Why run FeatureCounts after Stringtie? (Galaxy recommends!)

Question

[c_u]

Hi,

I have RNA-Seq data from patients, and I want to find novel transcripts that are differentially expressed in the treatment condition. For this, I followed the methodology given in this Galaxy tutorial - https://galaxyproject.org/tutorials/nt_rnaseq/. Here they use HISAT for mapping, followed by Stringtie for transcript reconstruction (using Stringtie allows them to find novel transcripts), and then use FeatureCounts for counting the number of reads per transcript. Then they use DESeq2 for differential expression analysis.

My question is that since Stringtie itself also gives the count (in terms of coverage, FPKM and TPM) for each transcript that it constructs, then why should I use FeatureCounts for these constructed transcripts?

[GenoMax]

DESeq2 requires raw counts, which featureCounts provides.

Answer 1

[Kristoffer Vitting-Seerup]

There is absolutely no reason to do that. Actually you should rather use StringTie's quantification since that is more accurate than featureCounts - you can read more about such considerations here. Acutally DEXSeq directly supports analysis of StringTie data via tximport::tximport() as described here - although they actually don't mention StringTie in the DESeq2 vignette tximport does support StrigTie as describe here.

[c_u]

Hi Kristoffer, did you mean DESeq or DEXSeq?