[OpenSwathWorkflow] Add direct in-silico library searching

[singjc]

It would be useful to support FASTA-driven in-silico assay generation directly inside OpenSwathWorkflow, similar in spirit to DIA-NN’s “library-free” mode and Spectronaut’s directDIA, where the user can provide a FASTA and run a single command end-to-end.

Currently, OpenSwathWorkflow requires a pre-built predicted assay library (e.g., generated with Prosit/PeptDeep/MS2PIP/DeepLC/DIA-NN) which is then passed into OpenSwathWorkflow.

We could integrate easypqp-rs (Rust in-silico library generation) in two ways:

• Add a C ABI (FFI) wrapper for easypqp-rs and call it from OpenMS.

• Call the easypqp-rs executable from OpenSwathWorkflow as an internal thirdparty preprocessing step (subprocess, probably the easiest approach?)

Considerations:

• Need to ensure some sort of reproducible library ID mapping, so that different MS run analyses have the same library IDs (precursor, peptide, protein, gene and transition IDs)
• We would need to perform direct corresponding decoy generation within OpenSwathWorkflow (currently done outside via OpenSwathDecoyGenerator)

Pros:

• Depending on how it's implemented, we could aid in reducing memory usage (from the large input transition list), by chunking how many in-silico proteins are digested at a time.
• Reduces the amount of extra work the user would need to do on their end
• It would be possible to fine-tune the prediction models based on the iRT calibration step, which would calibrate the predictions for the specific data file being analyzed.

@timosachsenberg, @jcharkow what are your thoughts?

[timosachsenberg]

The easiest option for now is to call an executable. I would probably go with that now and reconsider if we need a tighter integration.

[jcharkow]

As a first step I would probably just integrate it with nextflow and produce a .tsv or .parquet library file to load into OpenSWATHWorkflow afterwards rather than direct integration.