How to calculate TPM of transcripts with isoforms

Question

[marcosp.belem]

Hi, I used the Salmon tool for transcript quantification using as a reference transcriptome a "de novo" assembly (Trinity) and RNA-Seq sequences. I would like to calculate the TPM of the gene trim62 expressed below, according to the isoform results. In that case, the TPM of the trim62 gene is the simple sum of the transcripts of each isoform? Or should I choose the transcript with the best HIT (Blast)

GENE | TRANSCRIPTS | TPM | Num reads

trim62 | TRINITY_DN93372_c7_g1_i7 | 0,950919 | 684879

trim62 | TRINITY_DN93372_c7_g1_i3 | 0,0824602 | 607185

Answer 1

[Rob]

To aggregate abundance estimates to the gene-level, you should consider using a tool like tximport, and providing the transcript-to-gene mapping as input. Of course, the quality of the _de novo_ assembly can be a limiting factor in the accuracy of the abundance estimates, and you may want to remove assembled contigs that you believe may be assembly artifacts (though, in that case, I'd recommend re-quantifying abundances once you have filtered your contig set). However, if you believe all of these transcripts represent full or partial assemblies of distinct isoforms of the same gene, you should aggregate their abundances rather than pick a single representative.

[marcosp.belem]

I was careful to identify and remove artifacts. So I should sum of the TPMs of each transcript (isoforms), correct?

[Rob]

That's correct. However, I'd still recommend using tximport if you're going to be using these estimates for anything other than visualization downstream, since it can correct for things that aren't possible to do looking at each sample in isolation (like the average expected gene length across all samples in an experiment).