This proposal reviews all 188 seeded source annotations from 192 raw GOA rows. Four raw rows duplicate complete source tuples; the temporary projection accounts for every raw row including evidence, reference, qualifier and WITH/FROM. The 188 source objects are unchanged. No alternative-product list was seeded, and none has been invented. The normal UniProt record, raw GOA and 86 publication/Reactome caches remain byte-identical to the preparation baseline.
BLM is an ATP-dependent RecQ DNA motor with established 3-prime-to-5-prime polarity. The first core describes that motor and its DNA work in homologous recombination repair, fork processing and telomere maintenance. The original human helicase study establishes polarity, and selected full-length biochemical and complementation results directly test ATPase activity, fork unwinding, junction migration, double-Holliday-junction dissolution and sister-chromatid-exchange rescue. TOP3A supplies the topoisomerase chemistry in the BLM–TOP3A–RMI1–RMI2 complex; that chemistry is not assigned to BLM. The second core is DNA/DNA annealing, supported by the purified catalytic-fragment experiment and fork-regression biochemistry. Concentration and construct limits are explicit. These cores summarize existing annotations; no NEW assertion is proposed.
The draft distinguishes enzyme-partner binding from the partner enzyme's catalytic activity. TOP3A binding is refined to DNA topoisomerase binding; RAD51, FANCJ and SUV3 contacts support ATPase binding; the PLK1 contact supports protein kinase binding. FEN1 biochemistry explicitly reports direct interaction and helicase-independent stimulation of FEN1 endo- and exonuclease activities, supporting an endodeoxyribonuclease-activator refinement. The WRN study explicitly reports direct binding and inhibition of WRN exonuclease activity, supporting deoxyribonuclease inhibitor activity. DNA2 is refined to enzyme binding: substrate generation by BLM unwinding is not by itself direct activation of the nuclease.
Other generic interactions remain individually justified non-core records or unresolved source assertions. RPA1, RMI1, shelterin partners and SPIDR are not assigned an invented catalytic class. SPIDR is the scaffold in its study; BLM is not made an adaptor by reversing that relationship. Selected NABP2/hSSB1 Results support cellular association but did not detect binding between the purified recombinant proteins. The draft preserves that directness limit. The literal P42677 partner in PMID16864798 is RPS27, not the abstract's MPS1/TTK. Its exact pair is unresolved, while the separate P53350/PLK1 association is directly described by the abstract. MitoCheck, the HeLa interaction survey and OpenCell pair-level datasets were not inspected and remain undecided rather than being removed.
Selected irradiation Results distinguish BLM accumulation during a G2 delay from induction of the checkpoint itself. The checkpoint-signaling assertion is therefore refined to the supported cellular DNA-damage response. The CDC2-regulation assertion remains undecided because its abstract treats BLM phosphorylation as a readout and the complete experiment was unavailable. The precise meiotic lateral-element localization, the partial PRIMPOL-cache interaction record, and the camptothecin-specific BLM/SPIDR experiment also remain unresolved; no wrong-gene claim is inferred from an abstract or incomplete extraction.
The stem-cell telomere study distinguishes C-circles, T-circles and C-rich overhangs. Its BLM depletion result concerns C-circle abundance; the positive T-circle assay concerns NBS1, and separate XRCC3 experiments concern overhangs. This does not establish universal absence of a BLM role in t-circle formation. The existing TAS assertion remains undecided. The damaged telomeric D-loop paper was abstract-only in the normal cache; selected primary-paper search excerpts were accessible, but direct page access failed and a separate telomeric-G4 binding panel was not recovered. The exact G4 assertion remains undecided, while DNA engagement and D-loop unwinding retain the support actually available.
All 49 complete cached abstracts and all 37 Reactome summaries were read. Twelve caches have full_text_available=true, but that flag does not guarantee a complete extraction: the FANCJ, SPIDR, PRIMPOL and OpenCell bodies expose only specified portions. Selected prose from those twelve caches is itemized in the temporary reading-scope receipt; no whole-paper, image or supplementary-data audit is claimed. The human BLM transgene localization in Drosophila S2 cells is explicitly distinguished from host identity. The human SUV3 binding assay is distinguished from mouse knockout phenotypes. Reactome nucleoplasm annotations remain cellular-location assertions; event titles naming PALB2, BRCA2, nucleases or kinases do not assign those enzymes' activities to BLM.
