ADTRP (Q96IZ2) — propagation and route analysis

Generated by analyze_adtrp_propagation.py. generated: 2026-07-27

1. The worklist's 'no-IBA' name is stale

projects/paint/human-no-IBA-simple.csv lists ADTRP, but GOA carries 3 IBA rows: GO:0005901, GO:0016787, GO:0042758.
GOA TSV has 27 data lines (27 distinct) over 19 distinct GO terms; QuickGO returns 27 annotations.

2. PANTHER node reach — the MF/BP granularity asymmetry

Cached PAINT table (interpro/panther/PTHR10989/PTHR10989-paint.tsv), which agrees with
the live QuickGO WITH/FROM:

The finding. At one and the same node PTN001659973, scoped by PAINT to taxon:2759 Eukaryota and seeded by exactly two human proteins (ADTRP and AIG1), PAINT asserts the root of the hydrolase branch as the molecular function (GO:0016787) and a four-step-deep biological process (GO:0042758 long-chain fatty acid catabolic process). Both reach all 86 recipients, of which 14 are fungi, 5 are plants and 2 are other eukaryotes.

Only 7 of the 86 recipients are Swiss-Prot reviewed, and two of those are uncharacterised UPF0641 fungal proteins. The MF call is dated 20260528, i.e. after GO:0120573 FAHFA hydrolase activity was created (2026-03-14), so the general term is a deliberate judgement rather than a stale-term artefact.

What the conserved character actually licenses (this corrects a first-pass error)

My first pass argued that the clade's heterogeneity justified the general MF and therefore made the specific BP unwarranted. Measurement does not support that reasoning, and the corrected version is sharper. Aligning all 85 other recipients to Q96IZ2 Thr47/His131 (UniProt SITE features, ECO:0000269|PubMed:27018888) and requiring the aligned column to land on ADTRP's own annotated SITE position:

Positive control: AIG1 (Q9NVV5), whose own catalytic residues are independently annotated as Thr43/His134, scores dyad-intact at 36.5% identity, so the aligner recovers a known case.

So the catalytic dyad is broadly conserved: 73 of 85 recipients retain it, including every vertebrate and both Dictyostelium members. Note the fungal column: all 14 fungal recipients fall below 25% identity, where a pairwise alignment will manufacture residue matches out of noise, so their dyad status is undetermined, not negative — an absence of evidence, which is not evidence of absence.

The corrected finding, and it is a category distinction rather than a sloppiness claim. A conserved catalytic dyad licenses a mechanism inference but not a substrate inference. GO:0016787 hydrolase activity states mechanism only, and is therefore exactly scoped to what the conserved residues support family-wide - so the general MF is well founded, and better founded than the heterogeneity argument I first made. GO:0042758 long-chain fatty acid catabolic process is a substrate-level claim, and the substrate is known only for the four characterised animal members (human and mouse ADTRP and AIG1). That is the asymmetry worth reporting: not that one term is too specific and the other too general, but that the node propagates a substrate claim on evidence that can only support a mechanism claim.

Reciprocal question, benign answer. PTN002591065 — the node whose reach is almost exactly this gene's orthologue set — covers 25 gene products, all Vertebrata, scoped taxon:117571 Euteleostomi, and gives them GO:0005901 caveola. Caveolae are a vertebrate structure, so the scope and the term agree. The paralog AIG1 is correctly not among the recipients.

PTHR12242 vs PTHR10989 — distinct families, not a renumbering

PMID:27018888 names PANTHER family PTHR12242 for the non-mammalian AIG1/ADTRP-like proteins, whereas this analysis works from PTHR10989. Both are live and distinct, so the paper describes a sister set in another family rather than the recipients of the node reviewed here:

3. The three non-PAINT routes

InterPro2GO — non-confirmation. IPR006838 (ADTRP/AIG1, family) is a family-specific signature, not a bare fold. interpro2go maps it to GO:0016020 — a cellular component only, no molecular function. Of its 5788 proteins, 8 are reviewed (Swiss-Prot) (0.138% of the family), and 2 of those are uncharacterised UPF0641 proteins. The predicted fold-to-activity error is absent: this entry shows restraint.

