README.md

In siamakz/BNBR: Bayesian Negative Binomial Regression For Differential Expression Analysis

Bayesian Negative Binomial Regression for Differential Expression Analysis

This is a R package for differential expression analysis of seqencing count data. The main function NBregDE2 takes as input:

• counts: a matrix of counts, rows are corresponding to genes and columns are corresponding to samples
• X: design matrix
• cond: the index of covariate corresponding to the main treatment
• idx.cond: a vector of covariate indices that the effect of their combination on gene expression is under study
• Burnin: number of burn-in iterations in MCMC
• Collections: number of collected posterior samples after burn-in
• PGTruncation: the truncation level used for genrating random numbers from Polya-Gamma distribution
• randtry: to be used for set.seed()

Example

Given the gene count matrix and two covariates as sex and cond, the ranking of genes with respect to cond can be obtained as:

Burnin <- 1000
Collections <- 1000
X <- as.matrix(t(model.matrix(~sex+cond)))
idx.cond <- 3
res <- NBregDE2(dat,X,cond,idx.cond, Burnin,Collections)
res$kl            # the KL-divergence score of genes, high values correspond to high DE
res$beta          # posterior means of regression coefficient for the covariate of interest (i.e., cond)

siamakz/BNBR documentation built on Oct. 21, 2019, 5:50 a.m.