AADACL4

Summary: Correct, correctly scoped and validly transferred - and redundant. The donor node is functionally heterogeneous, and donor-by-donor testing shows bare hydrolase activity is not merely a cautious call there but the exact least common ancestor of the donor set, with no candidate term below it surviving every donor. What makes the row replaceable is therefore not granularity but redundancy: AADACL4's own subfamily signature already supplies the specific ester-hydrolase term on this same record.

Reason: All three human AADAC-family paralogs carry this row byte for byte: AADACL2, AADACL3 and AADACL4 each have GO:0016787 by IBA from GO_REF:0000033, transferred from PANTHER node PTN009058710, with the same 17 WITH/FROM tokens. That equality is measured for all three, not assumed, from two independent sources that have to agree: the committed GOA TSVs, and a per-accession QuickGO fetch for Q6P093, Q5VUY0 and Q5VUY2. It therefore has one correct treatment, and the donor set decides what that is. All 17 tokens were resolved and each donor's chemistry and nucleophile read off its own EC numbers, its own curated GO annotations classified by fetched ontology ancestry, and its own ACT_SITE features. GO:0016787 is true of 16 of the 16 donor proteins. GO:0052689 carboxylic ester hydrolase activity is true of 14 and false of 2: the kynurenine formamidases mouse Afmid (Q8K4H1) and yeast BNA7 (Q04066), EC 3.5.1.9, carry GO:0004061 arylformamidase activity, which sits under GO:0016810 on the carbon-nitrogen branch and not under the ester branch, by IMP and IDA respectively. GO:0017171 serine hydrolase activity is true of 15 and false of 1: soybean HIDH's nucleophile-elbow residue is Thr164. That last call rests on fold position rather than on a UniProt label - UniProt describes HIDH's ACT_SITE 164 as a proton acceptor (ECO:0000305) rather than naming a nucleophile, and 164 is the N-terminal-most member of a checked Thr-Asp-His triad - but the sequence corroborates it independently: 15 of the 16 donors read G-x-S-x-G at the nucleophile elbow while HIDH alone reads GETSG, so the elbow serine is substituted. The verdict does not depend on that rule: downgrading HIDH to undetermined still leaves GO:0017171 untrue of every donor, and downgrading every positionally-inferred nucleophile leaves it supportable for only the two donors UniProt labels explicitly. The two refutations lie on different axes, so neither candidate can be rescued by choosing the other, and because GO:0016788 and GO:0016810 are siblings whose only common ancestor below catalytic activity is GO:0016787, the term PAINT chose is exactly the lowest common ancestor of its donors. There is consequently no granularity and no term-scoping defect to correct on this row in any of the three genes, which is why root_cause is EVIDENCE_CIRCULAR_OR_REDUNDANT and no failure mode is recorded: the target already has stronger direct evidence, and the propagation itself did not fail. GRANULARITY_MISMATCH is deliberately not recorded: read literally its definition would fit any general term on any record, but the slot describes the biological shape of a propagation issue and this propagation has none, so the campaign rule applies - ask whether the term is the LCA of its donors, and reserve the mode for donors that agree while the term still sits above them. The schema's enum description does not yet distinguish the two readings and would be worth clarifying: all three paralogs now encode the row on the propagation-shape reading, which the enum text does not itself state. For AADACL4 that stronger evidence is the narrower arylacetamide-deacetylase signature (InterPro IPR017157, PIRSF037251), whose GO:0052689 is already on this record, backed by a catalytic apparatus that is demonstrably intact: the annotated Ser193, Asp347 and His377 really are Ser, Asp and His, the nucleophile sits in a canonical GESVG elbow, the HGG oxyanion-hole motif is present at 119-121, and all seven AADAC-family relatives project their own annotated triads onto exactly those three residues. So this row adds only a redundant ancestor of a term the record already has, and the replacement term is a statement about what AADACL4's annotation set should contain rather than a claim that PTN009058710 could itself license the child term. Neither GO:0016787 nor GO:0016020 is in gocheck_do_not_annotate (both are usage Unrestricted), so this rests on redundancy, not on a rule violation. GO:0017171 is separately true of AADACL4 on its own residues but cannot be reached from this row either; it is true of every donor PAINT cites at the family node PTN009058713 and IDA-supported by all three of them, so it is raised as a PAINT node-placement recommendation in suggested_questions.

Proposed replacements: carboxylic ester hydrolase activity

Summary: Accept, and accept it at this general level. The donors are the AADAC/NCEH1 clade, all single-pass type II membrane proteins, but they disagree about which membrane, so membrane is the correct call rather than a lapse into an over-general parent.

