All 46 distinct seeded GO assertions and five alternative-product objects are preserved. The 49 raw GOA rows contain three pairs differing only in fields excluded by the normal seed projection; the fetched GOA and seed remain unmodified as provenance. The review records 13 ACCEPT, 15 KEEP_AS_NON_CORE, 8 MODIFY and 10 UNDECIDED, with two core functions and no NEW assertion. The unresolved cases retain their experimental source tuples rather than calling uninspected interactions false.
The core functions are direct mRNA 3′-UTR binding and translation activation through protein recruitment. The complete official and normal cached abstract of PMID:24954906(https://pubmed.ncbi.nlm.nih.gov/24954906/) establishes PAR-CLIP contact with U-rich elements and transcript stabilization; its full Methods were not available. Increased protein abundance alone is not treated as a measurement of translation efficiency. In PMID:32989052(https://pubmed.ncbi.nlm.nih.gov/32989052/), the official/normal abstracts, all seven displayed figure captions and selected original Methods/Results were read. Reporter activity was normalized to transcript abundance, with polysome profiling as an independent endpoint. Human ATXN2 recruits PABPC1 and TENT2/PAPD4; TENT2 performs poly(A) addition. Partial PAPD4 depletion, a reconstructed human full-length cDNA and the engineered 906–1095 fragment establish bounded mechanisms, not the function of every natural product. Figure pixels and complete supplements were not reviewed.
The complete cached Abstract, Methods and Results through the PABP-abundance experiments, plus selected Discussion paragraphs, were read for PMID:17392519(https://pubmed.ncbi.nlm.nih.gov/17392519/). Endogenous human ATXN2 associates with DDX6 in nuclease-treated lysates and redistributes to stress granules after heat or arsenite. Depletion reduces granule number/size without eliminating residual granules. Overexpression of full-length normal/expanded ATXN2 or its DDX6-binding region disperses visible P-bodies, while depletion lacks an obvious P-body effect. The existing assembly assertion is therefore refined to negative regulation of P-body assembly in that experimental context. This does not establish universal physiological suppression.
Later stress-granule evidence is kept assay-specific. The complete official and normal abstract of PMID:27240544(https://pubmed.ncbi.nlm.nih.gov/27240544/) distinguishes mouse fibroblast deficiency from human neuroblastoma starvation. The complete abstracts and selected original knockout/construct Methods and initial Results of PMID:37295429(https://pubmed.ncbi.nlm.nih.gov/37295429/) show that human ATXN2-knockout HeLa cells still form arsenite-induced granules with altered size and number; U2OS domain-deletion experiments are separate. PABPC is the RNA-dependent chaperone, and ATXN2 its client. Complete abstracts, selected Methods/Results and the Figure 7 caption of PMID:38869059(https://pubmed.ncbi.nlm.nih.gov/38869059/) support an approximately 20% reduction after human-cell ATXN2 depletion and stronger rescue by full-length than engineered 1–1079 protein. These observations do not require universal necessity for assembly.
All 21 seeded publication abstracts were read in the independent annotation consultation, along with selected original target-level passages. Sources with only accessible abstracts remain explicitly limited. PMID:10814712(https://pubmed.ncbi.nlm.nih.gov/10814712/) assigns the RNP motif to A2BP1/RBFOX1, not ATXN2. PMID:16835262(https://pubmed.ncbi.nlm.nih.gov/16835262/) distinguishes human and fly polyribosome association. In PMID:17097639(https://pubmed.ncbi.nlm.nih.gov/17097639/), ATXN2 is the Parkin substrate, not the ubiquitin ligase. The ATXN2L-focused PMID:23209657(https://pubmed.ncbi.nlm.nih.gov/23209657/) explicitly tests ATXN2/G3BP association in Figure 4B; its title is not a paralog-error diagnosis. PMID:26184334(https://pubmed.ncbi.nlm.nih.gov/26184334/) identifies ATXN2 in the DDX6 interactome, but a DDX6-bait experiment does not alone establish the source-specific ATXN2/PABPC1 pair. Other uninspected high-throughput pair assignments remain UNDECIDED.