The actual PTHR13710 PAINT IBD records were inspected. Target BLM among experimental descendants is legitimate support, not circularity; short source lists were not treated as weakness. The cytoplasmic IBA remains a secondary localization inference because the read human nuclear studies do not demonstrate loss of every cytoplasmic pool. No phylogenetic re-curation is claimed. The cached GO-CAM entry places BLM in a glucocorticoid-response target map with a root molecular-function term; it does not justify a new catalytic assertion. No NEW term is proposed and no unperformed comparator check is claimed.
Author-selected term definitions and relationships were checked against the dated local GO-plus snapshot and available official AmiGO pages. The topoisomerase-binding definition covers DNA topoisomerases, and ATPase binding describes the bound enzyme class. Several live endpoints failed; this is a bounded ontology check, not an exhaustive current search. The local GO:0043138 definition text has a polarity inconsistency with its 3-prime-to-5-prime label; the current official complex page and primary BLM polarity assay support that established label. Machine-sourced GOA IDs/labels were not changed. The review raises the distinction between non-crossover BTR dissolution and endonucleolytic junction resolution as an ontology question.
The single authorized normal Falcon operation required a startup correction. The first invocation failed to initialize an inaccessible uv cache (68f3ad). With the authorized writable UV_CACHE_DIR/UV_TOOL_DIR/UV_TOOL_BIN_DIR, workspace virtualenv PATH and offline/no-sync settings, the corrected invocation failed because deep-research-client[cyberian]==0.2.7rc1 was unavailable to the offline resolver (1a1714). The scientific provider did not start in either invocation, no provider output exists, and no further attempt was made. These are startup failures, not generated research. No manually written provider-named report was created.
The proposal currently has 117 ACCEPT, 30 KEEP_AS_NON_CORE, 28 MODIFY and 13 UNDECIDED decisions, with no REMOVE, MARK_AS_OVER_ANNOTATED or NEW. Verbatim snippets are exact normal-cache substrings and total at most eighteen words per source across this proposal; that is the assistant's source-quotation boundary, not a repository validator rule. Notes paraphrase the evidence. ROOT's independent annotation review and canonical application remain pending; this preparation changes only temporary proposal files.
Focused temporary validation passed (start36cd56, terminal954d10), including schema, authored terms, reference titles and verbatim supporting text. GOA auto-discovery was disabled because this is a temporary proposal outside the gene directory; the exact 192-to-188 raw-source projection was checked separately. Fifteen warnings are the validator’s generic-binding policy advisories. The user’s ActionEnum governs retention of justified non-core interactions; no unsupported removals were made to silence those warnings. This is not a canonical or repository-wide validation.
The independent reviewer identified six descriptions of inference methods that did not match the unchanged source records. Row72 is an InterPro ATP-binding inference (IPR011545/IPR012532); row122 is an InterPro DNA-replication inference (IPR012532/IPR018982); row124 is an InterPro DNA-repair inference (IPR018982); row126 is an InterPro DNA-recombination inference (IPR004589). Row150 is the combined-IEA method GO_REF:0000120 with ARBA00043426 and RHEA:13065 in WITH/FROM, and row169 is the ARBA rule ARBA00086129 under GO_REF:0000117. The corresponding reasons now name those actual source methods. Two duplicated-article typos in the HeLa survey reasons were also corrected.
These eight changes affect review.reason only. All 188 source objects, actions, evidence links, quotations, references, core functions and the absent alternative-product slot are unchanged from version1. The complete first proposal is preserved in tmp/BLM-review-proposal. Provider startup failures and source-reading limits recorded above remain applicable. No source cache, canonical review or history was changed by this revision.
Version2 focused validation also passed (startb6015f, terminal46f1b1), with the same fifteen generic-binding policy warnings. Schema, authored terms, titles and exact supporting text were checked; the unchanged raw-source projection and source-reading evidence were reused.
ROOT independently assessed all 188 annotation decisions and both cores, then approved the corrected second proposal (78ddc1). The exact reviewed YAML was applied (f5e6b7) and passed focused canonical validation (c9e625), HTML rendering (c81be2), and history validation (2dc205). Fifteen generic-binding policy advisories remain for source-supported non-core associations under the supplied ActionEnum. Thirteen annotations remain undecided because the relevant assay or dataset could not be fully adjudicated. The rendered page embeds the exact YAML; no repository-wide validation, whole-paper audit or scientific-provider output is claimed. All raw source and reference cache bytes remain unchanged.
