Applied the review, annotation-reviewer and core-function-synthesizer workflows. The 37 machine-seeded annotations and 13 original reference identities were captured before editing in /tmp/ABHD12-baseline.yaml and /tmp/ABHD12-baseline-local.json. No source term, evidence code, reference, qualifier, supporting entity or alternative product was changed; no NEW annotation was added. Human ABHD12 is HGNC:15868, UniProt Q8N2K0; previous symbol C20orf22 and aliases ABHD12A, BEM46L2, DKFZP434P106 and dJ965G21.2 were checked against cached authoritative HGNC metadata. Canonical and five alias paths were absent from main, and six separate live open-PR queries were empty at main 71da2124e75b3e7673ee39a33e1ad1b1cfdc7945 (/tmp/ABHD12-current-preflight.json). The source3 seed/import provenance is retained in /tmp/ABHD12-root-baseline.json and tmp/source3-{canonical,auxiliary}-import-receipt.json.
The genuine default Falcon launch used a 1200-second limit and perplexity-lite fallback, concurrent with normal seed-publication caching. The first attempt could not resolve the pinned dependency from the offline tool cache; a justified online attempt failed DNS fetching the same dependency for both providers. Neither provider actually produced a report. Logs are /tmp/ABHD12-provider.log and /tmp/ABHD12-provider-online.log; these notes document the manual primary-source route, not a generated provider report. Normal seed caching found all four source PMIDs already available (/tmp/ABHD12-fetch-gene-pmids.log).
"PMC3466009" "Results" "HEK293". Human ABHD12 cDNA was expressed in HEK293 cells; glycerol-production assays establish MAG hydrolysis, with catalytic S246/D333/H372 mutant controls. These are human constructs, not mouse enzymes inferred from a human host. The tested DAG/TAG negatives bound the broad acylglycerol process to MAG catabolism. The early statement about lack of lysophospholipase activity is not a universal headgroup exclusion: the tested LPA/glycerol-readout context differs from later lysoPS fatty-acid assays.The two integrated cores describe the distinct lysoPS and oxidized-PS reactions. Their process nodes already occur in the seed, and the A2-type MF is the refinement of row 1, not an ungrounded added process. GO:0046475 is used for the lysoPS reaction and GO:0006660 for the distinct oxidized-PS reaction; these are related processes attached to different catalytic units, not duplicate new annotations. MAG hydrolysis remains supported but contextual. This includes a physiological possibility, not merely an in-vitro artifact.
"PMC3557017" "Table" "lyso". Mouse Abhd12 cDNA is expressed in HEK293T; knockout mouse brain demonstrates predominant lysoPS turnover and additional LPI hydrolysis. Bulk MAG/LPC/LPE hydrolysis was unchanged, which does not exclude recombinant capacity or local flux. The specific LPE panel result remains incompletely resolved here, so row 33 is UNDECIDED. The explicit LPI result supports row 37 as non-core."PMC4301979" "ABHD12" "lymphoblast" "extracellular". PHARC-subject lymphoblasts have reduced ABHD12/lysoPS-lipase activity and altered secreted lysoPS; mouse macrophage experiments establish the opposing ABHD16A-generation/ABHD12-degradation axis. Protease-protection and glycosylation comparisons support lumenal/extracellular orientation, not a soluble cytosolic catalytic domain. This does not by itself distinguish ER lumen from every possible cell-surface pool."PMC7611549" "HEK293T" "human ABHD12" searches. Human enzyme in HEK293T membranes prefers very-long-chain synthetic R-Me-lysoPS substrates. These are methyl esters; carboxylesterase conversion makes them useful surrogates/prodrugs, but they should not be silently described as unmodified natural substrates. Mouse knockout tissue and mast-cell experiments, and human THP1 lipid treatments, are separate scopes.The MGI comparative source graph links mouse Q99LR1 to PMID:27307232 for protein depalmitoylation, cytoplasm and dendrite cytoplasm, and to PMID:22632720 for AMPA-complex membership.
PMID:27307232, PMC5015780 was recovered as indexed original Methods/Results with the title and ABHD12, and "PMC5015780" "Figure 3" "ABHD12". Mouse Abhd12 NM_024465 is explicitly in the cDNA screen. Figure 1 measures reduced PSD-95 palmitate labeling after ABHD12 coexpression in HEK293T, with a smaller effect than ABHD17. Figure 3A compares FLAG-tagged proteins in rat hippocampal neurons. The detailed neuronal catalytic-mutant, endogenous-substrate and knockdown work concerns ABHD17; it should not be borrowed for ABHD12. Thus the protein-hydrolase and dendritic assertions are retained as bounded secondary annotations, while the broad cytoplasm term remains true for the established intracellular membrane pool. Cytoplasm is not a synonym for soluble cytosol. No source failure is inferred from the ABHD17-centered title.
