  modelstr,
  prior_count = 0.1,
  model_name = "binomial_nested"
)

Arguments

modelstr: right-hand-side model formula, for example "~ group_"
prior_count: non-negative symmetric pseudo-count
model_name: model identity

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyBinomial$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

dat <- data.frame(
  group_ = factor(rep(c("A", "B"), each = 4)),
  detected = c(1, 2, 1, 3, 4, 5, 3, 5),
  undetected = c(4, 3, 4, 2, 1, 0, 2, 0)
)
strategy <- StrategyBinomial$new("~ group_")
fit <- strategy$model_fun(dat)
coefficients(fit)
dat <- data.frame(
  group_ = factor(rep(c("A", "B"), each = 4)),
  detected = c(1, 2, 1, 3, 4, 5, 3, 5),
  undetected = c(4, 3, 4, 2, 1, 0, 2, 0)
)
strategy <- StrategyBinomial$new("~ group_")
fit <- strategy$model_fun(dat)
coefficients(fit)

R6 class for limma modelling strategy

Description

R6 class for limma modelling strategy

Details

Analogous to strategy_lm but for limma's matrix-based pipeline. Consumed by build_model_limma.

Value

An R6 class generator.

Public fields

formula: model formula
model_name: name of model
trend: logical, passed to eBayes
robust: logical, passed to eBayes
weights: either a character string (column name) or a numeric matrix

Methods

Public methods

StrategyLimma$new()
StrategyLimma$clone()

Method `new()`

Create a new StrategyLimma

Usage

StrategyLimma$new(
  modelstr,
  model_name = "limma",
  trend = FALSE,
  robust = FALSE,
  weights = NULL
)

Arguments

modelstr, model_name, trend, robust, weights: see strategy_limma

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyLimma$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyLimma$new("abundance ~ group_")
strat$formula
strat$model_name
strat <- StrategyLimma$new("abundance ~ group_")
strat$formula
strat$model_name

R6 class for limpa modelling strategy

Description

R6 class for limpa modelling strategy

Details

Consumed by build_model_limpa. Configures the limpa::voomaLmFitWithImputation call: formula for the design matrix, trend/robust flags for eBayes, and optional vooma plotting parameters.

Value

An R6 class generator.

Public fields

formula: model formula
model_name: name of model
trend: logical, passed to eBayes
robust: logical, passed to eBayes
plot: logical, plot the vooma mean-variance trend
span: lowess smoother span for vooma trend (NULL = auto)

Methods

Public methods

StrategyLimpa$new()
StrategyLimpa$clone()

Method `new()`

Create a new StrategyLimpa

Usage

StrategyLimpa$new(
  modelstr,
  model_name = "limpa",
  trend = FALSE,
  robust = FALSE,
  plot = FALSE,
  span = NULL
)

Arguments

modelstr, model_name, trend, robust, plot, span: see strategy_limpa

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyLimpa$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyLimpa$new("abundance ~ group_")
strat$formula
strat$model_name
strat <- StrategyLimpa$new("abundance ~ group_")
strat$formula
strat$model_name

Linear model strategy (R6 class)

Description

Linear model strategy (R6 class)

Details

Encapsulates everything needed to fit per-protein linear models and extract contrasts: the formula, model fitting function, singularity check, contrast computation, ANOVA, and residual statistics.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyLM

Public fields

weights: optional character string naming a column in the data containing per-observation weights, passed to lm.

Methods

Public methods

StrategyLM$new()
StrategyLM$clone()

Inherited methods

prolfqua::StrategyBase$contrast_fun()
prolfqua::StrategyBase$df_residual()
prolfqua::StrategyBase$isSingular()
prolfqua::StrategyBase$model_fun()
prolfqua::StrategyBase$sigma()

Method `new()`

Create a new StrategyLM

Usage

StrategyLM$new(modelstr, model_name = "Model", weights = NULL)

Arguments

modelstr: model formula string
model_name: name of model
weights: optional character string naming a column in the data containing per-observation weights

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyLM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyLM$new("Intensity ~ condition", model_name = "parallel design")
strat$formula
strat$weights
strat <- StrategyLM$new("Intensity ~ condition", model_name = "parallel design")
strat$formula
strat$weights

Linear mixed-effects model strategy (R6 class)

Description

Linear mixed-effects model strategy (R6 class)

Details

Encapsulates everything needed to fit per-protein linear mixed-effects models via lmer and extract contrasts.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyLmer

Methods

Public methods

StrategyLmer$isSingular()
StrategyLmer$contrast_fun()
StrategyLmer$clone()

Inherited methods

prolfqua::StrategyBase$df_residual()
prolfqua::StrategyBase$initialize()
prolfqua::StrategyBase$model_fun()
prolfqua::StrategyBase$sigma()

Method `isSingular()`

Check if model is singular

Usage

StrategyLmer$isSingular(model)

Arguments

model: fitted model

Method `contrast_fun()`

Compute contrasts from fitted model

Usage

StrategyLmer$contrast_fun(...)

