Is it correct to use Tophat2 directly followed by Cuffquant to only align to the reference transcriptomes without wishing to assemble new transcripts?

Question

[zoukai3412085]

Hi, friends. I only want to perform differential expression analysis on the annotated transcripts of my existing reference genome. I use tophat2 for alignment with --no-novel-juncs --no-novel-indels --transcriptome-only parameters, and then I directly use Tophat2's accepted_hits.bam file as the input file of Cuffquant, and use the genome's reference annotation file genome.gtf as the reference annotation file, can this achieve my purpose? I do not wish to perform an analysis of any new transcripts. So I have not used Cufflinks and Cuffmerge to assemble and merge transcriptomes. Is this correct?

Answer 1

[Rob]

If you don’t want to assemble novel transcripts, there is absolutely no reason to use TopHat2 + Cuffquant. In fact, even if you do want to assemble new transcripts, there are much better options, like aligning with STAR or HISAT2 followed by assembly with StringTie2 or Scallop2. In fact TopHat and TopHat2 have entered a low-maintenance mode and the developers explicitly suggest the use of HISAT2 instead.

However, if you only want to quantify the annotated transcripts, you’re much better off using a tool like salmon for transcript quantification. It is easier to use, faster, and more accurate, and it’s output can easily be integrated into downstream differential expression analysis workflows.

[zoukai3412085]

Thank you so much for your advice. As you say, I have tried HISAT2, but it is killed due to some reasons like a large genome, and so on. So I use Tophat2 to solve my problem. Now, I want to try salmon as you recommended. Thanks.