PMID:22632720(https://pubmed.ncbi.nlm.nih.gov/22632720/) was subsequently recovered as the original institutional PDF, also available from the Fakler laboratory. Table 1 (journal page 624) explicitly identifies the mouse protein accession Q99LR1/ABHD12. Results page 622 describe total-brain membrane preparations from adult rats, wild-type mice and antibody-target-knockout mice, with multiple anti-GluA antibodies and IgG controls. Figure 1B distinguishes rat-membrane antibody/IgG comparisons from mouse wild-type/knockout comparisons. The protein identifier does not establish the species of every preparation. The donor evidence now resolves row 17 as non-core complex membership. No human receptor-gating function is inferred.
Live GO:0120560 specifies hydrolysis of 1-acyl-sn-glycero-3-phospho-L-serine; GO:0004622 instead specifies phosphocholine. GO:0004623 is A2-type glycerophospholipase activity, cleaving the sn-2 acyl ester. The GO:0046475 definition includes glycerophosphate derivatives with at least one acyl/alkyl/alkenyl group, so lysoPS qualifies. GO:0008474 is a protein-palmitate hydrolysis activity, distinct from lipid ester hydrolysis.
The seeded GO:0098734 label is obsolete. AmiGO still displayed its August 6 snapshot, but the later official obsoletion notice 1174 and closed ontology issue 32290 explain that this is molecular function, not a separate BP. The source term is preserved and the inferred BP row removed; the underlying GO:0008474 is retained separately. Direct QuickGO API access failed; that failure is not claimed as an ontology result.
All PAINT rows preserve PTN000957701 and record only that node in the structured propagation assessment. The target's appearance among descendants is legitimate self-evidence, not circular. The imported PTHR12277 inventory places human Q8N2K0 and mouse Q99LR1 in SF61 and ABHD12B in SF69; no source node tree/MSA was reconstructed or invented from this inventory. No Q8N2K0/ABHD12 hit was found in the local GO-CAM index. That absence was not used to propose a process. Existing broad membrane/ER assertions retain their original evidence resolution.
Cached R-HSA-5694462 and R-HSA-426048 describe the MAG-to-arachidonate reaction and its parent pathway. The live ABHD12 entity R-HSA-5694474 supplies the contextual plasma-membrane location. Its compartment is retained as curated pathway context; the more direct ER evidence is the core location. No new Reactome record was authored or substituted.
The 13 cited PMIDs are the four cached seed sources plus nine additional sources listed above. All nine normal fetches failed with DNS errors, with terminal exit 1 and no records created: /tmp/ABHD12-extra-fetch.log, /tmp/ABHD12-donor-fetch.log, /tmp/ABHD12-current-fetch.log. An initial mistyped identifier in the first log also failed and was not used as evidence or retained in the citation inventory. Required missing PMIDs are 20797687, 22632720, 23297193, 25580854, 27307232, 30420694, 33571455, 40496808 and 41983263. Both cited Reactome records are cached. External verification is recorded separately from local full-text availability and does not close these cache gates. All provider and machine-source files remain unedited.
The review is DRAFT pending normal cache recovery. Final decisions are 15 ACCEPT, 16 KEEP_AS_NON_CORE, four MODIFY, one REMOVE and one UNDECIDED; there are zero NEW annotation rows and two catalytic cores. Immutable UniProt bibliography that was not used to support an authored assertion was not recursively treated as a requirement to review every paper. Full validation passed with one warning category covering the nine missing PMID records. Independent schema validation passed; all 37 source objects, 13 original reference identities and three alternative products are preserved, and all 39 ordinary supporting quotations match cached text case-sensitively after whitespace normalization. The coordinator read all decisions and both cores; the subsequent source-resolved AMPA membership refinement was reported separately for final independent review. A bounded peer consultation found no material concern but did not independently inspect the AMPA source or neuronal Figure 3A image, and is not represented as a full-source signoff. Exact final hashes and provenance are recorded in /tmp/ABHD12-local-manifest.json.
The expected published head is dec9a93c76c1f0f4ce91a3e3ed40b8ff69b54b99; the five canonical files matched the coordinator's exact-head baseline before editing. Formal review 5329661187 and full issue comment 5854462385 were read. The review/annotation-reviewer and PR-review workflows apply. This follow-up preserves all 37 source assertions, their evidence and qualifiers, all 22 existing reference identities, both catalytic cores, three alternative products, and the original history. One action changes: the human inferred protein-palmitate hydrolase moves from NON_CORE to MARK_AS_OVER_ANNOTATED. Totals are 15 ACCEPT, 15 KEEP_AS_NON_CORE, four MODIFY, one MARK_AS_OVER_ANNOTATED, one REMOVE and one UNDECIDED, with zero NEW rows or new core terms.
The mouse PSD-95 screen in PMID:27307232 is real, but its cellular labeling effect does not isolate direct ABHD12 protein-thioester chemistry. Public primary PMC5015780 Methods/Results were re-read directly and through the indexed query site:pmc.ncbi.nlm.nih.gov/articles/PMC5015780 "ABHD12" "80%". The topology controls in newly cached full PMID:25580854 create a concrete substrate-access limitation; this is not proof that the screen was an artifact. The separately retained dendrite-cytoplasm annotation does not claim a spine or plasma-membrane pool. The propagated source assessment records weak mechanistic inference rather than a false donor identifier. Live GO:0008474 requires palmitoyl-protein hydrolysis; it is not merely any change in protein palmitate labeling.
