Skip to content

TBD - #8

Draft
iskoldt-X wants to merge 8 commits into
protwis:mainfrom
iskoldt-X:prompt-routing
Draft

TBD#8
iskoldt-X wants to merge 8 commits into
protwis:mainfrom
iskoldt-X:prompt-routing

Conversation

@iskoldt-X

Copy link
Copy Markdown
Member

No description provided.

iskoldt-X and others added 8 commits July 22, 2026 22:47
- Multi-copy ligand alerts now gate on genuine per-copy binding-site
  divergence (>1 distinct site_ref, fail-closed on missing sites) rather
  than crystallographic copy count; per-copy role divergence is advisory.
- Separate native G-protein alpha subtype ambiguity from genuine chimeras:
  when the alpha5 helix confirms the family but cannot resolve the member
  (e.g. Gi1 vs Gi2), emit an advisory "[G-ALPHA SUBTYPE UNRESOLVED]" note
  instead of a critical "[CHIMERIC G PROTEIN]" warning; non-human orthologs
  ("[G-ALPHA SPECIES UNVERIFIED]"), model-declared chimeras, and unverified
  constructs remain gated. g_proteins.csv columns are unchanged.
- Stop mistaking peptide agonists and fusion/tag chains for receptors:
  extend the non-receptor slug-prefix list (collision-checked against the
  receptor set) so they no longer trigger the suspicious-7TM alert; the
  alert and its guidance are unchanged.
- De-noise validation gating: route already-corrected PubChem-mismatch
  notes and apo-placeholder site disagreements to advisory; make the
  apo-with-molecules check role-aware (advisory only when coexisting
  molecules are all structural cofactors/ions/lipids); add a source-side
  reconcile that downgrades gating records pointing at entities absent from
  the shipped output, behind a bracket-safe, fail-closed guard.
- Add unit tests for the new gating branches; full suite green.

Co-Authored-By: Claude Opus 4.8 <[email protected]>
… single receptor

Adds a chain-total check on the all-single stoichiometry proxy, scans every candidate assembly for a homo-oligomer record, and skips entries whose cited evidence only recites that assembly.
A fusion or splice construct can fragment the UniProt alignment so only five TM helices map, all of them resolved. The absolute six-helix threshold read that as missing density; requiring resolved == total with at least five annotated separates a mapping artefact from a genuinely unbuilt helix.
…etes a heterotrimer

Also stops reporting a relocation when the subunit is already recorded on that chain: nothing is recovered there, so the bucket entry is left in place rather than deleted and neither channel describes a move that did not happen.
The silent no-op path is for beta and gamma only. An auxiliary entry sharing the G-alpha chain is what a receptor-Galpha fusion or a dominant-negative chimera looks like, so alpha keeps its review, matching the split this function applies elsewhere.
The weak-match branch is the one place the sequence check returns nothing about G-alpha identity, so a slug asserted there ships unverified. An asserted slug now raises a conflict; an honest abstention stays a note, since there is no claim to check.
The compound row cannot record which copy sits at which site; mirror the gating alert so that table opens for review.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant