We use DESeq to normalize RNA-seq data for math. modeling purpose. Our concerns is the length bias. After reading DESeq paper, I think the data normalized using DESeq should be be more balanced than EdgeR since the estimate of dispersion is added. We need to calculate correlation between genes and construct gene regulatory network. What's your opinion on this. Questions: 1. How to arrange controls and treatments columns in a data set: WT-2 WT-3 WT-4 4-3-1 4-4-2 4-5-1 4-6-2 7-1-2 7-2-3 7-3-1 8-3-1 8-3-2 This is in one row, 3 wild types, 3 groups of replicates. There are 4 rep. in group 1 (4-3-1 4-4-2 4-5-1 4-6-2), 3 rep. in group 2 (7-1-2 7-2-3 7-3-1). 2 rep. in group 3 ( 8-3-1 8-3-2). 3 wild types are the controls for 3 groups. Similar to Multi-factor designs??? How should the data columns arranged? I only know conditions but not libType. The read counts/differential expression should be compared with its relevant controls. Even though the rep. are pooled for dispersion estimates, I would not think rep. in different groups are pooled together for estimates. 2. The output file contains statistical analysis results, I would think I should select the genes with padj < 0.05, but I don't see how to get the normalized counts, which shuld be decimal instead of integer. 3. MA plot, sometimes I could get the plot, sometimes I could not get it, don't know why? I just copied code from the documentation. plotDE <- function(res) plot(res$baseMean,res$log2FoldChange,log="x",pch=20,cex=.3,col=ifelse (res$padj < .1, "red","black")) This one doesn't work at all. plotDispEsts <- function(cds){ plot(rowMeans(counts(cds, normalized=TRUE)), fitInfo(cds)$perGeneDispEsts, pch = '.', log="xy") xg <- 10^seq(-.5,5,length.out=300) lines(xg,fitInfo(cds)$dispFun(xg),col="red")} Any guidance would be appreciated. Thanks, Xin [[alternative HTML version deleted]]