The weak MCF7 transition sentence was replaced on six cytoplasmic/ER/membrane rows by the exact cached result naming predominant ER-membrane localization. The plasma-membrane row instead cites the positive official Reactome entity, leaving the ER result as a limitation in its reason. PAINT and combined-mapping reasons now identify their actual source type. Broad membrane and cytoplasm assertions remain ACCEPT because they describe the compartment of the core enzyme at the source resolution; the live cytoplasm definition includes organelles. Broad phospholipid catabolism likewise remains a true description of the directly catalyzed core reaction. Omitting an ancestor from a concise integrated core does not turn it into a different non-core function. The existing MODIFY replacements converge on terms already present; no added coverage is claimed. The duplicate bare supporting reference was removed.
Inflammatory regulation now carries exact cytokine and infection evidence. PMID:25580854 Figure 4 uses mouse peritoneal macrophages, while its PHARC lymphoblast experiments supply human lipid evidence. Full PMID:30420694 includes DO264 versus inactive (S)-DO271 controls, recombinant human/mouse inhibition, primary-human-macrophage lipid measurements and mouse LCMV inflammatory outcomes. These scopes support the contextual process without universalizing immune direction. The 2026 PMID:41983263 abstract supplies the explicit 2-AG increase/arachidonate decrease result; the full species/construct protocol remains uninspected.
tmp/source8-canonical-import-receipt.json binds the nine exact normal publication records. Newly recovered full bodies are 25580854, 30420694, 33571455 and 40496808; 20797687, 22632720, 23297193, 27307232 and 41983263 remain abstract-only locally. Reference availability flags follow actual cache metadata, while prior external full-source access is described separately. Cached 33571455 Table 1 retains the synthetic methyl-ester versus natural-lysoPS distinction. The published 40496808 article is distinct from its preprint: its mouse sarmopathy/neuronal experiments support neuroimmune context, and its Discussion explicitly says oxidized-PS levels were not measured. No new human axon-maintenance assertion is inferred.
All 13 cited PMIDs now have normal caches. There are no provider reports, raw provider HTML/PDF artifacts or DOI-only research citations in this gene directory. Existing institutional PDF links resolve to the already cited PMID:22632720; publication metadata maps the other DOI/PMC identifiers to the same 13 sources. Unused UniProt bibliography is not treated as an authored source dependency.
The notes-inclusive census identified one additional pre-existing source dependency: Reactome R-HSA-5694474, ABHD12 physical entity, used for plasma-membrane placement. Its official page was rechecked for stable ID, human species/Q8N2K0 and Compartment; this is a curated physical-entity assertion, not experimental microscopy. It is now an explicit reference, bringing the list to 23. The initial guessed CLI command fetch-reactome was rejected as nonexistent (exit 2; /tmp/ABHD12-entity-fetch.log), so no retrieval occurred. The established cache_reactome_pathway wrapper was then invoked once with force=False and failed DNS resolving reactome.org, exit 1, no file (/tmp/ABHD12-entity-normal-fetch.log). It has been handed to the source16 owner; no provider or cache was manufactured. R-HSA-426048 and R-HSA-5694462 remain cached. R-HSA-5694474 is the sole remaining required cache and the PR remains draft. A future normal record may contain only identity/display-name metadata; external compartment verification must remain explicitly distinguished from cache contents.
The normal Reactome cache for R-HSA-5694474 is now present. Its exact 184 bytes
contain the stable identifier and display name, ABHD12 [plasma membrane];
there is no experimental body or explicit compartment field in that cache.
The official entity page
was independently rechecked for human ABHD12, UniProt Q8N2K0 and its
plasma-membrane compartment. This remains a curated location assertion,
separate from direct ER-localization experiments. The original cached reaction
R-HSA-5694462 is listed alongside the entity as pathway context; the entity
page supplies the compartment claim. None of this establishes a new localization
experiment or changes the KEEP_AS_NON_CORE judgment.
The current-head review of PR #3286, comment5855488055, was read in full.
Its topology suggestion is recorded as COMPARTMENT_OR_COMPLEX_MISMATCH on the
existing protein-depalmitoylase propagation assessment. The positive mouse
screen and the existing over-annotation decision are preserved; the added
failure mode records the already stated substrate-access limitation.
All 37 source assertions/actions, both complete cores, 23 reference identities
and three alternative products are unchanged. The 13-PMID/three-Reactome census
has no remaining missing record. The source16 run36313594604 at
408e7c41d93c2fd63f54ac1da5d7201edbc901bb supplied the exact normal bytes,
artifact10931450272 SHA2564251efe477f5575fe1f7f86a5b543f3188e81a87fd25fc88e65c47a2af862c01.
This supersedes the earlier transient retrieval gate. No source cache or
published history was rewritten. Source closure and warning-free validation
permit COMPLETE status; the unresolved annotation remains explicit.