| Title: | Replicate-based Imputation for Randomly Missing Data in Quantitative Proteomics |
|---|---|
| Description: | Implements replicate-based imputation for missing at random (MAR) values in proteomics data. The method leverages existing measurements from other replicates to determine imputation parameters (mean of observed values and global standard deviation). |
| Authors: | Dick Zijlmans [aut, cre] (ORCID: <https://orcid.org/0000-0003-1527-1834>) |
| Maintainer: | Dick Zijlmans <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.99.0 |
| Built: | 2026-11-12 16:40:39 UTC |
| Source: | https://github.com/BiocStaging/repli |
A small subset of a proteomics dataset used to demonstrate replicate-based imputation.
data(example_dataset)data(example_dataset)
A data frame containing protein abundances and experimental conditions in columns. Protein names are set as row names.
Data derived from a subset of the dataset reported in doi:10.1016/j.celrep.2025.115362.
data(example_dataset) head(example_dataset)data(example_dataset) head(example_dataset)
A SummarizedExperiment object containing the example proteomics dataset and experimental design information.
data(example_se)data(example_se)
A SummarizedExperiment object containing protein abundance
measurements and experimental metadata.
Data derived from a subset of the dataset reported in doi:10.1016/j.celrep.2025.115362.
data(example_se) example_sedata(example_se) example_se
Performs replicate-based imputation of Missing At Random (MAR) values. Missing values are imputed by sampling from a normal distribution using the mean of the observed replicates for the corresponding protein-condition pair and a global standard deviation derived from replicates of all protein-condition pairs across the dataset.
impute_replicate(prepared_data)impute_replicate(prepared_data)
prepared_data |
An object returned by
|
A data frame containing imputed data. MAR values are imputed, while values classified as MNAR or observed are left unchanged.
#in this instance data is a data frame with proteins as rows and #experimental conditions as columns. Protein names are set as rownames. data(example_dataset) prepared <- prepare_imputation_data(example_dataset) imputed <- impute_replicate(prepared)#in this instance data is a data frame with proteins as rows and #experimental conditions as columns. Protein names are set as rownames. data(example_dataset) prepared <- prepare_imputation_data(example_dataset) imputed <- impute_replicate(prepared)
Performs mixed imputation of missing values in a SummarizedExperiment object. Missing At Random (MAR) values are imputed using replicate-based imputation, while Missing Not At Random (MNAR) values are imputed using a specified method from the DEP R package. Both methods are applied independently to the original dataset, after which their results are combined.
impute_replicate_mixed_se( se, p = 0.05, MNAR_method = "MinProb", MNAR_args = list(q = 0.01) )impute_replicate_mixed_se( se, p = 0.05, MNAR_method = "MinProb", MNAR_args = list(q = 0.01) )
se |
A |
p |
A numeric value between 0 and 1 specifying the proportion of
proteins with the highest standard deviation across replicates, used to
estimate the global standard deviation. Default is |
MNAR_method |
A character string specifying the DEP imputation method
used for MNAR values, passed to |
MNAR_args |
A named list of additional arguments passed to
|
Protein-condition pairs are classified as MAR if there are missing values for some but not all replicates. Protein-condition pairs are classified as MNAR if there are missing values for all of the replicates of a condition.
#' MAR and MNAR imputation are performed independently on the original dataset. Missing MAR values are imputed using replicate-based imputation, while missing MNAR values are imputed using the specified DEP method. The results are subsequently combined, retaining all originally observed values.
When MNAR_method = "none", MNAR values are left missing.
