Counting intron reads in bulk RNA-seq

[Joe M]

We have an RNA-seq data set generated using illumina Truseq stranded total RNA depletion and not surprisingly, the data contains a large proportion of reads that map to introns (50-60%) using a recent standard gencode. Without getting into a debate of the merits of the assay, I would really like to use the intronic data somehow.

There are suggestions that counting intron reads might be useful although would be measuring a different facet of transcription:
https://academic.oup.com/nargab/article/2/3/lqaa073/5910008

I think a comparison of exon vs exon + intron could be useful and would not just throw reads out.

We could go to bwa, kallisto, or salmon, but since we used STAR for spliced bulk, I feel that a STAR vs STAR comparison might be better although maybe the mapping is so different that it doesn't matter.

Is there a principled way to count intron + exon reads to give total transcription gene counts ? We can create a simplified GTF with gene boundaries, but I am worried that STAR (bulk) algorithm fundamentally assumes a splicing model tuned to ligate out large chunks of sequence and negating that will be difficult.

Suggestions ? eg don't bother with STAR . Tuning tweaks ?

Read counting also possibly becomes an issue but haven't thought about that or RSEM yet.
Thanks, Joe.

[Alexander Dobin]

Hi Joe,

If you want to perform simple counting of all reads within gene boundaries, you can do it by supplying STAR with a modified GTF. You can add transcripts with gene start/end positions for each gene to the standard GTF. This will count intronic reads while also keeping the splice junctions that are used for mapping.

[Joe M]

Alex,
Thanks for confirming this.

Thanks also for continuing to support STAR and for your willingness to share knowledge and ideas to improve transcriptome practices generally.

closing this now.