Using Salmon for transcripts quantification in meta-transcriptome

[NathanVannier]

trafficstars

Dear Salmon Team,

Not sure if this message should go here or on the google group, sorry if it shouldn't go here.

I am trying to analyse meta-transcriptome RNAseq data to obtain genes expression from "complex" microbial samples (over a 100 strains of bacteria and fungi). I have over 500 millions reads so aligning these reads on references genomes would take forever... Salmon thus looks like a promising solution.

However, i am not sure Salmon is appropriate for meta-transcriptome data. Is there prerequisites if i want to use Salmon for that purpose ? What would be your recommended parameters ?

So far, since i have a reference genome (and thus the associated coding sequences) for all the strains i created a single Salmon Index (comprising all the strains cds) and ran the quantification tool on this index using these parameters: salmon quant -i index --libType A --validateMappings --seqBias --incompatPrior 0.0 -r input.fa -o output

I did not include gcbias correction so far because i am not sure it is appropriate in this case.

Thanks for you help,

Nathan

[AstrobioMike]

I'm curious too if there is any specific feedback from the devs on whether using salmon on bacterial coding sequences is generally seen as appropriate or not? (sorry if this information is already available elsewhere, I've looked a bit and not seen it yet)

[lmuenter]

I'm wondering the same thing! Can Salmon be used for quantification with a reference set of predicted CDS? IMO this would also have advantages for quantification with de novo assembled transcriptomes, as this would alleviate problems with chimeric contigs...

[akrinos]

Also would like to ask for advice on this question from the developers! TIA!