Get the RPKM value of the genes analyzed using DESeq or edgeR

[narges]

Hi,

I have done analyzation over RNA seq data using edgeR and DESeq to find DE genes (BAM files -> HTSeq -> edgeR and DEseq).
For some comparisons I need to have the RPKM values related to each gene. What is the best way of getting it?

Thank you in advance.

[Gordon Smyth]

The version of edgeR on the Bioconductor developmental repository has a function rpkm().

[ThePresident]

Originally Posted by Gordon Smyth

The version of edgeR on the Bioconductor developmental repository has a function rpkm().

Is it v3.08?

[Gordon Smyth]

Originally Posted by ThePresident

Is it v3.08?

No, the *development* repository:

http://bioconductor.org/packages/2.1...tml/edgeR.html

[narges]

Originally Posted by Gordon Smyth

The version of edgeR on the Bioconductor developmental repository has a function rpkm().

Thank you but can I ask how does this function calculate the gene length? Because my problem is that I do not know how to get the gene length to calculate the RPKM values. The gtf file I have used is the hg19 latest version from UCSC genome browser. I can do something like this: endposition - startpositoin+1. But there are different transcripts for each gene. which of these transcripts should be the as the source for the gene length? the average? the longest?

[sindrle]

Im wondering the exact same thing..

[Gordon Smyth]

An appropriate measure of gene length must be input to rpkm(). Computing gene length is a job for the read count software rather than for the differential expression software because the appropriate measure of gene length depends on the way the reads have been counted.

I use subread and featureCounts:

http://www.ncbi.nlm.nih.gov/pubmed/24227677

to count reads. For most RNA-seq analyses, I count reads that overlap any exon for each gene, so the appropriate measure of gene length is the total exon length. Gene length is returned as part of the output from featureCounts.

[sindrle]

Thanks!

Is there a difference in choosing CPM/logCPM/FPKM to represent gene expression level if I want to correlate the expression of one gene against i.e. body weight?

Or the change in gene expression of one gene against change in body weight etc.

[sindrle]

Originally Posted by Gordon Smyth

An appropriate measure of gene length must be input to rpkm(). Computing gene length is a job for the read count software rather than for the differential expression software because the appropriate measure of gene length depends on the way the reads have been counted.

I use subread and featureCounts:

http://www.ncbi.nlm.nih.gov/pubmed/24227677

to count reads. For most RNA-seq analyses, I count reads that overlap any exon for each gene, so the appropriate measure of gene length is the total exon length. Gene length is returned as part of the output from featureCounts.

I have used HTSeq for counting, I guess I can run featureCounts on one sample, just to get the gene lengths from my UCSC GTF file.

Then import the lengths to edgeR/DEseq fpkm()

Or am I missing some points here?