| { |
| "schema_version": "1.0.0", |
| "audit_date": "2026-07-27", |
| "task": "BioCandidateRanker governed absolute activity training from repository-local sources", |
| "target_model_field": "log10_activity", |
| "status": "blocked", |
| "decision": { |
| "training_performed": false, |
| "adapter_written": false, |
| "canonical_unit_defined": false, |
| "normalized_dataset_written": false, |
| "family_split_written": false, |
| "cpu_smoke_performed": false, |
| "checkpoint_written": false, |
| "predictions_used_as_labels": false, |
| "fitness_used_as_labels": false, |
| "reason_code": "INSUFFICIENT_COMPATIBLE_DIRECT_ABSOLUTE_ACTIVITY_ROWS_AND_FAMILIES", |
| "summary": "No repository-local source provides enough compatible, governed, direct absolute activity measurements for family-aware log10_activity training. Turnover constants remain kcat labels; EnzEngDB fitness remains campaign-relative; Lunzer values have an unresolved absolute unit; RFU/min is instrument-relative; and the only shared physical activity dimension, mass-specific rate, has at most 23 rows in three families before scientific gates and at most 10 coherent rows in one family after the locally documented assay and consistency gates.", |
| "claim_boundary": "This is a source and compatibility audit only. No labels were created, no endpoint was relabeled, no split or model was produced, and no model scoring was run." |
| }, |
| "search_scope": { |
| "root": "<redacted-local-path>", |
| "repository_is_git_worktree": false, |
| "included": [ |
| "source adapters and curation scripts under src/ and scripts/", |
| "repository-local tabular, JSON, archive, workbook, sequence, audit, metric, checkpoint, and prediction artifacts", |
| "all source directories under artifacts/external/absolute-kinetics-screen", |
| "all source directories under artifacts/external/temporal-absolute-kinetics", |
| "all EnzEngDB v1 campaign files and audits under artifacts/external/enzengdb-v1", |
| "existing UniKP manifests and audits under artifacts/ and configs/" |
| ], |
| "endpoint_terms": [ |
| "activity", |
| "specific activity", |
| "Vmax", |
| "U mg^-1", |
| "umol min^-1 mg^-1", |
| "RFU/min", |
| "fitness_value", |
| "kcat", |
| "rate" |
| ], |
| "exclusion_policy": [ |
| "Do not rename kcat, Km, kcat/Km, fitness, relative activity, inhibition, simulated flux, or predictions as activity.", |
| "Require a finite direct experimental value, an explicit physical unit and denominator, exact sequence and defined substrate inputs, row-level provenance, and permission-compatible local evidence.", |
| "Require assay semantics to be compatible before pooling values that happen to share dimensions.", |
| "Require complete protein families to stay within one train, validation, or test partition." |
| ] |
| }, |
| "candidate_canonical_dimension": { |
| "dimension": "amount of product per time per mass of protein", |
| "interconvertible_source_units": [ |
| "U mg^-1", |
| "umol min^-1 mg^-1" |
| ], |
| "possible_canonical_unit_if_unblocked": "log10(umol min^-1 mg^-1)", |
| "defined_for_training": false, |
| "reason_not_defined": "Dimensional convertibility is not sufficient. After directness, assay-validity, conversion-consistency, provenance, and family gates, the local pool collapses to one family with at most 10 rows. Declaring a model-wide canonical activity unit would falsely imply an adequate governed training endpoint." |
| }, |
| "source_buckets": [ |
| { |
| "source_id": "temporal-global-family-audit", |
| "path": "artifacts/external/temporal-global-family-audit/readiness-audit.json", |
| "endpoint": "direct steady-state kcat", |
| "unit": "s^-1 after source-specific normalization", |
| "records": 177, |
| "families": 21, |
| "substrates": 50, |
| "activity_eligible_records": 0, |
| "reason": "A molar turnover constant is a governed log10_kcat endpoint, not a mass-specific or condition-specific activity endpoint. Relabeling is prohibited." |
| }, |
| { |
| "source_id": "UniKP-kcat-development", |
| "path": "artifacts/unikp_audit.json", |
| "endpoint": "kcat", |
| "total_rows": 17010, |
| "adapter_accepted_rows": 16838, |
| "activity_eligible_records": 0, |
| "reason": "The source and implemented task are log10_kcat. It also lacks row-level accessions and a bundled independent split; no kcat value was relabeled as activity." |
| }, |
| { |
| "source_id": "EnzEngDB-v1", |
| "path": "artifacts/external/enzengdb-v1/audit.json", |
| "endpoint": "fitness_value", |
| "total_rows": 462092, |
| "adapter_accepted_rows": 245945, |
| "campaigns": 160, |
| "activity_eligible_records": 0, |
| "reason": "Fitness has campaign-specific meaning and units. The existing adapter intentionally leaves log10_activity unset and permits only within-campaign ranking. EnzEngDB was excluded from the activity candidate count." |
