The GO:0005123 SFT prediction is UNC, rather than automatically NPI or CNN. The current review marks death receptor binding over-annotated because DISC association does not establish direct receptor contact. The accessible rat spinal-cord-injury abstract describes Fas association with FADD, caspase-8, cFLIP and caspase-3 in a DISC [PMID:17518537, “forming a death-inducing signaling complex (DISC)”]. This supports complex association but does not resolve the direct-contact question. PubMed and the publisher landing page were checked; the latter lists restricted full-text access, and the cached record has full_text_available: false. Failure to verify the full experiment cannot justify declaring the binding prediction refuted. The main gene review and experimental GOA annotation are unchanged.
The supported SFT wrapper now records an explicit gene/term exception for MARK_AS_OVER_ANNOTATED with a source-based UNC rationale. A different rejection action or accepted negation does not inherit this exception. This follows the PredictionAssessmentEnum uncertainty definition; it is not an assertion that direct binding is absent.
GOA changes. Six new rows were seeded as PENDING: cytoplasm IBA (donor list now adds human CASP2, UniProtKB:P42575, and MGI:MGI:1335092), cytoplasm ISO (located_in, human CASP3), cytoplasm ISS (human CASP3), caspase complex ISO (human CASP3 p17/p12 chains), regulation of protein localization ISO (human CASP3), and response to hydrogen peroxide IEA (ARBA). Retired rows: protein binding IPI with Xiap (PMID:18485100; the MODIFY to GO:0031625 ubiquitin protein ligase binding is kept, since XIAP is an E3 ligase), negative regulation of CDK activity ISO (mouse), and the older cytoplasm IBA row whose WITH/FROM set GOA no longer carries. That last one was retired here because the validator reported it as absent from GOA.
PENDING resolved. Cytoplasm IBA/ISO/ISS: ACCEPT [UniProtKB:P55213 "SUBCELLULAR LOCATION: Cytoplasm"; PMID:21630211 "cleaved caspase-3 and DNase-1 were found in the epithelial cytoplasms"]. Caspase complex ISO: ACCEPT, because active caspase-3 is a p17/p12 heterotetramer [UniProtKB:P55213 "Heterotetramer that consists of two anti-parallel arranged heterodimers"]. Regulation of protein localization ISO and response to H2O2 IEA: MARK_AS_OVER_ANNOTATED, matching their existing siblings.
Action changes.
- Enzyme activator activity (IBA): MODIFY -> MARK_AS_OVER_ANNOTATED. This edit was carried over from the interrupted session. GO:0008656 is an activator (binding) term, but caspase-3 activates procaspases by catalysis. The only seed for the IBD node is the rat PLA2-activator IMP, which comes from inhibitor epistasis PMID:12492473.
- Negative regulation of cytokine production (IEA, ISO): REMOVE -> KEEP_AS_NON_CORE. Caspase-3 proteolysis suppresses cytokine signalling [PMID:30878284 "activated caspase-3 cleaved cGAS, MAVS, and IRF3 to prevent cytokine overproduction"; PMID:36002459 "caspase-3 can mediate the cleavage of NF-κB members p65/RelA, RelB, and c-Rel via its protease activity"].
- Pyroptotic inflammatory response and its positive regulation (IEA, ISO): MARK_AS_OVER_ANNOTATED -> KEEP_AS_NON_CORE. The old reason cited only gasdermin D inactivation. Caspase-3 also activates GSDME [UniProtKB:P55213 "Also involved in pyroptosis by mediating cleavage and activation of gasdermin-E (GSDME)"].
- Cellular response to staurosporine (IEA, ISO), neurotrophin TRK receptor signaling, response to wounding, DNA damage response, response to UV (ISO): REMOVE -> MARK_AS_OVER_ANNOTATED. Caspase-3 does act in these contexts as the executioner, so the annotations are stimulus-context over-annotations rather than wrong. Their propagation_review blocks were changed to TERM_SCOPING_PROBLEM.
- Sensory perception of sound (ISO, mouse): REMOVE -> KEEP_AS_NON_CORE PMID:11374883.
- Heart development, T cell homeostasis, negative regulation of activated T cell proliferation (ISO, mouse): KEEP_AS_NON_CORE -> UNDECIDED. No cached source supports them, and the cached Casp3-knockout immunology paper covers only B cells PMID:12970760.
Support added. About 60 ACCEPT/KEEP_AS_NON_CORE rows had no supported_by. They now cite UniProt CC lines or cached papers: synapse [PMID:24478350, PMID:21151119], erythroid PMID:11208865, keratinocyte PMID:15068794, brain development/apoptosis PMID:8934524, B cells PMID:12970760, nuclear fragmentation PMID:11350920. Missing reasons were added to the same rows.
Open questions. Which mouse papers support the heart-development and T-cell rows (MGI:MGI:107739 ISO)? The cached keratinocyte literature conflicts: PMID:15068794 supports a role, while PMID:11175259 reports that epidermal differentiation does not involve the executioner caspases. That paper has no cached abstract, so it was not used.
Review follow-up 2026-10-17. In response to the PR #4625 review: the GO:0008047 enzyme activator activity IBA row and the GO:0016005 phospholipase A2 activator activity IMP row it inherits from stay MARK_AS_OVER_ANNOTATED, and their reasons were rewritten so they make an over-annotation argument rather than a categorical one. The seed is not refuted: the abstract places caspase-3 upstream of, and required for, PLA2 activation PMID:12492473. It rests on inhibitor epistasis, though, and the full text is not cached, so REMOVE was not used for the experimental row or for the IBA built on it. The two rows are judged together. Other changes: the description now names GSDME activation (with GSDMD inactivation) PMID:28459430; PMID:28459430 was added as primary support on the IEA pyroptotic inflammatory response row; core_functions gained in_complex GO:0008303 caspase complex (the ACCEPTed ISO row; the reviewer's "GO:0043293" is the apoptosome); and the stale "non-core" summaries on the three mouse-donor UNDECIDED rows (heart development, T cell homeostasis, negative regulation of activated T cell proliferation) were rewritten to match their actions.