Audit & review
Scientific qualifications
Inspect recorded decisions, open scientific issues and the evidence supporting them.
The recorded scientific audit distinguishes corrected overlays, resolved interpretation limits and open scientific issues. It includes a purposeful 36-entry structural sample; it is not a manual validation of all 230 deposits or an external peer review.
Download the complete review queue · Six isoform corrections
| Finding / severity | Scientific issue | Decision or next action | Status and evidence |
|---|---|---|---|
| AF02 major | Local PWWP confidence and full-length confidence differ; inter-PWWP PAE is 25–28 Å in NSD1/2/3. | Use local core confidence and preserve domain-pair orientation uncertainty. Make no experimental-support claim from pLDDT. | resolved_caution alphafold_domain_confidence_audit.csv; canonical_interdomain_PAE.csv |
| FAM02 major | Six isoform/domain rows differ from source-anchored splice correspondence. | Apply separate v1.1 overlay: 106 complete/11 partial/14 absent models under explicit edit mapping. Preserve v1.0 values and method provenance. | corrected_overlay v1_1_isoform_core_overlay.csv |
| FAM03 major | PWWP4 (A0A494C071) is reviewed but PE 5; UniProt cautions that CDS prediction is dubious. | Keep annotation-supported candidate with uncertain protein existence; AF confidence is not biological confirmation. | open_biological_uncertainty sources/UniProt_A0A494C071.json |
| FAM04 major | MBD5 UniProt interval 1385–1409 is 25 residues, whereas CDD cd20141 spans 1383–1466. | Do not treat the 25-residue annotation as a complete PWWP fold or include it in global-domain comparisons without a separately justified boundary. | open_boundary_adjudication domain_boundary_audit.csv; https://www.ncbi.nlm.nih.gov/Structure/cdd/cd20141 |
| FAM05 major | KDM3A/B/JMJD1C are PF22988/IPR054504 predicted PWWP-like members. | Keep the separate predicted tier. 5RAM/4C8D are KDM3B catalytic-region structures starting at 1380, not observations of PWWP-like residues 90–188. | resolved_scope sources/InterPro_IPR054504.json; sources/RCSB_5RAM_entity1.json; sources/RCSB_4C8D_entity1.json |
| STR02 major | 11NZ/11OA have complete annotation-position coverage with many absent aromatic side chains. | Exclude affected chains from pocket-geometry/atomistic-ready claims; backbone comparisons retain their explicit limited meaning. | resolved_caution structure_residue_audit.csv; structure_expert_decisions.csv |
| STR03 major | 2NLU/5XSK have large sequence-fixed single-chain RMSDs in known domain-swapped contexts. | Discuss swapping; do not label as AF failure. Report diagnostic fragment/ensemble sensitivity separately from whole-core metric. | resolved_caution outlier_ensemble_sensitivity.csv; https://doi.org/10.1016/j.jmb.2007.01.010; https://doi.org/10.1038/s41598-017-18510-8 |
| STR04 major | 5XSK title/publication human designation conflicts with deposited rat Hdgf/Q8VHK7 source mapping. | Retain source-mapped rat reference and visible bibliographic conflict; provenance cannot independently settle biological specimen origin. | open_source_conflict structure_source_mapping_audit.csv; https://www.rcsb.org/structure/5XSK |
| STR05 major | The 6R3E entity has duplicated sequence but only its first copy is observed; 6L87 contains a deliberate deletion and substitutions. | Keep exact construct maps and engineering flags; no native tandem PWWP claim or unqualified wild-type designation. | resolved_caution structure_source_mapping_audit.csv; structure_expert_decisions.csv |
| STR07 major | Mixed DNMT3A/B deposits do not observe both PWWP cores; all 230 inventory rows have at most one observed native PWWP domain ID. | Multiple protein/domain associations must not be confused with co-observed tandem domains. The statement about all 230 deposits is based on inventory metadata, with risk cases independently sampled. | resolved_scope structure_expert_decisions.csv; original Inventories/experimental_records.json |
