Mapping samples to multiple reference fasta

[rbpisupati]

Hi,

I was thinking of adapting this pipeline to take care of multiple reference genomes. Something like each sample would be aligned to different reference fasta file. Any votes for such a feature?

The input if given would be a csv file with sample_id, path_to_reference_fasta, path_to_sample_fastq

Cheers,
Rahul

[phue]

I think this would be a nice addition, particularly for population (epi)genetics.
Something similar to a design file as was done over at nf-core/chipseq by @drpatelh

We need to think about how this could work together with iGenomes though

[drpatelh]

I have done something similar in the nanoseq pipeline. Check out the format of the input samplesheet here. The important thing is to have a validation script for the samplesheet that checks the entries provided by users. I wrote this function to resolve the iGenomes per sample if provided. Note, you will also have to configure the channels to only build genome indices once if multiple samples share the same reference. It will be tricky but not impossible 🙂

[phue]

There is now a dsl2 branch with this functionality.

Please give it a try 👍

[ewels]

NB: This was removed from the dsl2 branch so that we could work with centralised modules in nf-core/modules.

However, I then wrote nf-core modules patch to allow local modifications of remote DSL2 modules for exactly this use case ☝🏻

Now it just needs implementing again..

[njspix]

Hey all, thanks so much for all your hard work on the DSL2 pipeline!
Just wanted to note that an alternative to patching the DSL2 module could be a pattern something like the following:

input = Channel.of(
    ["homo_sapiens", [fastq_1, fastq_2]],
    ["homo_sapiens", [fastq_1, fastq_2]],
    ["mus_musculus", [fastq_1, fastq_2]]
)
refs = Channel.of(["homo_sapiens", path/to/ref], ["mus_musculus", path/to/ref])

input
    .combine(refs, by: 0) // "homo_sapiens", [fastq_1, fastq_2], path/to/ref
    .multiMap{ it ->
        fastqs: it[0..1]
        refs: it[2]
    }.set{ for_align }

aligner(for_align.fastqs, for_align.refs)

I've had succes using this in pipelines that I've written for personal use. The combine operation syncs everything up, and the multiMap operator kind of 'locks' the two output channels together so that they're always in sync.