In ChongC1990/scRMD: scRMD: Imputation for single cell RNA-seq data via robust matrix decomposition

Introduction

The emerging single cell RNA sequencing (scRNA-seq) technologies enable the investigation of transcriptomic landscape at single-cell resolution. However, scRNA-seq analysis is complicated by the excess of zero or near zero counts in the data, which are the so-called dropouts due to low amounts of mRNA within each individual cell. Consequently, downstream analysis of scRNA-seq woule be severely biased if the dropout events are not properly corrected. scRMD scRMD is developed to impute single cell RNA data with dropouts. scRMD assumes the the underlying expression profile of genes is low rank and the dropout events are rare compared with true zero expression.

scRMD can be applied to raw data count before the users perform downstream analyses such as

• dimension reduction of scRNA-seq data
• normalization of scRNA-seq data
• clustering of cell populations
• differential gene expression analysis
• time-series analysis of gene expression dynamics

Installation

scRMD can be installed by simplely run:

install.packages("devtools")         
library(devtools)           
install_github("ChongC1990/scRMD")

or

install.packages("scRMD")

Quick start

scRMD can be easily incorporated into existing pipeline of scRNA-seq analysis. Its only input is the raw expression matrix with rows representing genes and columns representing cells. It will output an imputed count matrix with the same dimension. In the simplest case, the imputation task can be done with one single function rmd:

rmd(Y, # The raw expression matrix
    tau = NULL, # Tuning parameter to penalize the sparsity of S
    lambda = NULL, # Tuning parameter to penalize the row rank of L
    initL = NULL, # The initionlization of L
    initS = NULL, # The initionlization of S
    initLambda = NULL, # The initionlization of Lambda
    maxiter = 100, # Maxmium iteration
    abstol = 0.001, # Convergence metrics
    reltol = 0.001, # Convergence metrics
    rho = 1, 
    overrelax = 1.5, 
    candidate = 0.05, # The cutoff to define candidate dropouts
    econ = 1 # If econ = 0, fast svd decomposition will be used
    )

library(scRMD)
set.seed(2017)
K=3; Kn=50; Ndiff=100; Nsame=10000; logMean=1.8; logSd=0.5; 
ZeroRate = 0.5; sigmahetero = 0.1; sigmahomo = 0.2; drbase = 1; dr = 0.2;
sData = sSimulator(K, Kn, Ndiff, Nsame, logMean, logSd, ZeroRate, drbase, dr, sigmahomo, sigmahetero, type = "cluster")
cutoff = quantile(sDate$de[sDate$de>0], 0.05)
res.rmd <- rmd(sData$de, candidate = cutoff)
pca.rmd <- prcomp(res.rmd$exprs)
cl.rmd <- kmeans(pca.rmd$x[,1:2],K,nstart = 100)
CalculateARI(sData$label, cl.rmd$cluster)

ChongC1990/scRMD documentation built on May 27, 2019, 4:05 a.m.