Reason: This is the narrower of the two PANTHER nodes used for AADACL4: PTN009058713 is supported by mouse Aadac (Q99PG0), mouse Nceh1 (Q8BLF1) and human AADAC (P22760), all of which are UniProt single-pass type II membrane proteins detected at protein level, and AADACL4 carries an equivalent predicted N-terminal signal anchor (its own TRANSMEM 5-25) with its catalytic domain and both predicted N-glycosylation sites on the far side. The is_active_in qualifier is appropriate for a membrane-anchored enzyme. The term should not be refined: the donors resolve to four distinct specific locations - endoplasmic reticulum membrane and microsome membrane for the two Aadac orthologues, cell membrane and microsome for Nceh1 - so choosing endoplasmic reticulum membrane or plasma membrane for AADACL4 would mean picking one donor over the other with no AADACL4 data to justify it.

Summary: The same conclusion as the IBA row, reached from AADACL4's own predicted topology rather than from donor localisation. Kept, with the caveat that its basis is the softer of the two and that the closest paralogs do not share the feature call.

Reason: This row traces to the UniProt subcellular-location line, which is an ECO:0000305 curator inference over an ECO:0000255 TRANSMEM prediction; SL-0162, ARBA00028763 and IPR017157 are the controlled-vocabulary, automatic-annotation and family handles for that same statement. The conclusion agrees with the independently derived IBA row and with the family, so it is retained. What is worth recording is how soft the underlying feature call is: among paralogs that share the catalytic register exactly, UniProt calls a signal anchor for AADACL4, a cleaved signal peptide for AADACL2 (which is consequently annotated Secreted), and nothing at all for AADACL3 despite a comparably hydrophobic N-terminal segment. All of those are sequence-analysis predictions, and mean hydropathy cannot discriminate a cleaved signal peptide from an uncleaved type II anchor, so no attempt was made to overturn the call from sequence alone. The two membrane rows are only partly independent: the IBA leans on experimentally localised donors, whereas this row leans on AADACL4's own topology prediction plus family similarity. So they are not simply duplicates, but neither contributes experimental evidence for AADACL4 itself, and this is the weaker of the two.

Summary: Redundant general parent, and correctly scoped for its own signature. IPR013094 is the generic Abhydrolase_3 domain, which can only be mapped to hydrolase activity, while the same protein also matches the subfamily-specific IPR017157 that yields the informative term.

Reason: InterPro2GO maps IPR013094 (alpha/beta hydrolase fold-3 domain) to GO:0016787 because that domain occurs in enzymes with different bond specificities; the mapping is correct for the domain and redundant for the protein. AADACL4 also matches IPR017157 (arylacetamide deacetylase) and PIRSF037251, which support the ester-bond specialisation, and its catalytic triad and oxyanion hole are intact, so GO:0052689 - already present on this protein from IPR017157 - is the term the evidence supports. As with the IBA hydrolase row, this is a granularity replacement, not a claim that the annotation is wrong - and specifically not a recommendation to change the IPR013094-to-GO:0016787 InterPro2GO mapping itself. That domain genuinely does occur in amidases as well as esterases, and in soybean HIDH, whose nucleophile is a threonine and whose physiologically important reaction is a dehydration, so the mapping is right for the domain and should stay. The replacement is a statement about what AADACL4's own annotation set should contain, where a subfamily-level signature is available too, and the classification matches the IBA row above: EVIDENCE_CIRCULAR_OR_REDUNDANT, with no failure mode, because the signature's own scope is correct and only its redundancy on this record is at issue: the fold genuinely spans amidases, esterases and a threonine-nucleophile dehydratase, so GO:0016787 is the right mapping for it. GRANULARITY_MISMATCH is deliberately not recorded here for the same reason as on the IBA row - the mode describes the biological shape of a propagation issue, and a signature whose own member set is heterogeneous has no granularity defect to record.

Proposed replacements: carboxylic ester hydrolase activity

Summary: The best-supported statement about AADACL4 and the one molecular-function term worth keeping. The subfamily signature is backed by a demonstrably complete GDXG catalytic apparatus, though no substrate has ever been tested.

Reason: IPR017157 is the arylacetamide-deacetylase family signature rather than a generic fold, and the activity it implies was checked against the sequence rather than taken on trust. AADACL4's UniProt active-site positions 193, 347 and 377 are genuinely Ser, Asp and His; Ser193 lies in a canonical GESVG nucleophile elbow; the HGG oxyanion-hole motif is present at 119-121; and the annotated triads of all seven AADAC-family relatives - AADACL3, AADACL2, human, rat and mouse AADAC, human and mouse NCEH1 - project onto exactly those three AADACL4 residues by global alignment with matching residue identity. This rules out the fold-retaining pseudoenzyme reading, so the annotation is a statement about an intact active site rather than a domain name transcribed into an activity. It remains an unverified homology inference: no substrate, inhibitor or activity assay has been reported, which is why the term is kept at this substrate-agnostic level and no child term is proposed. It is also the term the two general hydrolase-activity rows on this record should collapse onto, and the core molecular function recorded for AADACL2 and AADACL3 as well, whose catalytic registers are identical to AADACL4's.