1
$\begingroup$

I have a data set of de novo assembled RNA-seq datasets across different sample types.

When BLASTing, many of the matches of the individual transcripts match to the same gene on the reference genome. However, each individual transcript has its own unique FPKM value.

I'm confused, firstly, as to how you can have multiple sequences from the same gene with different FPKM values — and of course, I'm also wondering what would be a suitable approach for the subsequent analysis. Should I just add the FPKM values together for the sequences with the same matches?

$\endgroup$
1
$\begingroup$

If the quantification is done at the transcript level, each identified transcript of a given gene will have a different number of reads attributed to it, hence a different RPKM value.

For the subsequent analysis, you can continue at the transcript level.

I think you cannot sum the FPKM values directly, because they are inversely proportional to the transcript lengths ("K" stands for "by kilobase"). If you want to do the subsequent analysis at the gene level and use FPKM values, you would have to multiply the FPKM values of the transcripts by the corresponding transcript lengths before summing. Then you will have to divide this sum by the gene length, or something like that.

$\endgroup$
  • $\begingroup$ Indeed you can't sum FPKMs, and nowadays you should probably try to void using them in the first place (they have some bias, but are still widely used). But even if you get the raw counts per transcript, you should not directly sum them to get counts per genes. If you got your counts using Kallisto or Salmon, the easiest way is to either use sleuth to work directly on transcripts, or tximport to sum counts and input them into DESeq, edgeR or another program. $\endgroup$ – Alexlok Apr 2 '17 at 19:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.