featurecounts flags for lncRNA and mRNA co-expression

Question

[Sam]

Dear Biostars

I predicted some lncRNAs transcript according to RNA-seq data and then I used featurecounts to count the lncRNAs transcript (-t exon and -g transcript_id ) in the BAM files. for lncRNA/mRNA co-expression analysis, should I count mRNAs according to transcipt_id or gene_id (-t exon and -g gene_id)?

Thanks

Answer 1

[JC]

My suggestion is you need to test both, at the gene level, isoforms can be expressed differently between treatments, or not.

[Sam]

featurecounts with -g transcript_id give about 20% less DE mRNAs compare to grouping according gene_id. so is it expected to have less DE in transcript level? I think if we compare transcript vs. transcript( for mRNAs vs. lncRNAs transcript correlation), it could produce more robust data. is it true?

[JC]

Robust it not fair comparing genes (which are the sum of all isoforms) and just transcripts. Some genes can be no differential but some isoforms could be. That also doesn't give a good idea if the change is biologically important. I am more on the side to do both analyses and check if the results make sense biologically.