| }, |
| { |
| "source_id": "Lunzer-Dryad-7nd70", |
| "path": "artifacts/external/absolute-kinetics-screen/dryad-4964723/Lunzer Curated Biochemistry.xls", |
| "source_identity": { |
| "sha256": "7bfd71d78235ade06a9e117b8229ad99b1fe7adceb0a5ed72d708c3e925c2a69", |
| "size_bytes": 38761 |
| }, |
| "endpoint": "derived ln(Km) + ln(kcat/Km)", |
| "records": 1024, |
| "families": 1, |
| "activity_eligible_records": 0, |
| "reason": "The absolute source-unit constant is unresolved and all 512 sequences have development-corpus homology hits. Existing use is ranking within cofactor only." |
| }, |
| { |
| "source_id": "mLDHD-S-EPMC10589216", |
| "path": "artifacts/external/temporal-absolute-kinetics/biostudies-S-EPMC10589216/exclusions.csv", |
| "endpoint": "Michaelis-Menten Vmax reported as mass-specific activity", |
| "unit": "umol min^-1 mg^-1", |
| "candidate_rows": 9, |
| "families": 1, |
| "direct_dimensionally_compatible_rows": 9, |
| "scientifically_eligible_rows": 0, |
| "reason": "Every row uses a PMS-DCIP coupled assay without demonstrated saturation of either fixed electron acceptor. The local provenance audit excludes all nine apparent rates; none may become an activity label merely because Vmax has a compatible unit." |
| }, |
| { |
| "source_id": "Pantoea-GDH-S-EPMC12364357", |
| "path": "artifacts/external/temporal-absolute-kinetics/biostudies-S-EPMC12364357", |
| "endpoint": "Michaelis-Menten Vmax reported as mass-specific activity", |
| "unit": "U mg^-1", |
| "candidate_rows": 10, |
| "families": 1, |
| "direct_dimensionally_compatible_rows": 10, |
| "rows_passing_local_cosubstrate_saturation_rule": 1, |
| "rows_with_development_homology": 10, |
| "training_eligible_rows": 0, |
| "reason": "Only one row passes the local >=5x fixed-cosubstrate Km rule, and the complete family is excluded by development homology. The ten values are useful provenance evidence but cannot support a family-aware activity split." |
| }, |
| { |
| "source_id": "SlDefH-PMC13230308", |
| "path": "artifacts/external/temporal-absolute-kinetics/europepmc-PMC13230308/blocker-evidence.json", |
| "endpoint": "specific activity or Vmax", |
| "unit": "U mg^-1", |
| "candidate_rows": 4, |
| "families": 1, |
| "direct_dimensionally_compatible_rows": 4, |
| "training_eligible_rows": 0, |
| "reason": "The completed source audit found zero independently reconcilable rows. Each reported specific-activity/kcat pair implies approximately 599 kDa versus the stated 29.5 kDa sequence mass, while the article and complete supplement provide no kinetic enzyme molarity, conversion equation, oligomer measurement, active-site stoichiometry, or active-enzyme fraction. Reported values remain provenance evidence only; the local registry finalizes this source at zero rows." |
| }, |
| { |
| "source_id": "Zenodo-4448337", |
| "path": "artifacts/external/absolute-kinetics-screen/zenodo-4448337/source.zip", |
| "source_identity": { |
| "sha256": "8ef6d1a23ba3d5a614c5670fabc28daf8644a4d1d1320f0a0850a54f1088090a", |
| "size_bytes": 4637 |
| }, |
| "endpoint": "kcat, Km, kcat/Km, and cofactor preference", |
| "activity_eligible_records": 0, |
| "reason": "The archive contains turnover kinetics and a dimensionless preference series, not a governed absolute activity endpoint." |
| }, |
| { |
| "source_id": "Zenodo-20626217-Dataset-S1", |
| "path": "artifacts/external/absolute-kinetics-screen/zenodo-20626217/Dataset_S1.xlsx", |
| "source_identity": { |
| "sha256": "83f559a75277c87c307fadb93393b6090a628055fc4c4e8a96d8e0e6e48d5338", |
| "size_bytes": 84504 |
| }, |
| "endpoint": "BACE1 inhibition and instrument response rates", |
| "units": [ |
| "percent inhibition", |
| "IC50 in uM", |
| "Ki in uM", |
| "RFU/min" |
| ], |
| "activity_eligible_records": 0, |
| "reason": "RFU/min lacks a physical amount calibration and protein-mass denominator; inhibition endpoints describe peptides acting on BACE1 rather than enzyme-candidate catalytic activity." |
| }, |
| { |
| "source_id": "model-and-simulation-artifacts", |
| "paths": [ |
| "artifacts/candidate_predictions.json", |
| "artifacts/external/absolute-kinetics-screen/dryad-4964723/dlkcat-output.tsv", |
| "artifacts/external/absolute-kinetics-screen/dryad-4964723/biocandidate-full-evaluation.json", |
| "artifacts/external/enzengdb-v1/zero-shot-kcat-proxy.json", |
| "artifacts/simulated-flux-smoke" |
| ], |
| "activity_eligible_records": 0, |
| "reason": "Predictions, model metrics, checkpoints, and simulated flux are not direct experimental activity labels." |
| } |
| ], |
| "mass_specific_activity_funnel": { |
| "dimensionally_compatible_candidate_rows": 23, |
| "candidate_sources": 3, |
| "candidate_families": 3, |
