mcImpute

A matlab tool for imputation of scRNA-seq data

Paper link: A. Mongia, D. Sengupta, and A. Majumdar, “Mcimpute:Matrix completion based imputation for single cell rna-seqdata,”Frontiers in genetics, 2019

Brief Description

McImpute takes raw gene expression Data (cells x genes) as input. It pre-processed the data (see paper for detailed steps) and applies nuclear norm minimization to recover the full expression data, filling in the dropouts.

Implementation

Running the script would read the count matrix, processes it, call mcImpute and place the recovered matrices in a folder 'RecoveredMatrices'

run.m

Testing

Fix seed for reproducibility

rng(0);

Get annotations (single cell types)

actual_labels=eval(['get_numeric_labels_' dataname '( dataname , data_dir )']) ;

Run kmeans clustering algorithm after applying PCA on scRNA-seq data

list_ni=[]; list_mc=[];
for (i=1:100)
loc=randperm(length(actual_labels),length(unique(actual_labels)));
load(['processedData/' dataname '_processed.mat']);
ari_ni=call_kmeans(processed_data,'PCA', loc ,actual_labels); list_ni=[list_ni ari_ni];
ari_mc=call_kmeans(data_recovered,'PCA', loc ,actual_labels); list_mc=[list_mc ari_mc];
end
avg_ari_ni=mean(list_ni)
avg_ari_mc=mean(list_mc)