The complete PMID:18602463(https://pubmed.ncbi.nlm.nih.gov/18602463/) abstract and accessed publisher introduction support endophilin SH3-domain interaction and direction-specific receptor downregulation, while distinguishing human-library, mouse-brain and rat-neuron experiments. The exact EGFR-binding assay remains unresolved. Later SH3 mapping is explicit corroboration for earlier interaction assertions, not an experiment attributed to the original 2005 paper. The PAM2–MLLE peptide structure in PMID:20181956(https://pubmed.ncbi.nlm.nih.gov/20181956/) supports a defined PABPC1 interface without testing every full-length isoform. RNA-dependent TDP-43 complex association is not represented as isolated binary binding.
The normal UniProt function, subunit, interaction and alternative-product blocks and complete VAR_SEQ block were read. Product 2 replaces residues 980–995 and lacks 996–1313; product 3 lacks 1–981 and has further changes; product 4 lacks 1244–1313; product 5 lacks 1–265 and 277–300 and changes 1106–1124. These sequence facts do not establish activity losses. In particular, the experimentally sufficient 906–1095 recruitment fragment is not intact in products 2 and 3. The fetched product-3 sequence note is truncated after three VSP identifiers and a comma; it is preserved exactly, while the complete variation record remains in the raw UniProt file. None of the 46 seeded assertions carries an isoform field; no artificial isoform specificity is added.
The PAINT IBA rows are retained where consistent with target experiments. A target's appearance in its own WITH/FROM list is legitimate descendant experimental grounding. The full tree/MSA was not reconstructed, and no unsupported ancestral-placement or circularity claim is made. The local GO-CAM index contained no ATXN2/Q99700 match at inspection; that is not proof of global model absence. No new biological process is proposed. Official ontology records were checked for GO:0003730, GO:0008494, GO:0017124 and GO:0010607 before using the authored core or replacement terms.
Eight missing seeded publications and five additional publications were imported through the standard-fetch recovery workflow with exact fetched bytes. Existing publications were preserved. The five additional records contain four full-text extractions (27240544, 37295429 and 38869059 XML; 32989052 HTML) and one abstract-only record (24954906). All literal YAML excerpts are exact cache substrings, each source contributing at most 25 words.
The prescribed provider launcher failed dependency resolution before a request. A separate invocation of the installed client failed Falcon authentication and Perplexity DNS and produced no report. Manual primary-source reading supplies the research record; no provider-named report was fabricated. Focused validation and rendering results will be appended after execution.
Focused just validate human ATXN2 passed (exit 0), including schema, term and reference checks. Eleven advisories remain: nine generic interaction rows retained as non-core under the supplied action definitions, source-specific ACCEPT versus MODIFY for broad RNA-binding assertions, and the translation-activator core lacking a separate NEW row. The latter is a synthesis of primary functional evidence, not a reason to manufacture a redundant annotation. Specific RNA-binding replacements use later direct corroboration while the HDA rows retain their original broad experimental precision. just render human ATXN2 also completed successfully. No repository-wide validation or remote approval is claimed.
The original 46 GO assertions and their source fields remain exact. One molecular-function assertion, GO:0008494 translation activator activity, is added from the human reporter and polysome experiments in PMID:32989052(https://pubmed.ncbi.nlm.nih.gov/32989052/). Normalizing reporter output to mRNA abundance distinguishes translation efficiency from transcript stabilization. ATXN2 recruits PABPC1 and TENT2/PAPD4; TENT2 supplies polymerase catalysis. The existing regulation-of-translation process does not make this distinct molecular activity redundant. No new biological process or activity for all natural splice products is inferred.
The translation core now includes the general ribonucleoprotein complex class. The complete PMID:16835262(https://pubmed.ncbi.nlm.nih.gov/16835262/) abstract reports human and fly polyribosome association; the fly domain mapping does not establish equivalent effects of every human product. A third core summary captures the already accepted stress-granule assembly and localization assertions. It describes a contextual contribution through PABPC1 interactions and self-association, without inventing a scaffold activity or implying universal indispensability. The 2007 depletion phenotype, 2023 HeLa knockout with retained but altered granules, and 2024 partial-depletion/rescue results remain distinct experimental contexts. Translation activation is not placed inside granules: the Discussion of PMID:38869059 explicitly presents granule polyadenylation as an untested possibility.