| "excluded_coupled_assay_rows": 9, |
| "excluded_internal_conversion_inconsistency_rows": 4, |
| "coherent_direct_rows_before_homology_and_saturation_gates": 10, |
| "coherent_direct_families_before_homology_and_saturation_gates": 1, |
| "rows_passing_existing_saturation_and_homology_gates": 0, |
| "families_passing_existing_saturation_and_homology_gates": 0, |
| "maximum_rows_if_homology_exclusion_were_ignored": 1, |
| "maximum_families_if_homology_exclusion_were_ignored": 1 |
| }, |
| "family_split_audit": { |
| "required_partitions": [ |
| "train", |
| "validation", |
| "test" |
| ], |
| "assignment_unit": "complete protein homology family", |
| "label_independent_assignment_required": true, |
| "candidate_families_before_scientific_gates": 3, |
| "eligible_families_after_scientific_gates": 0, |
| "split_feasible": false, |
| "reason": "Even the permissive dimensional pool would place only one family in each partition with no within-partition family diversity. After existing scientific gates there are no eligible families, so a leakage-aware split cannot be written." |
| }, |
| "incompatibility_counts": { |
| "heterogeneous_fitness_rows_explicitly_excluded": 245945, |
| "governed_kcat_rows_not_relabelled": 177, |
| "unikp_kcat_adapter_rows_not_relabelled": 16838, |
| "unit_ambiguous_lunzer_rows_not_relabelled": 1024, |
| "mass_specific_candidate_rows": 23, |
| "mass_specific_candidate_families": 3, |
| "mass_specific_rows_rejected_by_assay_or_consistency": 13, |
| "mass_specific_rows_rejected_by_existing_saturation_or_homology_gates": 10, |
| "eligible_activity_rows": 0, |
| "eligible_activity_families": 0 |
| }, |
| "blocking_findings": [ |
| { |
| "code": "ENDPOINT_SEMANTICS_INCOMPATIBLE", |
| "severity": "fatal", |
| "detail": "kcat, kcat/Km, fitness, inhibition, instrument response, relative activity, and simulated flux are distinct endpoints and cannot be pooled as log10_activity." |
| }, |
| { |
| "code": "MASS_SPECIFIC_POOL_TOO_SMALL", |
| "severity": "fatal", |
| "detail": "Only 23 dimensionally compatible candidate rows in three families exist before scientific gates; at most 10 coherent rows remain in one family before homology and saturation gates." |
| }, |
| { |
| "code": "NO_ELIGIBLE_FAMILY_SPLIT", |
| "severity": "fatal", |
| "detail": "No family remains eligible after the locally documented assay-validity, conversion-consistency, saturation, and development-homology gates." |
| }, |
| { |
| "code": "CANONICAL_UNIT_PREMATURE", |
| "severity": "fatal", |
| "detail": "Defining log10(umol min^-1 mg^-1) as a training endpoint would imply unsupported comparability across apparent Vmax, coupled assays, fixed-condition specific activity, and inconsistent conversions." |
| } |
| ], |
| "conditions_to_unblock": [ |
| "Acquire direct purified-enzyme activity rows with a single prespecified physical denominator and auditable conversion to it; do not substitute kcat or fitness.", |
| "Preserve per-row source file/cell, DOI or stable record, exact construct sequence, defined substrate structure, assay conditions, replicate/error semantics, evidence tier, and license provenance.", |
| "Prespecify whether the endpoint is saturating mass-specific Vmax or activity at a fixed substrate protocol; do not pool those assay semantics without a validated normalization model.", |
| "Resolve coupled-assay saturation and SlDefH conversion inconsistencies before admission.", |
| "Acquire enough independent homology families to populate train, validation, and test with meaningful family diversity, then freeze a label-independent family assignment.", |
| "Only after those gates pass, define the canonical unit, implement a strict activity-specific adapter, and run a CPU training smoke." |
| ], |
| "evidence_files": [ |
| "README.md", |
| "src/biocandidate/data/normalized_kinetics.py", |
| "src/biocandidate/data/enzengdb.py", |
| "src/biocandidate/data/lunzer.py", |
| "configs/temporal_absolute_kinetics_protocol.json", |
| "artifacts/unikp_audit.json", |
| "artifacts/external/enzengdb-v1/audit.json", |
| "artifacts/external/absolute-kinetics-screen/dryad-4964723/selection.json", |
| "artifacts/external/absolute-kinetics-screen/dryad-4964723/homology-audit.json", |
| "artifacts/external/temporal-absolute-kinetics/registry-readiness-audit.json", |
| "artifacts/external/temporal-global-family-audit/readiness-audit.json", |
| "artifacts/external/temporal-absolute-kinetics/biostudies-S-EPMC10589216/provenance.json", |
| "artifacts/external/temporal-absolute-kinetics/biostudies-S-EPMC12364357/provenance.json", |
| "artifacts/external/temporal-absolute-kinetics/candidate_registry.json", |
| "artifacts/external/temporal-absolute-kinetics/third_round_discovery_registry.json" |
| ] |
| } |
|
|
