[BioC] edgeR - R script - results compared to DESeq

[Gordon K Smyth]

Hi Avinash,

Your analysis seems fine, from what you give, although I'd rather than you 
used estimateTagwiseDisp() as a general rule instead of just 
estimateCommonDisp().  One of the major motivations for the edgeR package 
is the ability to allow gene-specific variability without losing too much 
power, which is what the tagwise dispersion estimation does.  We also 
generally recommend normalization using calcNormFactors() prior to 
dispersion estimation and testing.  Either or both of these steps may 
affect the number of DE genes.

Best wishes
Gordon

> Date: Mon, 28 Nov 2011 09:58:48 -0600
> From: Avinash S <avins.s at googlemail.com>
> To: <bioconductor at stat.math.ethz.ch>
> Subject: [BioC] edgeR - R script - results compared to DESeq
>
> Hello List,
>
> I'm just starting with R and wanted to analyze my data for differential
> expression using edgeR. Here is the code which is working for me but I want
> to check if I'm missing something as I get more number of differentially
> expressed genes compared to DESeq
>
> MY Sample data
>
> GeneID        CR1 CR2 MR1 MR2
> 3119s00200.1 78 78 148 124
> 3119s00202.1    33 68 168 198
> 3119s00232.1     52 73 135 99
>
> CR1 and CR2 are replicates of controls and MR1 and MR2 are treated sample
> replicates
>
> library(edgeR)
> library(limma)
> raw.data <- read.delim("RNAseq-MycReadCountData-EdgeR.txt")
> names(raw.data)
>
> d <- raw.data[, 2:5]
> rownames(d) <- raw.data[, 1]
> group <- c(rep("CR", 2), rep("MR", 2))
> d <- DGEList(counts = d, group = group)
> d <- estimateCommonDisp(d)
> et <- exactTest(d)
> topTags(et)
> etTag = rownames(topTags(et)$table)
> sum(et$table$p.value <0.05)
> sum(p.adjust(et$table$p.value,method="BH") < 0.1)
> good = sum(et$table$p.value <0.05)
> goodList = topTags(et, n=good)
> sink("edgeR-MR-pVal005.csv")
> goodList
> sink()
>
> I compared the result with DESeq and I get about 2000 genes more in edgeR
> at pVal < 0.05, however, the the matched genes showed same log2foldchanges.
> Is it usual that edgeR gives more number of diff.expr genes?
>
> Thank you for your time,
> Avinash

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}