The prior unresolved RNA-transport judgment is superseded by a source-specific over-annotation judgment. The normal PMID:10814712(https://pubmed.ncbi.nlm.nih.gov/10814712/) cache remains abstract-only. Separately, the author-deposited original was read through complete Results, Discussion and Methods, with intervening captions. It does propose a possible neuronal transport role, but conditionally; the assays establish protein interaction, expression, fractionation and localization rather than movement of an RNA cargo. Thus the reviewer assertions that the normal cache is full text and that the original lacks any transport claim are both corrected. No figure pixels, raw data or supplements were reviewed. A publisher corrigendum surfaced during access concerns Figure 1 motif boxes, not a transport result. The function is not declared impossible; it remains an experimental question.
Broad RNA-binding HDA assertions retain the precision of their original assays; later direct target evidence supports separate IEA/NAS refinements. Uninspected ataxia-network pair assignments remain unresolved despite independent interaction mapping in other papers. Broad membrane HDA retains its source boundary, now with explicit independent membrane-association support from the complete PMID:18602463 abstract. Nine supported generic-binding NON_CORE decisions remain under the supplied ActionEnum policy; uninformative generality alone does not establish biological falsity.
The product-3 sequence-note truncation is corrected by executing only the independently tested candidate parser function against the immutable normal UniProt record. This restores the three missing VSP identifiers from its continuation line; the other four product objects and every other product-3 field are unchanged. The original fetched seed and raw UniProt are preserved. The separate parser change is PR #3567; no production parser code is changed by this gene follow-up. This supersedes the earlier note preserving the visibly truncated product-3 field, without asserting experimentally demonstrated isoform functions. All new literal quotations are exact normal-cache substrings, with cumulative quotations at most 25 words per source.
Focused validation passed after correcting only the newly authored stress-granule core label to the ontology label, stress granule assembly; source GOA labels were not changed. Ten advisories remain: nine supported generic-binding NON_CORE annotations under the supplied ActionEnum policy and the source-specific HDA versus IEA/NAS RNA-binding action difference explained above. Rendering passed. No global validation success or current-head approval is claimed.
The retained GO:0005515 decisions implement the user's explicit instruction for this ClinGen task: keep a supported, biologically correct interaction as KEEP_AS_NON_CORE when no defensible, more specific replacement has been established, rather than remove it solely for lacking functional information. This deliberately departs from the annotation-reviewer skill's informational-exclusion recommendation. That recommendation can remove an uninformative annotation without declaring the interaction false; the latest review is correct about that distinction. Earlier references here to the supplied ActionEnum are incomplete as an explanation of authority: the basis is the explicit instruction for this task, not the enum alone, a repository-wide policy change, an exemption granted by the agent, or an observed maintainer sign-off. No external approval comment is claimed.
The nine retained rows remain RBFOX1/A2BP1, three PABPC1 associations, Parkin, two DDX6 associations, TDP-43 and G3BP. The existing evidence boundaries are unchanged: Parkin acts on ATXN2 as its substrate, TDP-43 complex association is RNA-dependent, and DDX6 recovery is not uniformly a purified binary-binding measurement. The PABPC1 peptide/domain evidence does not test every natural ATXN2 isoform. Four endophilin rows retain their supported SH3-domain refinements, and all eight unresolved generic-binding rows remain UNDECIDED; evidence for an interaction in one paper does not verify an uninspected pair assignment in another.
All 47 review objects (46 original source assertions and one existing NEW translation-activator proposal), five alternative products, three core descriptions and the repaired Q99700-3 sequence note remain unchanged. The RNA-transport over-annotation decision continues to distinguish a published hypothesis from a measured cargo-movement endpoint.
This addendum clarifies the authority and exclusion criterion for the existing decisions. It adds no primary-source reading, biological assertion or new annotation adjudication. Earlier journal entries and validation results remain historical; this clarification does not claim reviewer approval or PR completion.