| STR08 major | 2NAS/6L87 discrepancies persist through 20 NMR conformers each. | Retain unexplained context-dependent discrepancies; do not attribute solely to first-conformer selection or imply that all predicted cores are experimentally validated. | open_structural_interpretation outlier_ensemble_sensitivity.csv |
| LIG-001 major | 31 numeric rows were liable to be interpreted as 31 distinct active ligands. | Use 31 measurements = 23 Kd + 8 interaction/cellular effects; 12 positively measured named small-molecule preparations, 7 exact CCD identities, and one nanobody. | resolved_v1_1 assay_audit_36.csv; graph data/counts.json |
| LIG-003 major | Compound11 mixture measurements cannot identify the affinity of either deposited stereoisomer. | Keep preparation-level affinity with chemical_id null; link A1IEG and A1IFS as deposited candidates without affinity transfer. | resolved_v1_1 S05 Table 1 and Results; 9FOC CCD descriptors |
| LIG-005 major | The H4K20me3 peptide and M3L monomer were conflated; M3L was absent from the chemical inventory. | Keep whole 11-mer preparation distinct, add source-verified M3L monomer identity, and retain its polymer context. | resolved_v1_1 S07 Table 1/Table 2; S11 CCD:M3L; 3QBY original CIF |
| LIG-006 major | Seven NSD3 counter-screen results in the same Collie table were omitted from v1. | Add four Kd >30 µM rows and three qualitative no-binding rows; do not invent thresholds for the latter. | resolved_v1_1 S05 Table 1 NSD3(263–398) column; new_evidence_records_10.csv |
| LIG-008 major | HS13 primary abstract includes Kd 8.33 µM, but method and exact identity have not been verified. | Retain abstract_reported_value separately; retrieve full article/SI and exact structure before promoting to quantitative assay inventory. | open_source_and_identity_review S08 PMID42442058 abstract; assay0035 |
| LIG-009 major | Several source checks rely on selected indexed full-text passages, rather than complete article/supplement inspection. | Use exact source_verification_scope. For detailed protocol comparisons inspect original figures and SI for S01/S02/S04; do not describe these as complete full-text reviews. | open_source_scope_limit source_ledger.csv; S01/S02/S04; Figure S2 remains inaccessible |
| LIG-010 major | Assay construct bounds are curated for 21/31 numeric rows; temperature remains uncurated for 31/31. | Before quantitative pooling recover missing constructs, buffers, temperatures, DMSO and incubation conditions. Uncurated values remain null. | open_protocol_curation audit_summary_counts.json; assay_audit_36.csv |
| LIG-011 major | Five positively measured preparations lack one strict full chemical identity: compounds5/6/11/12/13. | Keep 5/11 mixtures distinct; source-resolve 6/12/13 stereochemical structures and assayed preparations before increasing the 7 exact-identity count. | open_chemical_identity_review phase2_curation.json; S05 chemistry Results; 9FOC stereo CCDs |
| LIG-012 major | Crystal construct and assay construct can differ, including seven 9FOE surface substitutions. | Retain separate construct annotations. Do not propagate Table 1 affinity to a mutated crystal construct or pool211–350 with208–368 measurements. | resolved_modeling_rule; detailed_protocol_review_open S05 Results/Table 1; construct_dependence.csv |
| LIG-013 major | Extra-chain proximity can resemble a broader or alternative binding site. | Treat 7,684 pairs as geometry. Review author biological assemblies, density and packing before functional site assignments; prioritize7VLN/7QC,9EXW/A1H7X,9EXX/A1H7Y. | open_biological_site_review secondary_chain_contacts_review.csv; original CIF coordinates |
| LIG-015 major | Viral integrase inhibitors and methyltransferase cofactors occur in entries containing a PWWP-family protein. | Retain their actual contacted partners; exclude these structure associations from PWWP active-ligand counts. | resolved_v1_1 7OUF/1L0,7OUG/RLT,7OUH/KLQ;11NZ/SAH/ZN; principal_contact_summary.csv |
| LIG-016 major | Ideal CCD SDFs cannot substitute for a bound pose or a C294 covalent adduct. | Use original CIF and selected_source_pose_atoms; retain struct_conn and partner SG coordinates. No reconstructed free-ligand pose or inferred adduct chemistry is exported. | resolved_representation_rule deposited_connections_audit.csv; selected_source_pose_atoms.csv |
