Question: Incorporating 'offset' into edgeR pipeline
gravatar for alakatos
15 months ago by
United States
alakatos90 wrote:

Hello All,

I would like to adjust my RNAseq dataset for GC content.

I used EDASeq to calculate the offset.


data <-newSeqExpressionSet(counts=as.matrix(d$counts),featureData=feature,phenoData=data.frame(pheno),row.names=rownames(d))
dataOffset <- withinLaneNormalization(data,"gc", which="full",offset=TRUE)
offset <- offst(dataOffset)

My goal is to incorporate “GC offset” into my edgeR pipeline  (not to override edgeR calculated offset).

I found some suggestions online but I am not sure I am doing right.

Large logCPM values when using offsets in edgeR for normalization

[bioc] correlation analysis of RNA-seq


d <- calcNormFactors(d, robust=TRUE, method ="TMM", offset=offset)
d <- estimateDisp(d, design,)
fit <- glmFit(d, design)



Would you please advise?

Thanks a lot,


edger edaseq offset • 288 views
ADD COMMENTlink modified 15 months ago by Aaron Lun23k • written 15 months ago by alakatos90
Answer: Incorporating 'offset' into edgeR pipeline
gravatar for Aaron Lun
15 months ago by
Aaron Lun23k
Cambridge, United Kingdom
Aaron Lun23k wrote:

Supplying offset to calcNormFactors will do nothing, see ?calcNormFactors.

If you need to supply offsets to edgeR, use the scaleOffset function to store the offset matrix inside the DGEList. This will ensure that the offsets are interpretable as log-library sizes (natural log), which is necessary for sensible calculation of the average log-CPM.

However, any specification of the offset matrix will override the use of normalization factors in downstream calculations. There is no straightforward way to "combine" offset matrices with the TMM normalization factors. edgeR can't possibly know what biases are removed by your offset matrix; it just assumes that you know what you're doing, and that the offset matrix gets rid of all biases that might be relevant to your contrasts of interest.

In short, supplying the EDAseq offset matrix will mean that only the EDAseq normalization will be performed. The TMM normalization factors will be mostly ignored for the purposes of DE testing.

ADD COMMENTlink modified 15 months ago • written 15 months ago by Aaron Lun23k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 210 users visited in the last hour