one gene one RPKM cufflinks or DEseq

[frankyue50]

Hi, guys, I need your expert input. So, I am not worried about the isoforms, at least for now. I just want to assign a rpkm for each gene. From what I read here, DEseq should get the job done.

But I already have the cufflinks out and want to do something quick. Should I add the FPKM for different transcript or should I do something else? Thanks!

[Simon Anders]

DESeq is for comparing genes across samples, not for calculating RPKM.

Maybe you better first explain why you want RPKM values: what do you want to do with them once you have them?

Simon

[lpachter]

Yes, you can add the FPKM for individual transcripts in a gene to get an overall FPKM.

[carmeyeii]

If I were to un-normalize these FPKMs in order to get the raw reads that cufflinks assigned to each gene... what would the M in FPKM normalization factor be - the total number of alignments produced by tophat, or the unique number of reads that mapped at least one time? Or just the number of reads that mapped to the actual features, i.e, genes?

What subset of all the accepted_hits.sam file does cufflinks use to normalize the reads after accounting for transcript length?

Thanks,
Carmen

[Gonza]

Hi all, a question along the lines of the discussion:

I made a table count with htseq-count then run edgeR. Now I have a table with GeneID, logFC, logCPM, LR, PValue and FDR. Now I would like to add RPMK values to this table.

edgeR has an rpkm function (x, gene.length, normalized.lib.sizes=TRUE, log=FALSE, prior.count=0.25) but I know that figuring out 'gene.length' is a tough task (and I guess this is why is better to do transcript-based differential expression analysis as opposed to exon-based).

So what would be a good criteria to select gene.length for an RNA-seq using TAIR10 (Arabidopsis)??

Any ideas?

thanks
G


Thanks All
G

[dpryan]

Most people would just use a union exon model (i.e., add up the non-overlapping exons for each gene).