A SummarizedExperiment object containing the imputed assay. MAR values are imputed using replicate-based imputation and MNAR values are imputed using the specified DEP method.
data(example_se) imputed_se <- impute_replicate_mixed_se( example_se, p = 0.05, MNAR_method = "MinProb", MNAR_args = list(q = 0.01) )data(example_se) imputed_se <- impute_replicate_mixed_se( example_se, p = 0.05, MNAR_method = "MinProb", MNAR_args = list(q = 0.01) )
Performs replicate-based imputation of Missing At Random (MAR) values
in a SummarizedExperiment object. Missing values are imputed by sampling
from a normal distribution using the mean of the observed replicates for the
corresponding protein-condition pair and a global standard deviation derived
from replicates of all protein-condition pairs across the dataset.The assay
is prepared using prepare_imputation_data_se(), imputation is performed
using impute_replicate(), and the imputed values are returned in the
original SummarizedExperiment structure using reconstruct_se().
impute_replicate_se(se, p = 0.05)impute_replicate_se(se, p = 0.05)
se |
A |
p |
A numeric value between 0 and 1 specifying the proportion of
proteins with the highest standard deviation across replicates, used to
estimate the global standard deviation. Default is |
A SummarizedExperiment object with MAR values imputed. Completely missing protein-condition pairs, classified as MNAR, are not imputed.
data(example_se) imputed_se <- impute_replicate_se(example_se)data(example_se) imputed_se <- impute_replicate_se(example_se)
Prepares a data frame containing protein abundance values for replicate-based imputation. Data is transformed in long format, with protein-condition pairs as rows and replicates as columns. Missingness is classified as Missing At Random (MAR), Missing Not At Random (MNAR) or observed. For protein-condition pairs classified as MAR, the mean abundance is calculated. A global standard deviation is derived from the standard deviations of replicates of all protein-condition pairs across the dataset.
prepare_imputation_data(data, p = 0.05)prepare_imputation_data(data, p = 0.05)
data |
A data frame containing protein abundance values, with protein
names as row names and experimental conditions as column names. Column names
should follow condition-replicate naming convention such as
|
p |
A numeric value between 0 and 1 specifying the proportion of
proteins with the highest standard deviation across replicates, used to
estimate the global standard deviation. Default is |
Protein-condition pairs are classified as MAR if there are missing values for some but not all replicates. Protein-condition pairs are classified as MNAR if there are missing values for all of the replicates of a condition.
A list containing:
The data in a format prepared for imputation, including missingness classification and protein-condition means.
A vector specifing the replicate columns.
The global standard deviation among the proteins
in the upper p proportion of replicate variability.
data(example_dataset) prepared <- prepare_imputation_data(example_dataset)data(example_dataset) prepared <- prepare_imputation_data(example_dataset)
Extracts abundance data from a SummarizedExperiment and prepares it
for replicate-based imputation using prepare_imputation_data().
The experimental design is also extracted from the sample names for
downstream reconstruction of the SummarizedExperiment.
prepare_imputation_data_se(se, p = 0.05)prepare_imputation_data_se(se, p = 0.05)
se |
A |
p |
A numeric value between 0 and 1 specifying the proportion of
proteins with the highest standard deviation across replicates, used to
estimate the global standard deviation. Default is |
A list containing:
The data in a format prepared for imputation, including missingness classification and protein-condition means.
A vector specifing the replicate columns.
The global standard deviation among the proteins
in the upper p proportion of replicate variability.
A data frame describing the experimental
design, containing columns label, condition, and
replicate
data(example_se) prepared <- prepare_imputation_data_se(example_se) imputed <- impute_replicate(prepared)data(example_se) prepared <- prepare_imputation_data_se(example_se) imputed <- impute_replicate(prepared)
Places imputed data from impute_replicate() back into the original
SummarizedExperiment object supplied in prepare_imputation_data_se().
reconstruct_se(imputed_data, se, experimental_design)reconstruct_se(imputed_data, se, experimental_design)
imputed_data |
A data frame containing the imputed data returned
by |
se |
The original SummarizedExperiment object supplied in
|
experimental_design |
A data frame describing the experimental
design, containing columns |
A SummarizedExperiment object containing the imputed assay. The original row and column metadata are retained.
data(example_se) prepared <- prepare_imputation_data_se(example_se) imputed <- impute_replicate(prepared) result <- reconstruct_se( imputed, example_se, prepared$experimental_design )data(example_se) prepared <- prepare_imputation_data_se(example_se) imputed <- impute_replicate(prepared) result <- reconstruct_se( imputed, example_se, prepared$experimental_design )