search

nf-core / proteomicslfq

Proteomics label-free quantification (LFQ) analysis pipeline
https://nf-co.re/proteomicslfq
MIT License
33 stars 19 forks source link

proteomicslfq process fails when running with decoys #151

Closed veitveit closed 3 years ago

veitveit commented 3 years ago

I am running the UPS data set with the --add_decoy option:

nextflow run nf-core/proteomicslfq -r 1.0.0 -profile docker --input https://raw.githubusercontent.com/bigbio/proteomics-metadata-standard/master/annotated-projects/PXD001819/PXD001819.sdrf.tsv --database https://raw.githubusercontent.com/nf-core/test-datasets/proteomicslfq/testdata-aws/uniprot\_yeast\_reviewed\_isoforms\_ups1\_crap.fasta\_td.fasta -with-report -with-trace -with-timeline -resume --add_decoys true

The log file tells that there are "no proteins left after FDR filtering" and I get a lot of warnings of the kind

Am I missing any particular parameter setting?

Thanks!

jpfeuffer commented 3 years ago

I think the problem is, that the specific database in question already contains decoys. Instead, you would need to specify the decoy_affix. This messes up FDR calculation quite a lot.

If you want to keep+quantify+export decoys, this is not possible in 1.0.0 and was only added to dev to support Triqler output which expects decoys. This might need refinement if we make it the default for e.g. also pmultiqc. The param for this is quantify_decoys.

veitveit commented 3 years ago

OK, thanks! Alternatively, I could also use a database without decoys, right?

And I should update you on the ELIXIR project at some point. Did you get my email?

veitveit commented 3 years ago

Works now