Salmon or Kallisto without reference transcriptome - possible?

Question

[tanya_fiskur]

Hello everyone,

Do I understand it correctly that if I want to count reads on a transcript level, and have only the mapped Illumina reads and a .gtf annotation file, I cannot use the Salmon or Kallisto? What could you recommend to use for this purpose?

Thank you!

[antonioggsousa]

Hi,

I believe that you can use featureCounts: http://bioinf.wehi.edu.au/featureCounts/

I never use it, but from the description says that you need to provide a reference/genome GTF file and the mapped reads in sam/bam format.

I believe this answers your question,

António

[swbarnes2]

FeatureCounts will assign reads to genes or exons, but I don't think it will assign to transcripts.

[antonioggsousa]

Yes, you're right!

This link explains that is not possible to get transcript-level quantification with featureCounts: A: How to get read counts on transcript level using featurecounts?

Sorry, for the bad answer and thank you for correcting me.

António

Answer 1

[swbarnes2]

RSEM can assign reads by transcript.given a bam aligned to genome. You'll have to prepare the reference first, so you'll need the genome and gtf.

[Devon Ryan]

Are you sure RSEM can now handle BAM files with genome alignments? Last time I used it it needed transcriptome alignments.

[swbarnes2]

You might be right...I've always generated both, and run RSEM on that.

[Rob]

Right, you need the bam in transcriptomic coordinates. If you have that, you can already run salmon on it.