Repository navigation
P2: distinguish Gemmi superposition from experimental-frame placement validation #950
Description
Severity: P2. Reviewed PR #916 at e8769e6366736196a42909445f0beeffae6ef598, base 6bd461b5e9f005646431cd373e97ffad3aa5d80d.
ref/tools_and_measurements_guide.md:115,121,123 presents gemmi align as symmetry/origin-aware placement validation for T08/T09. The command instead performs sequence alignment and free structural superposition. Its fitted RMSD can erase the translation/rotation error being assessed.
The pinned Gemmi0.7.5 command source defines --rmsd as structural superposition. The fresh reviewer and PRIMARY separately ran the same controlled counterexample: copy retained archive member data/1sar_final.pdb, translate every ATOM/HETATM X coordinate by exactly20Å, then run gemmi align --query=A --target=A --rmsd original.pdb translated_plus20x.pdb. Installed Gemmi0.7.5 reports CαRMSD0 over96CA/P atoms. PRIMARY also checked the exact coordinate translation independently. Original archive/member bytes were preserved.
Suggested fix: distinguish free structural superposition from measurement in a common experimental frame. Placement validation must specify the frame and permitted crystallographic symmetry/origin operations, preserving the translation/rotation error being evaluated. Remove the claim that the existing CLI provides that integration; qualify catalog recommendations as recommendations rather than implemented frame-aware measurement. Check the T08/T09 matrix and prose siblings. No new placement implementation or licensed experiment is required.
Triage: in scope, because these unsupported runnable-capability claims are newly published by this PR. Complete38-open-issue queue, bodies and comments were checked: existing #937 concerns R-factor execution, #938 ligand/audit capabilities, and #927 thresholds; none covers this placement-frame defect. No scientific threshold or retained result changes are needed.
Evidence is retained in PRIMARY 012-primary-finding-verification.json, 012-primary-probes.log and the fresh-context review report. The reproduction is a local command-capability counterexample, not a new crystallographic benchmark.
Activity
realmarcin commentedon Nov 5, 2026
P3 residual at
2f52b15e60be0161811a1a07e71ee43ef0a9e972,ref/tools_and_measurements_guide.md:115: the new option list names--query-chainand--target-chain, which pinned Gemmi 0.7.5 rejects. PRIMARY ran the listed names against the retained original/translated 1SAR copies: exit 2,Invalid option.The identical inputs with--query=A --target=A --rmsdexit 0 and reproduce fitted Cα RMSD 0. Installedgemmi align --helpalso gives--queryand--target.Suggested fix: replace the two invalid names with
--queryand--target, or omit the unnecessary option inventory. Preserve the corrected distinction between free superposition and experimental-frame placement. The scientific remedy is otherwise supported by PRIMARY's source check and counterexample.Triage: in scope; this is a newly introduced runnable-capability detail in the #950 repair. Deduplicated under this existing open issue after refreshing the complete 40-issue queue. No new issue or scientific measurement is needed. Final acceptance remains pending this correction and fresh review/QC.
Address fresh review findings #950, #951, and #934 in PR #916
Address prose residuals in #950 and #926
realmarcin commentedon Nov 5, 2026
Fixed in PR #916 at
d039ebd8786e7b90d1e76197a418ce352fec6548. The guide distinguishes freely fitted Gemmi RMSD from experimental-frame placement validation, qualifies catalog recommendations and marks the missing frame-aware integration unavailable. Canonical IDs are preserved. A fresh review also caught invalid chain-selection option names in the first repair; the final guide now uses--queryand--target, independently checked against pinned Gemmi 0.7.5 help and a successful command. Final fresh read-only review found no remaining defect; locked final-head QC passed. CI and conditional merge gates remain pending.