ATN1 is reviewed as a transcriptional coregulator with one nuclear corepressor molecular-function core. All 43 original source assertions are preserved: 9 ACCEPT, 17 KEEP_AS_NON_CORE, 15 UNDECIDED, 1 MODIFY and 1 MARK_AS_OVER_ANNOTATED. There are no NEW annotations or invented alternative products. Generic binding remains contextual where supported; unavailable exact pair evidence is kept uncertain.
Human ATN1 reporter repression and ETO association support the core. The GAL4 fusion supplies DNA binding; colocalized histone deacetylases supply their own catalytic activity. Neither activity is assigned intrinsically to ATN1. Nuclear-matrix association is experimentally contextual, not a claim that all ATN1 is permanently in an insoluble aggregate or a particular named N-CoR complex. Mouse reporter activation and rodent vascular migration/orientation remain correctly labeled contextual inferences. Species, reporter and cell-line differences are not proved causes of the different regulatory outputs.
The neuronal apoptosis annotation is over-specific for the cited non-neuronal assay. This does not deny the cytotoxicity result or exclude participation by a toxic fragment. The detailed external source boundaries are below. The CNS-development assertion and 14 interaction records remain uncertain. Positive curated interaction counts are acknowledged; lack of access to a particular pair table is not biological evidence against the pair.
IBA assertions agree with direct human evidence. The PAINT ancestral node and alignment were not replayed. Human ATN1 appearing among descendant evidence is legitimate, not circular. No new process, partner enzyme activity or family identifier is inferred from these associations.
The normal gene seed and all 14 originally cited publication caches are in place. Ten preexisting publication records were preserved. Falcon with its configured fallback failed without producing a report, so this is manual research rather than a provider-attributed report. Complete cached abstracts were read by the annotation consultant; targeted original Methods/Results and the core paper were inspected as specified below. Cached full-text availability is distinct from actual reading completeness. Figure pixels and complete interaction supplements were not inspected.
The additional mouse coactivation paper, PMID:17150957, is now available as an unchanged normal abstract-only cache. The recovered record confirms its title, DOI and abstract; it does not supply the previously read external Methods. All 15 PMID caches are present. All 43 annotation actions and the single core are unchanged by this source closure.
Source66 imported the unchanged normal PMID:17150957 abstract-only record with no overwrite. Its external assay-reading limits are recorded above. The six earlier literal findings were preserved and one short abstract finding was added; that source closure did not alter annotations, core function, products or source objects.
The one supported specificity change is the MAGI1/AIP3 interaction: its reviewed human UniProt entry explicitly states, “Interacts via its WW domains with DRPLA,” and attributes this to PMID:9647693. This agrees with the unchanged local MAGI1 UniProt record. GO:0050699 describes binding to a WW domain, with GO:0019904 protein domain specific binding as its parent. The replacement relies on this explicit curated domain attribution, not simply the abstract's statement that all five partners contain WW domains. Sources: https://www.uniprot.org/uniprotkb/Q96QZ7/entry; https://amigo.geneontology.org/amigo/term/GO:0050699.
The original PMID:9647693 abstract was read completely through official PubMed and the publisher. Publisher full-text/API access failed; the original ResearchGate page explicitly offers no full text. Original domain constructs and the pair-specific in-vitro confirmation table were not independently read. The other four AIP associations remain non-core with no invented interface specificity. Binding to a multidomain protein alone does not establish which of its domains mediates binding. Review articles were encountered as leads, but do not substitute for an original assay read in this assessment.
The LYST point concerns two different proteins. Original Tables 1 and 3 identify a LYST BEACH-domain-containing bait (U67615 cDNA nucleotides 9502–10590) paired with ATN1 prey (U23851 first nucleotide 2659, one colony). The precise ATN1 residues remain unresolved, but the partner bait is mapped. A controlled human-library yeast interaction supports the existing domain-binding annotation; neither a minimal interface nor purified ATN1–LYST binding is asserted. This original text was accessible externally despite the normal cache being abstract-only. The highlighted biochemical-confirmation examples concern other partners. Sources: https://www.researchgate.net/publication/11385760_The_Chediak-Higashi_Protein_Interacts_with_SNARE_Complex_and_Signal_Transduction_Proteins; https://link.springer.com/article/10.1007/BF03402003.
Six source assertions name WDR5/P61964, and the ATN1 UniProt record reports positive interaction evidence. Together these corroborate a reported association with a chromatin-regulatory partner. They are not automatically six independent experiments: reuse of interaction data, exact constructs and assay provenance were not resolved here. The individual source-specific rows therefore remain uncertain without implying biological noninteraction or assigning WDR5 activity to ATN1. The related ATN1–RERE record likewise has positive curated support, but membership in the same protein family does not identify the assay or construct behind a particular source row.
For PMIDs 28514442, 33961781 and 40205054, article text is available in the normal caches; exact case-sensitive ATN1/DRPLA searches did not find the target pair in those body texts. This is a bounded search result, not a claim that the original supplements lack the pair. The full_text_unavailable flags remain unchanged. References supplying contextual interaction records are rated MEDIUM rather than uniformly HIGH; the corepressor and directly informative functional sources retain HIGH relevance.
The activation/repression comparison does not establish a species-dependent switch. Selected original PMID:17150957 activation Results and Discussion were reread from the author-uploaded paper: mouse Atrophin-1 activates a GAL4 reporter, whereas the cited human repression assays use a different cell and promoter context. Several variables differ, so species is an experimental provenance limit, not a demonstrated cause. No new core or process is proposed. Source: https://www.researchgate.net/publication/6648376_Functional_Architecture_of_Atrophins.
Seven supported generic interaction rows remain non-core under the user-provided action definitions. The request to remove a biologically supported annotation solely because it is uninformative remains a policy disagreement; uncertainty or removal is not manufactured to silence that warning. All 43 machine source objects, the one core, the seven earlier literal cache findings and their availability flags are preserved. One exact curated MAGI1 UniProt finding and its file reference make the domain-specific premise reproducible; the source record itself is unchanged. No new annotation or alternative product is introduced.
The MAGI1 record is an explicit additional file reference for row 26 and a protected source dependency for publication. Its local bytes must be compared with the current-main source before publication; it is not an authored or regenerated gene file.