| LIG-017 major | The 36-row inventory is a selected evidence set, with other primary-paper assays uncurated. | Scope manuscript and database coverage explicitly. A later systematic expansion should cover full tables, other nanobodies, additional NSD2 series, covalent compounds and BPP negative targets with verified methods. | open_coverage_scope_limit S05/S06/S07 full tables; deposited primary citations for7VLN,9EXW,9EXX,9KN9 |
| AF03 minor | Q586Y0 official versus community core mean pLDDT is 92.424 versus 67.422. | Use the official canonical model as primary comparison and label the community model separately; neither confidence score establishes truth. | resolved_provider_policy alphafold_domain_confidence_audit.csv |
| FAM06 minor | Reviewed annotations list RRM domains for PUF60 and an MBD domain for MBD6, with no positive PWWP assignment. | Retain exclusions; absence of annotation is not proof against every future structural discovery. | resolved_scope sources/UniProt_Q9UHX1.json; sources/UniProt_Q96DN6.json |
| STR06 minor | 9KNB A259 has C/O but no Cα; any-atom and Cα counts differ. | Use any-heavy-atom coverage for the inventory and explicitly Cα coverage for structural fit eligibility. | resolved_denominator structure_residue_audit.csv |
| LIG-002 moderate | A1A39 and W0J share the same complete InChIKey. | Merge exact identity nodes while retaining both CCD aliases; preserve other shared-connectivity groups as separate full identities. | resolved_v1_1 identity_alias_and_stereochemistry_audit.csv; all 140 original CCD metadata reread |
| LIG-004 moderate | Bis-Tris records0033/0034 lacked CCD identifiers. | Resolve to BTB and retain crystallization-buffer role; do not create potency values. | resolved_v1_1 S07 Results/crystallization methods; 3LLR/3QKJ original CIF |
| LIG-007 moderate | NbH10 functional IC50 lacked the accompanying direct-binding evidence. | Add apparent SPR Kds of 94.4 ± 9.1 nM (PSIP1) and 81.2 ± 1.8 nM (HDGFL2); keep nanobody separate from small molecules and absence of bound complex explicit. | resolved_v1_1 S06 Table 1 and SPR methods |
| LIG-014 moderate | Twenty short-construct contacts lie outside both chosen annotation windows. | Keep candidate context status, especially UNX/ions. Do not label all as validated PWWP-fold ligand contacts. | open_boundary_context_review binding_sites contact_class=isolated_construct_contact_outside_annotations |
| LIG-018 moderate | UNL and UNX remain unidentified chemical components. | Preserve separate unresolved placeholders and do not count them as chemically identified or active compounds. | unresolved_source_identity 140 CCD source metadata; deduplicated_chemicals.json |
| LIG-019 moderate | Native core and raw Pfam intervals can overlap without one fully containing the other. | Interpret their separate contact lists as source-specific windows; combined contact evidence and native-domain boundary audit are required for fold-level conclusions. | documented_method_limit EVIDENCE_ONTOLOGY.md; structure_chains.json |
| AF01 verified | All 131 coordinate sequences, accession/taxon identities and pLDDT records agree with original official records. | Separate 39 official canonical, 91 official isoform and 1 community model; source snapshot validation is not a fresh AFDB census. | resolved_scope alphafold_model_audit.csv |
| FAM01 verified | All 39 protein identities and all 42 annotation boundaries match original official source records. | Retain identities and boundaries as source annotations; do not call all 27 established-tier proteins experimentally confirmed. | resolved_scope family_identity_audit.csv; domain_boundary_audit.csv |
| STR01 verified | The 36 sampled deposits independently reproduce presence and any-heavy-atom coverage, with two parser backends. | State 36 purposefully sampled entries, 113 chain-domain rows and 4,576 residue rows; never call this a full manual 230-entry audit. | resolved_scope structure_sample_summary.json; parser_crosscheck.csv |
No records match this filter. Reset to view the complete table.