• P38842 UPF0641 membrane protein YHR140W — Saccharomyces cerevisiae (strain ATCC 204508 / S288c)
• Q5M828 Androgen-dependent TFPI-regulating protein — Rattus norvegicus
• Q60534 Androgen-dependent TFPI-regulating protein — Mesocricetus auratus
• Q8C138 Androgen-dependent TFPI-regulating protein — Mus musculus
• Q96IZ2 Androgen-dependent TFPI-regulating protein — Homo sapiens
• Q96WV4 UPF0641 membrane protein PJ4664.05 — Schizosaccharomyces pombe (strain 972 / ATCC 24843)
• Q9D8B1 Androgen-induced gene 1 protein — Mus musculus
• Q9NVV5 Androgen-induced gene 1 protein — Homo sapiens

Bulk classification imports carrying TAS — non-confirmation. ADTRP has no TAS row.
Reference-projection test on every supporting reference, fully paginated:

Identical term sets are a necessary but not sufficient condition for a projection. PMID:27018888 gives ADTRP and AIG1 the same four terms, but the paper individually mutated and individually assayed both proteins (T47A/H131A for ADTRP, T43A/H134A for AIG1), so this is parallel per-protein curation of two characterised enzymes, not one finding projected onto a set. The ACTR8-style defect — a complex-level or single-gene phenotype spreading unchanged across every member — is absent here: no reference on this gene annotates more than four entities, and the two multi-entity references give each entity a different term set.

ARBA rules — non-confirmation. No WITH/FROM on any ADTRP row names an ARBA… rule;
the two automatic routes present are GO_REF:0000044 (Swiss-Prot subcellular-location
mapping) and GO_REF:0000116 (Rhea mapping), both of which name their source explicitly.

4. Reaction direction and substrate identity

GO:0120573 carries 12 RHEA cross-references and its definition states the hydrolytic direction explicitly: Catalysis of the hydrolysis of the ester bond in a fatty acid ester of a hydroxy fatty acid (FAHFA), yielding a free fatty acid and a hydroxy fatty acid. FAHFAs are a class of endogenous bioactive signaling lipids in which a fatty acid is esterified to a hydroxyl group on a second fatty acid backbone.

ChEBI classification of the reaction participants:

So the substrate itself is a long-chain fatty acid, and the annotated direction is hydrolysis (removal), matching UniProt's PhysiologicalDirection=left-to-right on all 12 catalytic-activity lines. The predicted direction/substrate inversion is not present: GO:0042758 names the right chemistry for this gene.

5. Ancestry relations the review depends on (all fetched, none inferred)

4b. The two GO:0005515 rows are ONE screen counted three ways

IntAct returns 10 interactions for Q96IZ2. The two rows GOA carries both come from PMID:32296183 (HuRI), and each partner is logged under three sub-methods of the same screen:

All are yeast two-hybrid, MI-score 0.56, and this is what UniProt's NbExp=3 is counting — not three independent experiments. There is no orthogonal assay for either partner and no follow-up anywhere in the ADTRP literature.

5b. Cell-type branch check on GO:2000402 (sibling, not ancestor)

PMID:28341552 is annotated to GO:2000402 negative regulation of lymphocyte migration (IMP, BHF-UCL). Cell-type words in the cached abstract: monocyte 4, lymphocyte 0, leukocyte 0 (full text available: False).

Neither closure contains the other, so the annotated term does not merely generalise the data — it names a different leukocyte lineage. Monocytes are myeloid mononuclear phagocytes; lymphocytes are lymphoid. The abstract's phrase is "transendothelial migration of monocytes", which is precisely GO:2000438 negative regulation of monocyte extravasation. GO:0071676 negative regulation of mononuclear cell migration is a verified ancestor of both and is the safe fallback if the full text (not cached) does contain a lymphocyte assay. The leukocyte-level claim is in any case already annotated separately from the same paper as GO:0002686.

6. Logical-opposite citation cross-product

8 positive/negative regulation terms; opposed pairs on the same base process: 0. Verdict: NEGATIVE (no positive/negative pair on one base process).

7. Does PMID:27018888 support GO:0005886 plasma membrane?

Occurrence counts in the cached full text (full_text_available: true):

Conclusion: no plasma-membrane or cell-surface localisation experiment in the cached full text; compartment evidence is recovery in the membrane fraction of transfected HEK293T cells plus six topology predictors (supplementary figures are not in the cache).

8. Paralog AIG1 (Q9NVV5)

Q9NVV5 AIG1_HUMAN, reviewed (Swiss-Prot), catalytic sites [(43, 'Important for catalytic activity'), (134, 'Important for catalytic activity')]. It is a genuine co-seed, not a spurious donor: it carries its own experimental annotations to GO:0005515, GO:0005886, GO:0016020, GO:0042758, GO:0120573 and shares the identical IBA rows GO:0016787, GO:0042758 with ADTRP from the same node.