Salmon quantification vs HTSeq count quantification

Question

[pablo]

Hi guys,

I've got a question. Can we compare the quantification files from Salmon and from HTseq-count?

I mean, Salmon quantifies reads over transcripts and HTSeq-count can quantify reads over exons. So, is it meaningful to say : "the exon AT5G00000.exon.1 gets 14 reads (HTSeq count) and the isoform AT5G00000.1 get 114 reads (Salmon) , then, these tools quantify differently and there is a significant difference between reads estimated by Salmon and HTSeq". ??

Best, Vincent

[GenoMax]

Hello vincentpailler ,

Don't forget to follow up on your threads.

If an answer was helpful, you should upvote it; if the answer resolved your question, you should mark it as accepted. You can accept more than one if they work.

Please do the same for your previous posts as well.

Answer 1

[Rob]

The way you have described it, the answer is emphatically "no". You cannot compare transcript-level abundance estimates from salmon (or any other transcript-level abundance estimation tool) to exon-level quantifications from HTSeq in a meaningful way. The most reasonable comparison you could make would be to quantify _gene-level_ expression in HTSeq, and then use a package like tximport to aggregate the transcript-level abundances from salmon to the gene level. Then, you could compare the gene-level abundances (read counts) reported by the different methods. One would still expected differences for certain genes since resolving ambiguity at the transcript level and then aggregating abundances to the gene level is typically more accurate than simply counting reads hitting genes (since, for example, the latter method either discards or uniformly allocates genomically multimapping reads).

[pablo]

I was meaning, if I want to compare the log2foldchange between Salmon and the reads counted by HTSeq-count, is it meaningful ? (Since HTSeq count estimates reads over exons and Salmon over transcripts)

[Rob]

If you compare log2 fold change of exons to log2 fold change of transcripts, then no, it is _not_ meaningful. To have a meaningful comparison, you must be comparing the same targets of quantification. You cannot compare exon read counts to transcript read counts (or the fold-change of exon read counts to the fold-change of transcript read counts).

Answer 2

[h.mon]

No you can't do this. An exon and an isoform are not the same thing, an isoform is composed of many exons and most likely will have higher counts than a particular exon of the same gene.

Do you think exon AT5G00000.exon.1 and transcript AT5G00000.1 are equivalent because they have the same digit?