Arguments

...: passed to compute_lmer_contrast

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyLmer$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = FALSE)
istar <- prolfqua::LFQData$new(istar$data, istar$config)
istar$set_data(istar$data_long() |> dplyr::group_by(protein_Id) |>
  dplyr::mutate(abundanceC = abundance - mean(abundance)) |> dplyr::ungroup())
strat <- StrategyLmer$new("abundanceC ~ group_ + (1|peptide_Id)",
  model_name = "random_example")
mod <- build_model(istar, strat)
sum(mod$model_df$has_model_fit)
istar <- prolfqua::sim_lfq_data_peptide_config(Nprot = 10, with_missing = FALSE)
istar <- prolfqua::LFQData$new(istar$data, istar$config)
istar$set_data(istar$data_long() |> dplyr::group_by(protein_Id) |>
  dplyr::mutate(abundanceC = abundance - mean(abundance)) |> dplyr::ungroup())
strat <- StrategyLmer$new("abundanceC ~ group_ + (1|peptide_Id)",
  model_name = "random_example")
mod <- build_model(istar, strat)
sum(mod$model_df$has_model_fit)

Firth's logistic regression strategy (R6 class)

Description

Firth's logistic regression strategy (R6 class)

Details

Encapsulates everything needed to fit per-protein Firth's bias-reduced logistic regression via logistf and extract contrasts.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyLogistf

Methods

Public methods

StrategyLogistf$new()
StrategyLogistf$df_residual()
StrategyLogistf$sigma()
StrategyLogistf$clone()

Inherited methods

prolfqua::StrategyBase$contrast_fun()
prolfqua::StrategyBase$isSingular()
prolfqua::StrategyBase$model_fun()

Method `new()`

Create a new StrategyLogistf

Usage

StrategyLogistf$new(modelstr, model_name = "logistf")

Arguments

modelstr: model formula string
model_name: name of model

Method `df_residual()`

Get residual degrees of freedom

Usage

StrategyLogistf$df_residual(model)

Arguments

model: fitted model

Method `sigma()`

Get residual standard error (always 1 for logistic)

Usage

StrategyLogistf$sigma(model)

Arguments

model: fitted model

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyLogistf$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyLogistf$new("bin_resp ~ condition")
strat$formula
strat <- StrategyLogistf$new("bin_resp ~ condition")
strat$formula

Rank-based regression strategy (R6 class)

Description

Rank-based regression strategy (R6 class)

Details

Encapsulates everything needed to fit per-protein rank-based linear models via rfit and extract Wald contrasts. The fit is augmented (model frame, terms, rfit_prolfqua subclass) so it satisfies the same coef() / vcov() / terms() contract as lm.

Unlike StrategyLM, rfit takes no observation weights, so nr_children weighting is not supported by this backend.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyRfit

Methods

Public methods

StrategyRfit$new()
StrategyRfit$clone()

Inherited methods

prolfqua::StrategyBase$contrast_fun()
prolfqua::StrategyBase$df_residual()
prolfqua::StrategyBase$isSingular()
prolfqua::StrategyBase$model_fun()
prolfqua::StrategyBase$sigma()

Method `new()`

Create a new StrategyRfit. Residual df and scale come from df.residual() and sigma() of the augmented fit.

Usage

StrategyRfit$new(modelstr, model_name = "rfit")

Arguments

modelstr: model formula string
model_name: name of model

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyRfit$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyRfit$new("Intensity ~ condition", model_name = "parallel design")
strat$formula
strat <- StrategyRfit$new("Intensity ~ condition", model_name = "parallel design")
strat$formula

Robust linear model strategy (R6 class)

Description

Robust linear model strategy (R6 class)

Details

Encapsulates everything needed to fit per-protein robust linear models via rlm and extract contrasts.

Value

An R6 class generator.

