Deprecate openms tsv as an option for SDRF input and allow csv AND tsv

[jpfeuffer]

A lot of plugins and other systems interfacing with nextflow require or default to creating samplesheet inputs as CSV for nextflow/nf-core pipelines. And cloud storage interface like Seqera Datastudios and Explorer have special rendering cases for CSV and TSV only.
In general I think the nf-core way is to have only ONE schema for the input parameter.

Therefore I think we should:

• deprecate allowing two formats (i.e. deprecate the openms schema)
• revert or only additionally allow .sdrf
• definitely allow csv and tsv as alternative suffixes

@ypriverol

[ypriverol]

@jpfeuffer We have multiple users in the past talking about the tsv as input. The main use case is the identification workflow. While SDRF is amazing 😉 for quant analysis, for some users looks like an overkill for quant.

Im in favor of removing the openms tsv as input but we can also consider standardizing and put more examples for a tsv for id workflow rather than quant. What do you think @jpfeuffer @timosachsenberg @daichengxin @Copilot

[jpfeuffer]

The problem is that having two samplesheet schemata has basically no support in any upstream interface with nextflow.
No Seqera Platform, no other services that interoperate with nextflow like Lamin

In this case we should move the ID workflow out as a subworkflow together with our modules and make a new workflow that just does ID. This would also significantly decrease the number of if cases and complexity in quantms.
And you could exactly specify which columns are needed for each type of analysis.

[jpfeuffer]

Another solution would be to make SDRF easier to use 😄
I.e. by splitting it into two samplesheets. One for files and one for samples. To reduce duplication significantly.
But it might be harder to statically validate because you need to make sure that all the samples that are referenced in the file samplesheet are also in the sample samplesheet (but I guess this is not so much of a problem because there are other things that need runtime validation).

[jpfeuffer]

Mayyybe we could have two input parameters. One for openms tsv and one for sdrf. Then we could specify a fixed schema for each.
But then you will have a hard time specifying what is required and what not because it depends on if you are using ID mode or not.

[jpfeuffer]

Also, I just gave the SDRF spec to AI and said "convert to this" and it worked. So, I don't think it is very hard to create an SDRF these days. You can host an LLM based converter for this easily.

[jpfeuffer]

Not being able to create an SDRF is no excuse.

[ypriverol]

Lets deprecate it. The main problem is the TMT channels issues but to be honest this is a problem of reprocessing not of quantms.