--outFilterMultimapNmax option

[Jose Manuel Latorre Estivalis]

Hi Alex,

First of all, I would thank you for providing support to all STAR users.

I am interested on sensory receptor genes and I studied their expression by means of RNA-Seq in several insect vectors trying to identify if some of them are differentially expressed. Most of these genes have similar sequences due to their evolutive origin (mainly gene duplication events). For this reason,  and in order to identify accurately those sensory genes that are differentially expressed, I usually use --outFilterMultimapNmax 1 during the mapping process. Do you think that is it correct or do you recommend other value for this parameter?

Thanks for your time and help.

Best,

Jose

[Alexander Dobin]

Hi Jose,

--outFilterMultimapNmax 1 limits the output to just the uniquely-mapping reads (i.e. reads that confidently map to only one locus).

If the genes of interest have high sequence similarity, this option will probably eliminate a large number of reads that map to multiple loci.

Working with multi-mappers, on the other hand, requires a strategy to assign them to specific loci.

Cheers

Alex

[Jose Manuel Latorre Estivalis]

Hi Alex,

Thanks a lot your answer. I forgot to mention in my previous post that I use "Genecounts" options on the STAR script. Then, as far as I understand, "GeneCounts" will deal with the multi-mapped reads, not counting those reads that mapped in multiple locus. Is it correct?

Thanks again for your time and feedback

Jose

[Alexander Dobin]

Hi Jose,

yes, presently GeneCounts option counts only uniquely mapping reads, without regard to the --outFilterMultimapNmax value.

Cheers

Alex

[Matteo Di Bernardo]

Just following up on this, does this effectively mean that setting --outFilterMultimapNmax 1 will bar reads from even ending up mapped to one locus (so in your Aligned.bam file, you would see no multimapped reads), while the GeneCounts option will bar these from ending up presented as counts in the final UMI x gene output? Therefore, if you allow --outFilterMultimapNmax to be ~20 as default, yet provide the GeneCounts option, you'll see multimapped reads in your .bam file but these won't figure in the final output, correct?

Thanks so much,

Matteo

[Alexander Dobin]

Hi Matteo,

You are correct: GeneCounts output always counts only unique mappers - it does not depend on --outFilterMultimapNmax option.