Add design file for functionality for differential expression analysis, multiple species

[drpatelh]

@apeltzer

If we are able to provide some sort of experiment design file to the pipeline then it will be relatively straightforward to perform the differential analysis on the counts.

For example the atacseq pipeline uses the R script below with a matrix of counts as input:
https://github.com/nf-core/atacseq/blob/dev/bin/featurecounts_deseq2.r

Pipeline should be able to run without this feature for backward compatability? @ewels

[apeltzer]

I think making this available as an optional step would be really nice - especially if we can generate these anyways automatically in e.g. a tsv/csv table format 👍

[ewels]

Yes, this would be great! Needs some thought about how to refactor the input channels whilst retaining backwards compatibility, but hopefully shouldn't be too tricky.

[olgabot]

For reference, here's an implementation of input sequences that can take SRA, **{R1,R2}*.fastq.gz, a csv file, or fastas: https://github.com/czbiohub/nf-kmer-similarity/blob/master/main.nf#L80

[olgabot]

(updated title to include "design file" for easy searching and added species for my own suggestions :)

Is there any interest in multispecies support for rnaseq? e.g. for PRJNA143627 (https://ewels.github.io/sra-explorer/#) there’s 9 species and it would be really awesome to give like a tab-delimited (no csv since commas are in the R1,R2 definition) file that said reads,genome and it would align all samples. even globs would be good, e.g.

reads                                genome     singleEnd
s3://data/human/**{R1,R2}*fastq.gz    GRCh38    false
s3://data/mouse/**.fastq.gz           GRCm38    true

I'm about to run 18 separate nf-core/rnaseq runs for 9 species x both single end and paired end so this pain is quite real for me right now :) Plus, having all species in one multiqc report would be super awesome!

[ewels]

My suspicion is that support for multiple genomes per run will add quite a lot of complexity, and that your use case is relatively rare 😉 If your work folders are kicking around still then it's pretty easy to re-run MultiQC on the different MultiQC workdirs from each run. This is what I've done in the past when I've had similar situations.

[drpatelh]

Goodness this one 🙈

[drpatelh]

Initial functionality for samplesheet input added in #459. Input format will be:

group,replicate,fastq_1,fastq_2

which should be enough information to perform a basic differential analysis.