DnaK annotation review notes

2026-08-29 comprehensive audit

• Reconciled all 61 qualifier-aware review signatures against the current DnaK GOA cache. The qualifier distribution is 39 enables, 13 involved_in, 6 located_in, 2 part_of, and 1 is_active_in. The physical GOA contains additional lines where one reviewed IPI signature has multiple WITH/FROM interactors; those interactors are retained as annotation-level supporting_entities rather than duplicated review decisions.
• DnaK's central activity is the ATP-dependent Hsp70 foldase cycle with DnaJ and GrpE. The direct folding evidence states that “Several rounds of ATP-dependent interaction with DnaK and DnaJ are required for fully efficient folding” PMID:7937953. GO:0140662 is therefore the most specific core molecular function, while GO:0044183 remains a valid broader IBA-supported foldase term.
• DnaK also has directly demonstrated ATP-independent antiaggregation/holdase activity: it “protects the host RNA polymerase enzyme from heat inactivation in an ATP-independent reaction” PMID:2203539. This is distinct from the ATP-dependent disaggregation/reactivation reported in the same abstract. GO:0140309 is carrier-specific and no acceptor/delivery step is demonstrated, while GO:0051082 is obsolete; the review therefore proposes the project-standard general holdase chaperone activity NTR rather than placing either term in the authored core-function slot [file:projects/UNFOLDED_PROTEIN_BINDING.md].
• Current PANTHER PAINT provenance was retrieved with the public wrapper just fetch-panther-paint PTHR19375 --extra-uniprot P0A6Y8 on 2026-08-29. All five IBA claims remain present: GO:0016887, GO:0031072, GO:0044183, and GO:0042026 at PTN000452648, and GO:0005829 at PTN000751549. Each node and its exact descendant evidence set is recorded in propagation_review; target self-evidence is treated as valid descendant evidence, not circularity. GO:0031072 is accepted because the PAINT placement is sound and the term preserves experimentally supported heat-shock-partner interactions with HtpG and ClpB rather than collapsing them into DnaK's separate GO:0051087 assertions.
• The sole previous UNDECIDED, GO:0008270 zinc ion binding, is now KEEP_AS_NON_CORE. The abstract directly reports that “nine zinc-binding proteins were newly identified including: acetate kinase (AckA), DnaK, serine hydroxymethyltransferase (GlyA)” PMID:11985624. The cache is abstract-only and does not establish physiological relevance, binding site, or specificity, so the observation is retained without promoting it to a core function.
• The CAFA GO:0051087 IPI signature citing PMID:9103205 has WITH/FROM UniProtKB:P08622 (DnaJ), while the cached abstract focuses on GrpE-DnaK and does not mention DnaJ. Because the cache is abstract-only, this absence is not sufficient to overrule an experimental curator with full-text access. The row remains ACCEPT; independently cited DnaK-DnaJ IPI rows corroborate the biological interaction.
• Generic GO:0005515 interaction rows were handled partner-by-partner. Functionally diagnostic interactions with DnaJ, GrpE, or ClpB are MODIFY to GO:0051087 or GO:0031072; heterogeneous interaction screens and true but mechanistically uninformative partners are KEEP_AS_NON_CORE. Exact GOA partner accessions are recorded in supporting_entities.
• The two RpoH/sigma32 GO:0005515 IPI rows are KEEP_AS_NON_CORE: physical interaction evidence does not by itself assert sigma factor antagonist activity. The latter remains a core function on its independent PMID:8599944 IDA evidence.
• The obsolete GO:0051082 IEA is replaced by the general chaperone term GO:0044183 because the InterPro family mapping does not itself establish ATP dependence; the two direct biochemical IDA rows are replaced by GO:0140662 because DnaK's ATP-dependent cycle is experimentally established.
• Evidence restraint was applied throughout. Of the physical PMID caches, only PMID:18304323, PMID:24561554, and PMID:35289645 contain full text; the remaining cached records are abstract-only. Experimental annotations were not removed merely because an abstract lacked assay detail.
• Final physical-row actions: 33 ACCEPT, 14 MODIFY, 14 KEEP_AS_NON_CORE, 0 REMOVE, 0 MARK_AS_OVER_ANNOTATED, 0 UNDECIDED, and 0 PENDING. The review is marked COMPLETE because every current signature has an evidence-aware decision, all IBA rows have current node provenance, and the remaining biological uncertainties are captured as questions and experiments rather than undecided annotations.
• Reconciled the coupled BioReason-Pro SFT prediction for GO:0008270 from NPI to CNN. The direct radioactive Zn(II)-binding result and pre-existing GOA IDA make the prediction correct but not novel; the negative structural survey remains useful as a limitation on mechanism and physiological interpretation, not as a refutation of the assay PMID:11985624.