Read counts at gene and transcript level - hisat2+Featurecounts or Stringtie?+deseq2

Question

[tianshenbio]

I hope to do DE analysis at both gene and transcript level.

First I mapped the reads to the genome using hisat2. Now I need to generate a raw count matrix for deseq2.

Since I have a well-annotated gff file, I am not interested in finding new isoforms. If I chose Stringtie, it would only be used to generate raw count matrix. And it can calculate both gene and transcript in one run (right?). If I chose Featurecounts, I have to run it twice, for gene and transcripts separately. But someone told me Featurecounts is not suitable to quantify isoforms...

What would be a better choice?

[lieven.sterck]

for instance salmon is well suited for isoform quantification (better than FeatureCounts of HTseq-count indeed)

with 'gene' you mean gene-locus and "transcript' = isoform(s), correct?

[tianshenbio]

Yes, with 'gene' I will count reads mapped to all exons of that gene locus, with 'transcripts' I hope to quantify isoforms. Is Stringtie suitable for isoform quantification as well?

[lieven.sterck]

not sure if I (can) agree with your approach here :/

sorry, don't have much hands-on experience with StringTie, but I know it can (and is used) for creating de-novo transcripts (and isoforms) from mapped reads. Others will chip in here I assume.