The review feedback on PR3706 was assessed against the unchanged source rows. The WRN inhibitor refinement is retained with explicit laboratory scope and a cellular-context question. Selected publisher Results, including Figure 6 controls, support the GO:0060703 definition. This activity remains outside the biological cores. The source IPI qualifier and partner are unchanged; the biochemical interpretation is recorded in the review rather than relabeling the original evidence.
DNA2 binding is refined from enzyme binding to ATPase binding, using the original human interaction experiment and the independently checked human DNA2 enzyme class. Binding an ATPase is distinct from activating its ATPase activity. FEN1 retains the narrower endodeoxyribonuclease-activator proposal supported by its abstract. That term represents the endonuclease component; the reported exonuclease stimulation remains explicit in its reason, and no unverified second regulator term is added.
The already accepted GO:0071139 process is now represented in core1 with the existing PMID23543748 reference. Its mechanism is BTR dissolution with non-crossover products. BLM performs DNA remodeling; TOP3A performs strand cleavage and rejoining. This is synthesis of an existing assertion, not a NEW annotation or an endonuclease assignment. The existing ontology question remains because the distinction between dissolution and endonucleolytic resolution should remain explicit. Short exact cached anchors were added for the TRF2, DNA2 and dissolution rows.
The fifteen retained generic-binding rows remain source-supported non-core interactions under the supplied ActionEnum. A generic term is uninformative, but that alone does not show a real interaction is unlikely to be correct. No action was changed to suppress the validator's differing policy. No blanket claim of physiological mechanism is made from partner-class binding.
Reading for this follow-up covered five complete cached abstracts (11919194, 12181313, 14688284, 21325134 and 23543748), selected indexed publisher text for PMID11919194, the human DNA2 record's relevant fields, and the specified GO definitions/parents. Several direct publisher, PMC and ontology endpoints failed; accessible indexed primary excerpts are distinguished from successful direct full-page access in the temporary source-reading record. No full-paper, figure-image or supplementary-data audit is claimed. Normal cache availability flags and bytes are unchanged. The earlier provider startup failures remain applicable; no provider or normal fetch was invoked for this follow-up.
All 188 source objects, absent alternative-product slot, original evidence/qualifiers/partners, reference identities/titles and action counts remain unchanged (117 ACCEPT, 30 KEEP_AS_NON_CORE, 28 MODIFY, 13 UNDECIDED). The two cores remain; only the first gains the already accepted process term and its source. Quotes total at most eighteen words per source, including prior snippets, below the assistant's 25-word boundary; this is not a repository validator limit. This temporary proposal awaits independent science review and canonical application.
Focused temporary validation passed (corrected start97d125, terminalbe14f1): schema, authored terms, reference titles and supporting text passed with the same fifteen generic-binding policy warnings. The first invocation stopped at the unwritable default uv-cache initialization (360168/af9ea6); its exact logs were preserved, and the established writable-cache/offline environment resolved startup. GOA auto-discovery was disabled only because this proposal is outside the canonical gene directory; all 188 source objects were compared exactly with the frozen canonical review. No canonical file or source cache was changed.
Independent science review passed for this focused follow-up on 2026-10-01 UTC. The exact approved YAML and appended notes were applied after matching canonical preimages and the published PR head. All 188 source objects, the absent product slot, and all actions are preserved. This follow-up clarifies the laboratory scope and controls for WRN inhibition, refines DNA2 binding to the verified ATPase class, adds exact evidence anchors and represents the existing recombination-intermediate process in the BTR core. Focused canonical validation, rendering and a new history record follow this application.
Focused canonical validation, HTML rendering and history validation passed on 2026-10-01 UTC. The rendered page embeds the approved YAML. Fifteen generic-binding policy advisories remain for supported non-core associations under the supplied ActionEnum. Those advisories do not change the bounded evidence decisions or create a new annotation. No full-repository validation is claimed. All normal sources and previous history remain unchanged.
Fifteen GO:0005515 protein binding rows were migrated from KEEP_AS_NON_CORE to REMOVE to follow the repository policy for new and re-reviewed generic binding annotations. Their partner evidence remains recorded in review.reason, but no retained GO annotation now keeps the literal generic term solely because the interaction itself is source-supported.
The RPA1, RMI1, shelterin, RAD51D, SPIDR, UIMC1/RAP80 and NABP2/hSSB1 rows still deliberately avoid unsupported partner-class refinements. RPA1 is a DNA-binding cofactor rather than an ATPase or kinase. RMI1 is represented through BTR complex membership rather than an invented BLM adaptor function. TRF1, TRF2, POT1 and SPIDR regulate or recruit BLM in the cited assays rather than showing that BLM activates those nonenzymatic partners. UIMC1/RAP80 binds ubiquitylated BLM via RAP80 UIMs, and the BLM-NABP2 association is cellular but not detectably direct with the purified proteins tested. Removing GO:0005515 is therefore a specificity decision; it does not assert that those interactions are false.