Super class

prolfqua::StrategyBase -> StrategyRLM

Methods

Public methods

StrategyRLM$sigma()
StrategyRLM$clone()

Inherited methods

prolfqua::StrategyBase$contrast_fun()
prolfqua::StrategyBase$df_residual()
prolfqua::StrategyBase$initialize()
prolfqua::StrategyBase$isSingular()
prolfqua::StrategyBase$model_fun()

Method `sigma()`

Get the robust scale estimate used by vcov()

Returns model$s, the robust scale MASS::rlm builds its vcov() / summary() standard errors from. This must match the scale embedded in vcov() so that ContrastsModerated's sigma / sqrt(var.post) rescaling stays coherent; stats::sigma() returns the ordinary-residual scale instead and would distort moderated rlm statistics.

Usage

StrategyRLM$sigma(model)

Arguments

model: fitted model

Method `clone()`

The objects of this class are cloneable with this method.

Usage

StrategyRLM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

strat <- StrategyRLM$new("Intensity ~ condition", model_name = "parallel design")
strat$formula
strat <- StrategyRLM$new("Intensity ~ condition", model_name = "parallel design")
strat$formula

compute var sd etc for all factor levels

Description

compute var sd etc for all factor levels

Usage

summarize_stats_factors(lfqdata)
summarize_stats_factors(lfqdata)

Arguments

lfqdata

LFQData object

Value

The computed result.

Examples

res2 <- prolfqua::sim_lfq_data_2factor_config()
res2$config$factor_depth <- 2
lfq2 <- LFQData$new(res2$data, res2$config)
xx <- summarize_stats_factors(lfq2)
stopifnot(nrow(xx) == 80)
stopifnot(length(unique(xx$interaction)) == (2 + 2 + 2 * 2))
res2 <- prolfqua::sim_lfq_data_2factor_config()
res2$config$factor_depth <- 2
lfq2 <- LFQData$new(res2$data, res2$config)
xx <- summarize_stats_factors(lfq2)
stopifnot(nrow(xx) == 80)
stopifnot(length(unique(xx$interaction)) == (2 + 2 + 2 * 2))

table facade to easily switch implementations

Description

table facade to easily switch implementations

Usage

table_facade(df, caption, digits = getOption("digits"), kable = TRUE)
table_facade(df, caption, digits = getOption("digits"), kable = TRUE)

Arguments

df

data.frame to display

caption

table caption

digits

number of digits (default from options)

kable

if TRUE use knitr::kable

Value

The requested plot, table, or transformed object.

Examples

table_facade(head(iris, 3), caption = "Iris preview")
table_facade(head(iris, 3), caption = "Iris preview")

UpSetR plot from interaction_missing_stats

Description

UpSetR plot from interaction_missing_stats

Usage

upset_interaction_missing_stats(lfqdata, tr = 2)
upset_interaction_missing_stats(lfqdata, tr = 2)

Arguments

lfqdata

LFQData object

tr

if less than tr observations in condition then missing

Value

The computed result.

Examples

istar <- sim_lfq_data_peptide_config()
lfq <- LFQData$new(istar$data, istar$config)
pups <- upset_interaction_missing_stats(lfq)
stopifnot(ncol(pups$data) == 5)
UpSetR::upset(pups$data, order.by = "freq", nsets = pups$nsets)
istar <- sim_lfq_data_peptide_config()
lfq <- LFQData$new(istar$data, istar$config)
pups <- upset_interaction_missing_stats(lfq)
stopifnot(ncol(pups$data) == 5)
UpSetR::upset(pups$data, order.by = "freq", nsets = pups$nsets)

Named variance-covariance matrix for an augmented rfit fit

Description

'Rfit::vcov.rfit' returns an unnamed matrix; 'compute_contrast()' indexes the covariance by coefficient name, so this method restores the dimnames from 'coef()'.

Usage

## S3 method for class 'rfit_prolfqua'
vcov(object, ...)
## S3 method for class 'rfit_prolfqua'
vcov(object, ...)

Arguments

object

an 'rfit_prolfqua' fit

...

passed to the next method

Value

the rfit covariance matrix with row/column names set to the coefficient names.

