compomics/ms2rescore

Support different levels of confidence estimates when using mokapot

Opened this issue · 2 comments

jonasscheid commented

Currently PSMs are evaluated using psm-level FDR when using mokapot see here
Peptide-level confidence estimates are also present in the object mokapot returns. Would be of interest to give the user the freedom to choose the FDR-level

RalfG commented

Agreed that this could be reported somehow. The main issue blocking this is that the (output) object in MS²Rescore is the PSM list, which has only PSM-level entries. Would you repeat each peptide/protein-level q-value and PEP for each PSM linked to these? Remaining issue then is how to deal with PSMs linked to multiple protein groups. Or am I missing something? Ideas are very much welcome!

jonasscheid commented

Would you repeat each peptide/protein-level q-value and PEP for each PSM linked to these?

Yes I had that also in mind! But for peptides mapping to multiple proteins that is indeed a tricky one..
One ad-hoc idea I have for protein-level fdr is to store a dict of proteins with the confidence estimates in provenance_data, but then again provenance_data : dict[str, str] so that is currently not possible. Maybe only support peptide-level-fdr for now? 🤔