Select molecular states and record preparation decisions for docking

Start with devguide/pending_proposals/select_molecular_states_and_record_preparation_decisions_for_docking.md, then review the related MolSysMT issues #217, #218, #220, #221, and #222. Determine how docking results can be associated with ligand and receptor states, pH, variants, retained waters or cofactors, and transformation parameters. Done means these preparation decisions and source inputs are explicit and traceable in a docking run.

Saved human-case recovery — 2026-10-29

#58 is completed at 710567d9350873cee9cf9bb2aa09e1bd5f9ff249. The documented developer operation verifies the externally expected qualification, all 16 retained members, supplied compressed-source bytes and admission/prepared/protocol/domain/captured-byte/result links. It recovers the original five-pose record after relocation, preserving original producer attribution, absent source fingerprints and problem is None.

Source and ordinary-installed selections each pass 137 tests, including 26 recovery controls; all 389 historical evidence files remain unchanged. The qualified c384ed2 runtime wheel is reused without rebuilding or relabelling, with all 58 runtime files identical. Exact-head CI passes all seven jobs, 1,367 tests per Python 3.11–3.14, zero test skips and actual coverage upload; policy and publication governance also pass.

This broader issue remains partial for other source families and general molecular persistence/replay. No native system, chemical state or mechanical assignment is reconstructed by record recovery. General molecular persistence remains uibcdf/molsysmt#256; partial/reference reading stays #49/#33 with uibcdf/molsysmt#521. Incoming #56 does not change current pins. No new preparation, energy calculation, search, viewer acceptance or public release is claimed.

Previous checkpoints retained

Prepared-input recovery guard — 2026-10-29

#57 addresses a concrete recovery gap found in this continuation: a supplied prepared object could change its state/charge declaration while the old result recovered successfully. Final source c384ed2be10e8b24d130658ba74a31fd44810ddc compares recorded object declarations, rejects backend-text substitution, checks retained Vina evidence for both mixed roles before export, and verifies actual prepared bytes through the existing public MolSysMT-backed writer. Contract and precise limits.

The final source and ordinary-installed selections each pass 273 tests, zero skips, including 40 recovery guards. Initial source/wheel 37d83d0 and its 271-case receipts remain distinct; two later mixed-role controls expose its uncovered preflight boundary. The corrected wheel 0.0.0+210.gc384ed2 retains SHA-256 d86bd904e4e25da75acc0fb5254e270ceec7fe2641f68821a7208641c26ed8e0; all 58 runtime files match Git/source/wheel/installed prefix. All 64 frozen historical/authority/initial-evidence files remain unchanged.

This is available-evidence consistency, not a source bundle or molecular snapshot. Missing old declarations/hashes establish no extra identity; matching writer bytes do not establish subprecision coordinate or original chemical identity. No chemical preparation, energy, search or molecular parser is added. Original source fingerprints can remain absent. General native mechanics persistence stays uibcdf/molsysmt#256; current provider pins and incoming #56 adoption are unchanged. This broader issue remains partial.

#57 is completed. Exact final-head CI 40916957436, attempt 1, passes all seven jobs and 1,341 tests per Python 3.11–3.14, zero test skips, ordinary installed/context/closure gates and actual coverage upload. Policy 40916958461 and publication governance 40916958776 pass. Full terminal evidence is verified; no publisher remains pending. No public package/release, original-population rerun or visible C4/C6 acceptance is claimed.

Previous profile checkpoints retained

Declared profile and current-composition admission — 2026-10-29

Published profile, DMT-043 and indexed evidence. Documentation/evidence head 12b536046f6dc5fa9bd823d1f974be540037d225; runtime and ordinary wheel remain the original 8d40870724e21cf579ce4d5289d46b5d8a5aaa1c, with all 58 runtime file hashes unchanged. The wheel was not rebuilt or relabelled.

The separately registered control passes on the unchanged 3PTB hypothesis: receptor 2,011 atoms, rigid/flexible ligand 13 atoms each, named charges/types, conserved +6/+1 e projection, native input parsing and actual native export/map/readback for both ligands. A pm/fs repeat preserves exact payloads/backend boxes and caller policy. All 44 frozen historical/source/authority files remain unchanged. Zero native searches, energy evaluations or generated poses; reader input is the actual export, not new Vina output. Four expected warnings retain the explicitly allowed occupancy/B-factor loss. Receipt SHA-256 d1b8c434a68dbb5eb869511a09bea5360982af01234eb62e767c4afe87efbaf3.

A separate ordinary-installed selection passes 108 tests, zero skips, 63.79 s, 35 warnings; 24 administrative and five governance cases pass separately. Seven synchronized guides and 214 relative links verify. An initial guard selection named two nonexistent modules and was refused before pytest; only the corrected selection qualifies. The scientific producer/registration did not change.