This entry supersedes the action policy used in the 2026-10-05 migration above; the older journal entries are retained as history. The applicable authority is the published ClinGen project instruction:
retain a supported, biologically correct
GO:0005515(protein binding) annotation asKEEP_AS_NON_COREwhen no evidence-backed, more specific replacement has been established.
The fifteen removed rows were reassessed individually against their cited evidence. All fifteen now retain supported associations as non-core. No partner activity is transferred to BLM, and no narrower molecular function is inferred merely from a partner name. The other actions remain unchanged, including the eight unresolved generic-binding assertions and the camptothecin-process assertion discussed below. Final counts are 117 ACCEPT, 30 KEEP_AS_NON_CORE, 28 MODIFY and 13 UNDECIDED across the same 188 source annotations. Both core functions, all 92 reference identities/titles, all source objects and the absent alternative-product slot are preserved.
The source-specific gaps have now been checked in selected external primary Results, beyond the immutable partial caches:
The remaining decisions use the complete cached source abstracts and, for RMI1 and NABP2/hSSB1, selected cached Results. Matching IntAct records additionally verify human TRF1/TRF2/RPA1/NABP2 pair-to-publication linkage; they share literature provenance and are not independent replication. Empty query results for other papers are not evidence that an interaction is absent. RPA-dependent BLM fragment assays and the POT1 splice-variant context remain explicit. The hSSB1 co-IP does not establish direct purified binding. The RAP80 association remains explicitly ubiquitin-dependent and does not make BLM a ubiquitin reader.
A bounded RAD51D partner-class check found an older DNA-dependent ATPase description. However, PMID:37344587(https://pmc.ncbi.nlm.nih.gov/articles/PMC7614784/) provides human BCDX2 structural and biochemical evidence distinguishing RAD51B/C ATP hydrolysis from RAD51D/XRCC2 nucleotide binding, including catalytic-residue and mutant analyses. Selected Results were read from the genuine publication cache already in the repository. Accordingly, ATP binding or a RAD51-paralog name is insufficient here to justify changing the original BLM–RAD51D assertion to ATPase binding. This does not claim that BLM cannot associate with an ATPase in other contexts.
The newly accessed SPIDR Results also describe BLM foci after camptothecin and hydroxyurea exposure. The associated access statements are corrected. The camptothecin process row remains UNDECIDED because the inspected panels lack an untreated comparison for stimulus-induced recruitment. This is an explicit assay-interpretation limit, not an inaccessible-paper claim or a requirement for survival evidence. SPIDR perturbations are not attributed to BLM. Other source-associated actions remain unchanged. No NEW assertion is proposed.
All inherited supporting quotations are unchanged and remain at most eighteen words per scientific source across the review; no new scientific quotation was added to these notes. Reference access notes for the three newly inspected papers distinguish external access from cache completeness. All normal caches remain unchanged. This is a bounded follow-up, not a whole-paper audit or a new gene-wide review.
The camptothecin row now follows the positive recruitment evidence in the
selected Results and Figures 1G/2 of PMID:23509288(https://pmc.ncbi.nlm.nih.gov/articles/PMC3696769/).
The authors report CPT-induced BLM foci and SPIDR-dependent recruitment; the
broad cellular-response term does not require a BLM survival endpoint or an
untreated control in the subset of panels inspected here. The previous
UNDECIDED rationale was an unnecessarily restrictive interpretation of that
access boundary and is superseded. SPIDR depletion/rescue is still attributed
to SPIDR. Hydroxyurea has additional independent support from PMID:28506294,
but that asymmetry does not invalidate the original CPT recruitment evidence.
No direct survival protection, new process annotation or core function is added.
The 188 source assertions now comprise 117 ACCEPT, 31 KEEP_AS_NON_CORE, 28 MODIFY
and 12 UNDECIDED; all other actions remain unchanged.
The interaction evidence extract supplies
16 exact curated records for the cited TRF1/TRF2, RPA1 and NABP2 contacts,
including their IntAct/IMEx identifiers, source-specific query URLs, methods,
species, construct features and response hashes. Colocalization-only records
are excluded from this binding-evidence extract. These database records share
provenance with the source publications and do not constitute independent
replication. No new scientific excerpt is added.