Filter mouse reads from PDX samples

[ssamberkar]

Hi,

I have some mouse PDX samples for which I need to run a standard DGE workflow later. While the standard nf-core rnaseq is more than enough for the purpose, for my use-case it is missing a mouse reads filtering step.
@drpatelh - suggested a Kraken based solution.

I was wondering if someone can suggest a process with BBmap (https://bioinformaticsonline.com/pages/view/35033/bbsplit-read-binning-tool-for-metagenomes-and-contaminated-libraries)

Thanks.

[apeltzer]

Out of personal experience, I'd also rather jump on the kraken2 train.

Both because it's more widely supported, used by more people and therefore subsequently also more tested. I also doubt that there will be a problem getting always updated kraken2 DBs / bioconda scripts, whereas all of the bb* tools are from a single developer who might just stop development at some point.

[drpatelh]

FIxed in #700

[drpatelh]

Docs will be here after v3.4 of the pipeline has been released.

So to summarise you provide the pipeline with --bbsplit_fasta_list of the format:

mm10,/path/to/mm10.fa
ecoli,/path/to/ecoli.fa
sarscov2,/path/to/sarscov2.fa

This means you can use custom names to name your contaminant genomes and the reads relative to the main reference will always be called *primary*fastq.gz.

The BBSplit index will have to be built at least once with this pipeline and then can be re-used for future runs so it doesn't have to be re-built over and over again.