This manual review preserves all 41 seeded annotation objects from 43 source rows, including the original evidence codes, references, qualifiers and partner identifiers. The three named UniProt products remain unchanged. None of the unqualified annotations establishes equal activity of every product. No new GO annotation is proposed.
The standard installed deep-research client attempted Falcon and the configured Perplexity fallback; both failed DNS and produced no research file. The following account is manual primary-literature work, not provider output. Source ingestion and curation are separate: existing normal publication caches are preserved, including abstract-only caches where authenticated supplemental full text was available for reading.
Sabath, Qiu and Jonas (2024), PMID:39032489 directly identify BRAT1 and WDR73 as assembly factors for the human INTS9–INTS11 cleavage module. The inspected abstract reports active-site blocking, prevention of premature INTS4 association, nuclear import assistance and IP6-dependent BRAT1 release. It supports specific endoribonuclease inhibition and Integrator assembly, not RNA cleavage by BRAT1. The complete original Methods and images were not inspected.
Lin and colleagues (2024), PMID:39032490 supplies complementary human and fly structural and functional evidence. The normal cache is abstract-only. An authenticated, separately retained full-text variant was read for all unique main Introduction, Results, Discussion and limitations paragraphs, plus the relevant construct, fractionation, imaging, structural, knockout, rescue and organoid Methods. These supplemental readings are paraphrased here; no snippet absent from the normal cache is used as a validated YAML quotation. Images, complete supplementary tables and every analytical Methods subsection were not independently audited.
The human ternary INTS9–INTS11–BRAT1 structure places the BRAT1 C terminus in the INTS11 active site. The detailed initial DCY metal-coordination account is in the fly complex; human BRAT1 is 821 amino acids and the fly Tyr974 numbering is not transferred to it. BRAT1-bound INTS9/INTS11 is incompatible with the mature INTS4-containing cleavage module. Knockout reduces INTS9/INTS11 protein levels, and wild-type BRAT1 rescues the processing phenotype more effectively than the DCY/CTE mutants. No UBE3D ligase or CPSF73 activity is assigned to BRAT1.
The two papers support one core assembly mechanism, represented by GO:0060698 endoribonuclease inhibitor activity coupled to GO:0160234 integrator complex assembly. The former is a child of the two seeded GO:0008428 assertions, so both are refined instead of adding a redundant NEW assertion. BRAT1 is an assembly factor and not a permanent holo-Integrator component; no mature Integrator in_complex assertion is introduced.
Import assistance is not universal transport dependence. Lin and colleagues found that residual INTS11 remained nuclear in BRAT1-null HCT116 cells. That observation qualifies, rather than erases, the import-assistance finding in the companion study. HCT116 and human embryonic stem-cell knockout cultures also lacked overt growth defects, whereas neural organoids developed transcriptional disruption. The latter experiments do not establish the activity of each patient allele or each BRAT1 product.
Aglipay and colleagues (2006), PMID:16452482 was read from the normal abstract. It reports BRCA1 yeast two-hybrid association, ATM binding, double-strand-break localization and reduced ATM/CHK2 phosphorylation after BRAT1 depletion. It also reports that purified BAAT1 partially blocks PP2A-mediated ATM dephosphorylation. Thus the DNA-damage evidence is more than a generic knockdown phenotype. Substrate protection remains a possible explanation: this does not establish that BRAT1 is a kinase or a general PP2A inhibitor.
So and Ouchi (2011), PMID:22977523 remains abstract-only in the normal cache. The authenticated supplemental full main text was read separately, including Methods and Results. Human BAAT1 fragments and full-length FLAG-BAAT1 recover DNA-PKcs and ATM from 293T lysates; fragment assays also recover SMC1. The NCS time course alters these associations, and knockdown reduces DNA-PKcs and SMC1 phosphorylation in human MCF10A cells. Lysate recovery does not prove purified binary binding. SMC1A is a cohesin ATPase and phosphorylation substrate, not a protein kinase; the broad wording in the UniProt functional comment is not copied into the synthesis.
