--quantMode GeneCounts options?

1,867 views
Skip to first unread message

Jenny Zadeh

unread,
Aug 20, 2015, 3:43:14 PM8/20/15
to rna-star
Hello,

We've switched over to using STAR for all our RNA-Seq alignments and I'm happy to see the addition of generating gene counts with --quantMode GeneCounts. However, currently there is no way to specify other options, like alternatives to feature type = "exon" and attribute = "gene_id" and therefore it's of limited use for us (lots of non-model organisms with non-standard gtf/gff files). Will these and other options available in htseq-count be included in a future release?

Thanks,
Jenny Drnevich Zadeh

Alexander Dobin

unread,
Aug 24, 2015, 5:03:46 PM8/24/15
to rna-star
Hi Jenny,

--sjdbGTFtagExonParentGene (=gene_id by default) should take care of the gene parent attribute, though I have not tested it thoroughly, please let me know whether it works.

The --sjdbGTFfeatureExon takes care of the feature type, however, it will affect both the read counting (which you want) and the annotated junctions. The latter may not be good if the feature types you want to count against are not true exons (e.g. CDS or UTR) - in this case you might be get some spurious junctions included in the database.

I will add the ability to count against the different feature types in the future releases.

At the moment, you could do the following:
1. Generate the genome with the normal parameters, i.e. with "exon" features.
2. In a separate directory, generate the genome with the different --sjdbGTFfeatureExon.
3. Copy the exonGeTrInfo.tab and geneInfo.tab from the 2nd directory into the first one.
4. Map using the 1st genome directory.

I believe this should work, but I have not tried it myself, if you try it, please let me know whether it works.

Cheers
Alex

Zhuofei Xu

unread,
Oct 20, 2015, 9:33:53 AM10/20/15
to Alexander Dobin, rna-star
Hi Alex,

Can I use --sjdbGTFtagExonParentGene transcript_id --quantMode GeneCounts to get read counts per transcript?

Many thanks,
Zhuofei

--
You received this message because you are subscribed to the Google Groups "rna-star" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rna-star+u...@googlegroups.com.
Visit this group at http://groups.google.com/group/rna-star.

Zhuofei Xu

unread,
Oct 20, 2015, 10:00:44 AM10/20/15
to Alexander Dobin, rna-star
Hi,
I have run STAR with --sjdbGTFtagExonParentGene transcript_id --quantMode GeneCounts. It can get read counts per gene rather than transcript.

Is there other options available for supporting read counting per transcript?

Thanks
Zhuofei

Alexander Dobin

unread,
Oct 23, 2015, 4:38:50 PM10/23/15
to rna-star, ado...@gmail.com
Hi Zhuofei,

you can do that in principle, however, only the reads that map to one transcripts will be counted, and the reads that map to multiple transcripts will be considered ambiguous.
So you will lose ~1/2 or more of your reads. The RSEM/Cufflinks/eXpress and other quantifiers use the MLEs to count these ambiguous reads, but STAR does not have this option yet.

Cheers
Alex
To unsubscribe from this group and stop receiving emails from it, send an email to rna-star+unsubscribe@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages