Dear Gordon and dear all,
I'm using the voom transformation on RSEM "raw" counts from TCGA (RNASeq experiments). To avoid problems associated with zero values, I added +1 to the original read counts.
The problem is that count distributions are shifted towards negative values, after voom transformation. And then if I perform a two-class comparison, I obtain some differentially expressed genes (with statistically significant p.values) that have negative values in both classes... I really don't know how to manage these negative values.
The RSEM log2 raw counts have the following distribution (I just selected 10 samples):
After voom transformation, the distributions are shifted towards negative values.
The code I used is the following:
library (limma) library (edgeR) a <- read.table ("RSEM_genes_data_raw_count+1.txt", header=T,sep="\t") dim(a) counts <- as.matrix(a[,2:11]) rownames(counts) <- a$gene_id design=cbind(NH=c(1,1,1,1,1,0,0,0,0,0),NL=c(0,0,0,0,0,1,1,1,1,1)) v <- voom (counts, design, plot=TRUE)
Could you please help me in solving my problem?
Thank you very much in advance