Package 'cellpaintr'

Aggregate predicted leave-one-out probabilities over meta variables

Description

Aggregate predicted leave-one-out probabilities over meta variables

Usage

aggregateYhat(sce, target, group, assay_type = "tfmfeatures")

Arguments

sce	SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
assay_type	A string specifying the assay

Value

data.frame

---

Aggregate predicted leave-one-out probabilities over meta variables over a list of SingleCellExperiment objects

Description

Aggregate predicted leave-one-out probabilities over meta variables over a list of SingleCellExperiment objects

Usage

calculateStats(sce, target, group, assay_type = "tfmfeatures")

Arguments

sce	A SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
assay_type	A string specifying the assay

Value

data.frame

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

calculateStats(sce_single, target = "Drug", group = "Patient")

---

Internal function to compute y_hat on a subset of the features

Description

Internal function to compute y_hat on a subset of the features

Usage

compute_y_hat(
  feature_name,
  sce_feature,
  starts,
  target,
  group,
  interest_level,
  reference_level,
  weights,
  n_threads,
  assay_type
)

Arguments

feature_name	String with feature name
sce_feature	SingleCellExperiment
starts	Starting string
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
interest_level	Factor interest level in 'target' variable
reference_level	Factor reference level in 'target' variable
weights	Weights variable when features are aggregated
n_threads	Number of parallel threads for fitting of models
assay_type	A string specifying the assay

Value

tibble data frame

---

Simulate CellProfiler data and write to a temporary file

Description

Simulate CellProfiler data and write to a temporary file

Usage

generate_data()

Value

path to csv file

Examples

set.seed(23)
generate_data()

---

Load cell painting data from file and convert to a SingleCellExperiment

Description

Load cell painting data from file and convert to a SingleCellExperiment

Usage

loadData(cell_file)

Arguments

cell_file

path to csv file from CellProfiler, e.g., MyExpt_Cells.csv

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
loadData(cell_file)

---

Plot AUC comparison

Description

Plot AUC comparison

Usage

plotAUC(sce, target, group, assay_type = "tfmfeatures")

Arguments

sce	A SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
assay_type	A string specifying the assay

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

plotAUC(sce_single, target = "Drug", group = "Patient")

---

Plot number of cells per image

Description

Plot number of cells per image

Usage

plotCellsPerImage(sce, bins = 100)

Arguments

sce	SingleCellExperiment object
bins	Number of histogram bins

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
plotCellsPerImage(sce)

---

Plot predicted leave-one-out probabilities

Description

Plot predicted leave-one-out probabilities

Usage

plotLOO(sce, target, group, assay_type = "tfmfeatures")

Arguments

sce	SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
assay_type	A string specifying the assay

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

plotLOO(sce_single, target = "Drug", group = "Patient")

---

Plot number of cells per image

Description

Plot number of cells per image

Usage

plotPCACor(
  sce,
  filter_by = 1,
  top = 20,
  pcs = seq(5),
  assay_type = "tfmfeatures"
)

Arguments

sce	SingleCellExperiment object
filter_by	PC to use for feature selection
top	Number of top features to select
pcs	Number of PCs to plot
assay_type	A string specifying the assay

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce, robust = TRUE)
sce <- scater::runPCA(sce, exprs_values = "tfmfeatures", ncomponents = 10)
plotPCACor(sce, filter_by = 1)

---

Plot ROC curves

Description

Plot ROC curves

Usage

plotROC(sce, target, group, assay_type = "tfmfeatures")

Arguments

sce	SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
assay_type	A string specifying the assay

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

plotROC(sce_single, target = "Drug", group = "Patient")

---

Predict target from features

Description

Predict target from features

Usage

predictLOO(
  sce,
  target,
  group,
  interest_level,
  reference_level,
  types = NULL,
  channels = NULL,
  weights = NULL,
  n_threads = 1,
  assay_type = "tfmfeatures"
)

Arguments

sce	SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
interest_level	Factor interest level in 'target' variable
reference_level	Factor reference level in 'target' variable
types	Vector of strings of feature types
channels	Vector of strings of feature channels
weights	Weights variable when features are aggregated
n_threads	Number of parallel threads for fitting of models
assay_type	A string specifying the assay

Value

tibble data frame

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

---

Filter low variance features

Description

Filter low variance features

Usage

removeLowVariance(sce, threshold = 0, robust = FALSE)

Arguments

sce	SingleCellExperiment object
threshold	Keep features that have larger variance than this threshold
robust	If true use median absolute deviation, otherwise use variance

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- removeLowVariance(sce)

---

Remove cells with missing features

Description

Remove cells with missing features

Usage

removeNAs(sce)

Arguments

sce	SingleCellExperiment object

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- removeNAs(sce)

---

Remove cells if not enough or too many in one image

Description

Remove cells if not enough or too many in one image

Usage

removeOutliers(sce, min, max)

Arguments

sce	SingleCellExperiment object
min	Remove cells below that number
max	Remove cells above that number

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- removeOutliers(sce, min = 0, max = 300)

---

Remove zero-inflated features

Description

Remove zero-inflated features

Usage

removeZeroInflation(sce, proportion = 0.2)

Arguments

sce	SingleCellExperiment object
proportion	Remove features exceeding this zero-inflation proportion

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- removeZeroInflation(sce)

---

Filter low variance features

Description

Filter low variance features

Usage

transformLogScale(sce, robust = FALSE)

Arguments

sce	SingleCellExperiment object
robust	If true robust z-score, otherwise standard z-score

Value

SingleCellExperiment object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

---

Plot predicted leave-one-out probabilities

Description

Plot predicted leave-one-out probabilities

Usage

volcanoPlot(
  sce,
  target,
  group,
  p_cutoff = NULL,
  fc_cutoff = 1,
  assay_type = "tfmfeatures"
)

Arguments

sce	A SingleCellExperiment object
target	Name of target variable for prediction
group	Grouping variable for cross-validation, e.g., patient
p_cutoff	Cut-off for statistical significance. A horizontal line will be drawn at -log10(p_cutoff).
fc_cutoff	Cut-off for absolute log2 fold-change. A vertical lines will be drawn at fc_cutoff.
assay_type	A string specifying the assay

Value

ggplot2 object

Examples

set.seed(23)
cell_file <- generate_data()
sce <- loadData(cell_file)
sce <- transformLogScale(sce)

sce$Drug <- as.factor(sce$Drug)
sce$Drug <- relevel(sce$Drug, ref = "D1")
types <- c("AreaShape", "Intensity", "Texture")

sce_single <- predictLOO(
    sce,
    target = "Drug", group = "Patient",
    interest_level = "D7", reference_level = "D1",
    types = types,
    n_threads = 1
)

volcanoPlot(sce_single,
    target = "Drug", group = "Patient",
    p_cutoff = 0.05, fc_cutoff = 0.5
)

---