Examples

fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
vcov(fit)
fit <- strategy_rfit("Sepal.Length ~ Species")$model_fun(iris)
vcov(fit)

volcano plotly

Description

volcano plotly

Usage

volcano_plotly(
  .data,
  effect = "fc",
  significance = "BFDR",
  contrast = "condition",
  proteinID = "Prey",
  color = "modelName",
  palette = NULL,
  xintercept = c(-2, 2),
  yintercept = 0.1,
  minsignificance = NULL,
  title_size = 25,
  group = "BB"
)
volcano_plotly(
  .data,
  effect = "fc",
  significance = "BFDR",
  contrast = "condition",
  proteinID = "Prey",
  color = "modelName",
  palette = NULL,
  xintercept = c(-2, 2),
  yintercept = 0.1,
  minsignificance = NULL,
  title_size = 25,
  group = "BB"
)

Arguments

.data

data frame

effect

effect size x-axis

significance

significance

contrast

column with contrast labels

proteinID

column with protein ids

color

column used for colouring points

palette

named colour vector for the colour aesthetic

xintercept

vertical abline at x

yintercept

horizontal abline at y

minsignificance

optional minimum significance value (floor for -log10 axis). If NULL, only exact zero and negative values are replaced with the smallest positive observed value in the same score column.

title_size

font size of the subplot title annotation

group

crosstalk group name for linked brushing

Value

The requested plot, table, or transformed object.

Examples


data <- data.frame(fc = c(-1,0,1,2,8), BFDR = c(0.01,1, 0.01, 0.005,0),
condition = rep("A",5), Prey = LETTERS[1:5], modelName = c("A","A","B","A","A"))

dataB <- data.frame(fc = c(-1,0,1,2,8), BFDR = c(0.01,1, 0.01, 0.005,0),
condition = rep("B",5), Prey = LETTERS[1:5],modelName = c("A","A","B","B","B"))
data <- dplyr::bind_rows(data, dataB)
bc <- volcano_plotly(data, xintercept = 1, yintercept= 0.01, palette = c(A = "black" , B = "red"))
bc |> plotly::subplot()

data <- data.frame(fc = c(-1,0,1,2,8), BFDR = c(0.01,1, 0.01, 0.005,0),
condition = rep("A",5), Prey = LETTERS[1:5], modelName = c("A","A","B","A","A"))

dataB <- data.frame(fc = c(-1,0,1,2,8), BFDR = c(0.01,1, 0.01, 0.005,0),
condition = rep("B",5), Prey = LETTERS[1:5],modelName = c("A","A","B","B","B"))
data <- dplyr::bind_rows(data, dataB)
bc <- volcano_plotly(data, xintercept = 1, yintercept= 0.01, palette = c(A = "black" , B = "red"))
bc |> plotly::subplot()

add missing values to x vector based on the values of x

Description

add missing values to x vector based on the values of x

Usage

which_missing(x, weight_missing = 0.2)
which_missing(x, weight_missing = 0.2)

Arguments

x

vector of intensities

weight_missing

greater weight more missing

Value

A logical vector indicating which values are missing.

Examples

which_missing(2**rnorm(10,2,0.4))

which_missing(2**rnorm(10,2,0.4))

Package 'prolfqua'

Help Index

adjust columns

Description

Usage

Arguments

Value

Examples

AggregateLimpa

Description

Details

Value

Super class

Public fields

Methods

Public methods

Method new()

Usage

Arguments

Method aggregate()

Usage

Returns

Method plot()

Usage

Arguments

Returns

Method write_plots()

Usage

Arguments

Returns

Method clone()

Usage

Arguments

See Also

Examples

AggregateMedpolish

Description

Details

Value

Super class

Methods

Public methods

Method aggregate()

Usage

Returns

Method clone()

Usage

Arguments

See Also

Examples

AggregateRlm

Description

Details

Value

Super class

Methods

Public methods

Method aggregate()

Usage

Returns

Method clone()

Usage

Arguments

See Also

Examples

AggregateTopN

Description

Details

Value

Super class

Public fields

Methods

Public methods

Method new()

Usage

Arguments

Method aggregate()

Usage

Returns

Method clone()

Method `new()`

Method `aggregate()`

Method `plot()`

Method `write_plots()`

Method `clone()`

Method `aggregate()`

Method `clone()`

Method `aggregate()`

Method `clone()`

Method `new()`

Method `aggregate()`

Method `clone()`

Method `set_response()`

Method `get_response()`

Method `pop_response()`

Method `factor_keys()`

Method `factor_keys_depth()`

Method `hierarchy_keys()`

Method `hierarchy_keys_depth()`

Method `id_vars()`

Method `value_vars()`

Method `annotation_vars()`

Method `clone()`

Method `new()`

Method `fun()`

Method `clone()`