The ATM, DNA-PKcs and mTOR partner assertions can be refined to GO:0120283 protein serine/threonine kinase binding. This describes the class of the binding partner and does not confer that partner's catalysis on BRAT1. The generic BRCA1, SMC1A and Raptor associations remain non-core with their source-specific assay boundaries.
Low and colleagues (2015), PMID:25631046 is abstract-only in the normal cache; a direct PMC request returned a browser challenge. Its abstract supports NDFIP1-dependent BRAT1 ubiquitination and nuclear translocation during damage responses, and the reviewed UniProt record identifies NDFIP1 binding. The physical-association row is retained with explicit curator deference; its original target-specific binding assay is not represented as independently inspected. BRAT1 is not assigned ubiquitin-ligase activity. Neither the newer Integrator mechanism nor the abstract access limit is a reason to delete the experimentally grounded DNA-damage-response annotations.
So and Ouchi (2014), PMID:25070371 was read through the complete unique extracted main text and available figure captions. Human cancer-cell knockdowns show reduced population expansion, altered cell-cycle behavior, increased basal apoptosis, increased glucose consumption/lactate and mitochondrial dysfunction. The migration evidence includes mitomycin-C pretreatment before the wound assay and a separate migration chamber, so it is not dismissed as cell-count confounding alone. These remain contextual phenotypes, not intrinsic BRAT1 mitogenic, motor or glycolytic activities.
The glucose-metabolic-process assertion is over-specific about participation in a chemical pathway: the study establishes altered metabolism after depletion, without identifying a glucose-metabolic step performed by BRAT1. The apoptosis assertion is better expressed as negative regulation of apoptosis in the tested survival context, not participation in the execution program. The positive-cell-growth assertion is over-annotated relative to its assay: GO cell growth concerns individual-cell size and biosynthetic mass, whereas the study measures cell counts, cell cycle and tumor expansion. The independently seeded cell-population-proliferation assertion already captures the measured population effect.
Mitochondria become more condensed in depleted cells, supporting a non-core organelle-distribution phenotype. This is not localization of BRAT1 to mitochondria. Mitochondrial ATP production and PDH activity fall, while total cellular ATP does not significantly change; SC79 partly rescues growth and ROS. These observations do not identify BRAT1 as a mitochondrial enzyme. Later HCT116/hESC growth findings limit generalization of the older knockdown results without refuting those cell-specific experiments.
So and Ouchi (2013), PMID:25657994 was read through the full extracted Methods and Results. Human GST-BRAT1 fragments recover mTOR and Raptor from 293-cell lysates; the negative Akt result applies to that pull-down, not every possible BRAT1–Akt relationship. Human HeLa depletion and mouse embryonic fibroblasts provide distinct contexts. The full Methods describe E13.5 embryos, while an indexed Figure 1 caption says E15; the discrepancy is recorded rather than resolved by assumption. No DOI is invented.
HuRI, PMID:32296183 supplies general screen methodology but the inspected cache does not expose the four BRAT1 source edges (A6NI15, Q13287, Q8N6Y0 and Q9H0I2). These remain UNDECIDED. The neurodegeneration interactome, PMID:32814053 is abstract-only and does not resolve the O14901/KLF11 edge. Preservation of a different BRAT1 interaction would not independently certify these particular edges.
The YTS membrane proteome, PMID:19946888 includes both integral and associated proteins. Its abstract does not expose the BRAT1 detection/peptide entry, so the membrane annotation remains UNDECIDED. Cytoplasmic or nuclear identity and lack of a transmembrane segment do not themselves refute a membrane-association experiment.
The project curation instruction retains supported, biologically correct generic binding as KEEP_AS_NON_CORE when a justified specific replacement has not been established. This is an explicit task-specific departure from the annotation-reviewer's informational-exclusion default, not an external maintainer sign-off or global policy change. Unadjudicated target evidence remains UNDECIDED. Source-specific kinase refinements are retained where supported.
