Repository navigation
Description
Goal: given any protein sequence — including a hypothetical protein BV-BRC has never seen — produce the same evidence report the genome route produces, with every source that could not answer shown as such.
Today the pipeline starts from a genome and depends on BV-BRC for the protein list, product names, specialty genes (virulence, AMR, drug targets, essential genes, human homologs), essentiality and taxonomy. About 2,500 lines (PDB/AFDB/Foldseek evidence, M3, the report) already have no BV-BRC coupling, so broadening the input is mostly a new entry point plus replacements for what BV-BRC supplied — not a rewrite. This is the first concrete step of Direction C in the post-workshop plan (#84).
python -m s2f.run --run runs/q1 --from-fasta query.faa [--organism "..."]
What BV-BRC provides, and what replaces it
| BV-BRC normally gives | Without BV-BRC | Status |
|---|---|---|
| Product name / function | UniProt (exact sequence, then search), eggNOG, InterProScan | exists (--map-ids, --annotate) |
| Organism / taxonomy | --organism, or inferred from the closest UniProt match and labelled as inferred |
small |
| Virulence, AMR | VFDB / CARD / AMRFinderPlus by DIAMOND | new (see #30) |
| Essentiality | DIAMOND against essential-gene reference sets | #29 exists for genomes only |
| Drug targets | ChEMBL via UniProt (--kg) |
exists |
| Human homolog | DIAMOND against the human proteome (--human-homology) |
exists |
| Structure | PDB, AFDB by accession, prediction | #54 for prediction |
| Function for true unknowns | Foldseek, structure-based GO, embeddings | #65, #86, #43 |
Sub-issues
Do #89 and #90 first. Together they show what already works without BV-BRC.
- M1: --from-fasta route — start a run from protein sequences, no genome #89 M1:
--from-fastaroute - Report: record BV-BRC-only evidence as not available, not zero, on non-genome runs #90 Report: BV-BRC-only evidence recorded as not available, not zero
- M2: virulence and AMR evidence without BV-BRC — VFDB and CARD by DIAMOND #91 M2: virulence and AMR without BV-BRC (VFDB, CARD)
- M2: essentiality for a sequence with no genome #92 M2: essentiality for a sequence with no genome
- M2: scoring profiles — the triage policy becomes a choice #93 M2: scoring profiles
- M2: what breaks for sequence-only and viral input #94 M2: what breaks for sequence-only and viral input
- M3: structure collection for sequence-only and viral input #95 M3: structure collection for sequence-only and viral input
- Decision: viral proteins — scope, and what changes per module #88 Decision: viral proteins
Open questions
- Viruses — Decision: viral proteins — scope, and what changes per module #88; it changes M1 (M1: --from-fasta route — start a run from protein sequences, no genome #89), M2 (M2: what breaks for sequence-only and viral input #94) and M3 (M3: structure collection for sequence-only and viral input #95).
- Short peptides (5–20 aa) defeat PDB search and structure prediction (see docs: post-workshop plan and the three directions for the restart #84); out of scope unless decided otherwise.
Unassigned on purpose: comment to claim this task, then assign yourself.
Activity