Source/replay boundary: the case bundle links original bytes/axes, declared decisions, provider transformations and actual execution records. Prepared problem inputs honestly retain source_fingerprint=None; no native origin or experimental state is invented. Standalone object reconstruction still requires explicit receptor/partner inputs. Captured PDBQT is backend-byte evidence, not a complete molecular/assignment snapshot. This issue stays partial for an actual caller's reusable bundle/recovery workflow. Inspect provider persistence/report contracts first; general molecular storage stays MolSysMT-owned (uibcdf/molsysmt#256), with no consumer snapshot implementation.

Exact documentation-head CI 40907435829, attempt 1, is completed/success: all seven jobs, 1,301 tests per Python 3.11–3.14, zero test skips, 318 warnings per lane, ordinary installed/context/dependency-closure checks and actual coverage upload. Full terminal captures verify all steps and expected conditional skips. Suite seconds: 3.11 640.94; 3.12 623.56; 3.13 590.74; 3.14 566.30. Policy 40907436525 and publication governance 40907436151 pass; no publisher remains pending. The previous 1,301-case matrix remains the separate runtime qualification, not a relabelled new run. Visible C4/C6 remains #52; public delivery stays #30/#47. This broader issue remains partial.

Earlier checkpoints retained

Domain-provenance follow-up — 2026-10-28

#55 is completed at 8d40870724e21cf579ce4d5289d46b5d8a5aaa1c: optional selection/redocking derivation is preserved through serialization and linked to the backend projection, with public MolSysMT resolving molecular data. Maintained contract.

Source and ordinary-installed selections each pass 188 tests. Exact-head CI passes all seven jobs and 1,301 tests per Python 3.11–3.14, zero test skips, installed/context/closure gates and actual coverage upload. Policy and publication governance also pass. Original human/scientific producers and provider pins remain unchanged; no public package is published.

This broader issue remains partial. Next work concerns the bounded preparation/scoring profile (#5/#33) and general prepared-source/decision provenance (#4). Visible C4/C6 remains deferred under #52. The earlier audit below retains its original runtime/evidence scope; its missing-domain-record observation has now been resolved by #55.

Current provenance acceptance — 2026-10-28

Nonvisual Core MVP audit at documentation/evidence head aa2f317c132e255470b0512649c3691c56364ed9; inspected runtime remains c9c8196. All 32 C0–C5 criteria have evidence, limits and owning follow-up. Original 24-search/167-pose and five-pose human records retain their bytes, producers and different provider composition. Six existing ordinary-installed controls pass (36 explicitly deselected, zero skips); the frozen receiving guard forbids new preparation, maps, scoring and searches. All 16 human manifest members, submitted-byte hashes and score/state bindings verify. No new scientific experiment, runtime/dependency change, visual acceptance or public publication is made.

Declared prepared state IDs, selected frames/axes, named methods, captured backend bytes, actual parameters and per-score descriptors are present for the preserved case. Source-file identity is established by the admission/qualification/result bundle: the standalone prepared-result JSON has source_fingerprint=None and absent native state labels; do not invent missing provenance or experimental pH from its descriptive state IDs.

Keep this issue partial for supported source/decision provenance across arbitrary prepared inputs and replay contexts. The concrete selection-domain recipe gap is resolved by #55, as recorded in the follow-up above. General molecular storage/manipulation remains MolSysMT-owned. C4 visibility/C6 stay deferred under #52.

Exact documentation-head CI 40771841699, policy 40771842081 and publication governance 40771842548 are now completed/success: 1,279 tests per Python 3.11–3.14, zero test skips, all seven CI jobs, executed installed/context/closure gates and actual coverage upload. Verified terminal recovery. The original c9c8196 ordinary-file/scientific producers remain distinct.

Completed audit/local checks and explicit hosted recovery. The audit documentation and its terminal hosted evidence are complete; the earlier pending closeout is preserved as historical.

Original scope retained

What — Make the chosen ligand and receptor states and preparation decisions explicit in a docking run.
How — Record ligand chemical state and conformer, receptor state, pH and residue variants, retained waters/cofactors, and any transformation parameters; associate each result with those inputs.
Why — The Core MVP requires prepared states and backend representations to be traceable to source systems.
Record — devguide/pending_proposals/select_molecular_states_and_record_preparation_decisions_for_docking.md
Related — uibcdf/dockingmt#3. DockingMT consumes MolSysMT state selection and preparation capabilities tracked in uibcdf/molsysmt#217, uibcdf/molsysmt#218, uibcdf/molsysmt#220, uibcdf/molsysmt#221, and uibcdf/molsysmt#222.