Official AmiGO definition/parent views for the endoribonuclease inhibitor and Integrator assembly terms were inspected. Protein-kinase binding and apoptosis definitions were also inspected; FlyBase supplies the specific serine/threonine-kinase-binding definition. A read-only GO 2026-03-25 SQLite snapshot independently supplies exact labels/definitions, including cell growth versus population expansion. Failed direct term requests are retained as failures, not described as successful live refreshes.
All 41 seeded annotations have been assessed, with six explicitly UNDECIDED assertions. COMPLETE records completion of this assessment; it does not resolve those outstanding experiments or imply that all three products have identical activities. Existing source records and abstract-only cache variants are preserved. Source-import, validation and publication provenance are recorded separately from these biological conclusions.
The GO:0051646 decision is now MARK_AS_OVER_ANNOTATED. This supersedes the earlier non-core wording in these notes; the underlying mitochondrial-distribution observation remains reported. GO:0051646 concerns establishing or maintaining mitochondrial position. The main Results and available Figure S2 caption of PMID:25070371(https://pmc.ncbi.nlm.nih.gov/articles/PMC4129107/) describe a mitochondrial-dye morphology assay in control and BRAT1-depleted HeLa cells. Condensation after depletion does not identify a positioning step performed by BRAT1 or separate it from the accompanying mitochondrial dysfunction. The supplemental image was not inspected, and the change does not allege an erroneous experimental observation. The current tally is 12 ACCEPT, 13 KEEP_AS_NON_CORE, seven MODIFY, three MARK_AS_OVER_ANNOTATED and six UNDECIDED, across the same 41 source objects. No NEW annotation is added.
The NDFIP1 decision rests on target-specific context, not a general rule that UniProt SUBUNIT text outranks its interaction section. The abstract of PMID:25631046(https://pubmed.ncbi.nlm.nih.gov/25631046/) explicitly studies NDFIP1-dependent BRAT1 ubiquitination and trafficking; the reviewed UniProt binding statement corroborates the curated interaction. The source binding assay itself remains uninspected, and the retained row explicitly defers to its experimental curator. By contrast, the five unresolved screen edges have no inspected target-pair assay or similarly explicit target-specific account. This distinction does not invalidate the screens or independently verify the NDFIP1 binding assay.
The single core expresses the directly defined INTS11-inhibition and assembly mechanism. Earlier ATM/DNA-PK interactions and phosphorylation effects remain represented in the description, annotation decisions and research questions. Their relationship to Integrator biogenesis and to separable BRAT1 activities remains unresolved; no claim is made that GO lacks a suitable term for an independently established regulatory activity. A new research question asks about a component term for the transient BRAT1–WDR73–INTS9–INTS11 intermediate, without inventing an identifier or assigning mature Integrator membership.
The canonical abstract of PMID:22977523(https://pubmed.ncbi.nlm.nih.gov/22977523/) already supports DNA-PKcs/SMC1 association and phosphorylation changes; the abstract of PMID:39032490(https://pubmed.ncbi.nlm.nih.gov/39032490/) supports INTS11 inhibition and assembly. Detailed assay and structural descriptions elsewhere in these notes also draw on the authenticated longer variants whose reading limits are stated above. These manual readings do not convert the preserved canonical caches into full-text records. No supporting quotation or source cache is changed. The review still contains 15 references: ten PMID entries, four GO_REF entries and one local UniProt record.
The review status is corrected from COMPLETE to DRAFT. Under the repository schema, COMPLETE requires no PENDING annotations and no validation warnings; DRAFT permits a fully assessed review with warnings. The normal validation passed with four warnings for supported generic-binding rows, and the subsequent status dry-run reported COMPLETE as inconsistent with DRAFT. The earlier sentence in these notes that used COMPLETE simply to mean completion of assessment is superseded by this correction.
All 41 source assertions and their review decisions, three alternative_products, 15 references and the single core function are unchanged. The six UNDECIDED judgments remain explicit. Supported generic binding remains retained under the published ClinGen task instruction; no correct interaction is removed merely to suppress a warning. No ontology definition, source text, quotation, annotation qualifier or product assignment is changed. The prior approved merge remains a campaign event, while this separate status correction must complete its own review and publication before being described as published.