View original ARBA rule on UniProt
Assigns GO:0006720 "isoprenoid metabolic process" to proteins matching any of 94 alternative condition sets built from InterPro entries, CATH FunFams, PANTHER families and taxon constraints. The GO term is in the right branch for most of the rule: 64 of the 94 sets identify genuine isoprenoid enzymes - terpene synthases and cyclases, the MVA and MEP precursor pathways, prenyl diphosphate synthases, carotenoid backbone and cleavage enzymes, gibberellin and ABA oxidases, named retinoid enzymes and JHAMT. Thirteen sets should be removed - twelve off-target (LDLR/LRP1, CYP2C9, CYP83B1, ADH5, ADH7, hormone-sensitive lipase, PNPLA2, phenylalanine aminomutase, bare UGT and Rossmann FunFams) plus one ALDH FunFam conjunction that appears unsatisfiable - one set is mixed and needs splitting, and sixteen cannot be audited from their identifiers. The rule's most serious defect is not biological: the 8,974 annotations it currently emits cannot be reproduced from its published condition sets.
Condition-set counts describe the sets recorded in this review, which may omit the full rule.
ARBA00027853 is a 94-branch omnibus rule whose GO term is correct and whose biological core is sound, but which is not auditable as published. GO:0006720 is a genuinely appropriate broad parent for the families the rule mostly targets: its definition covers compounds "containing or derived from linked isoprene residues", and QuickGO ancestor closures checked on 2026-08-22 confirm that retinoid metabolic process (GO:0001523), terpenoid metabolic process (GO:0006721), gibberellin biosynthetic process (GO:0009686) and the tetraterpenoid/diterpenoid biosynthetic terms are all descendants of it. That definition also answers the objection recorded in geneontology/go-annotation#5835 that a metabolite "is derived from an isoprenoid, but not an isoprenoid" - derived-from is explicitly in scope. The boundary that does bite is steroids: GO:0008202 steroid metabolic process is NOT a descendant of GO:0006720 (its QuickGO ancestor closure is GO:0008150, GO:0008152, GO:0006629, GO:0008202, GO:0044238, GO:0009987), which makes two branches wrong-branch rather than merely broad. Note also that the specific protein cited in #5835, S. pombe SPAC31G5.16c = dpm1 (O14466), is not reachable from any of this rule's condition sets and carries no GO:0006720 today; that part of the issue belongs to ARBA00028538 and ARBA00028655. This puts ARBA00027853 in a different position from its sibling rules in #5835: the GO term itself is in the right branch here, so the rule's problems are ones of branch construction and reproducibility rather than of a mis-chosen term. Branch triage gives 64 of 94 condition sets on-target, 13 to remove, 16 unresolvable from their identifiers, and one (CS39) mixed and needing a split. The empirical picture is better than the branch count suggests. The UniProt API reports zero annotated proteins for this rule - the figure the commissioned deep research was given, and on which it built its "no present production set" reasoning - but a QuickGO census on 2026-08-22 finds 8,974 live GO:0006720 / ECO:0000256 / GO_REF:0000117 annotations citing ARBA:ARBA00027853, across 1,570 taxa. A protein-name census of all 8,974 shows 47.6% carotenoid/retinoid cleavage enzymes (BCO1, BCO2, RPE65, NinaB), 30.4% prenyl and polyprenyl diphosphate synthases (PDSS1, PDSS2, FPPS, GGPPS), 16.0% uncharacterized entries, and a demonstrable false-positive tail of only a few percent. The 47 human annotations, enumerated exhaustively, are 100% on-target. The serious finding is a reproducibility failure: in a random 300-protein sample, 187 (62%) carry InterPro IPR004294 "Carotenoid oxygenase" but 167 of those 187 are not mammals, while the rule's only IPR004294 branch (CS7) requires taxon Mammalia; and 89 (30%) carry IPR000092 while only 25 of them also carry the IPR039702 that the rule's only IPR000092 branch (CS16) requires. Either the published conjunctions are not enforced at annotation time or the rule content served today has diverged from the annotating release. Either way, the constraints that make this rule look safe on paper are invisible in the data it is credited with, and that is a plausible mechanism for the "lots of off target inferences" reported by curators.
MODIFY rather than DEPRECATE, because the rule is currently producing roughly 8,900 annotations that are in the large majority correct and useful - carotenoid oxygenases and polyprenyl diphosphate synthases across 1,570 taxa, with a 100%-correct human subset. Retiring the rule would discard far more good annotation than bad. This is the opposite situation from ARBA00028655 in the same GO issue, where under 1% of emitted annotations were defensible and DEPRECATE/SPLIT was the right call. MODIFY rather than ACCEPT, because 13 branches are indefensible on their own terms and because the rule cannot be audited against its own output. The commissioned deep research recommends SPLIT into pathway-specific rules (precursor/backbone, terpene synthases, carotenoids, gibberellin/ABA, retinoids, sterol/triterpenoid, juvenile hormone), which is a reasonable longer-term target and would also let each fragment carry a more specific descendant term than GO:0006720; but the immediate, minimal fix that answers the curators' complaint is branch surgery plus reconciliation of the published rule with its emitted annotations, not restructuring. The single highest-priority action is not biological at all: UniProt should reconcile the condition sets with the annotation set and fix the statistics block that reports zero proteins for a rule with 8,974 live annotations - that field caused an entire commissioned literature review to be conducted on a false premise. The confidence value below is confidence in this recommendation, not in the rule.
ON-TARGET. IPR001906 (TPS N-terminal) + IPR005630 (TPS metal-binding C-terminal) + PANTHER PTHR31225:SF93. The PANTHER subfamily resolves to 'ALPHA-HUMULENE_(-)-(E)-BETA-CARYOPHYLLENE SYNTHASE' (panther.obo), a sesquiterpene synthase - the deep research treated bare PANTHER ids as unauditable, but the repository PANTHER build settles this one. Requiring both TPS domains plus a named subfamily is one of the better-constructed branches. No taxon constraint, which is fine for a subfamily this narrow.
KEEP WITH CAUTION. IPR008930 (terpenoid cyclase/protein prenyltransferase alpha-alpha toroid) + IPR008949 (isoprenoid synthase domain superfamily) + Ocimeae. Both conditions are homologous- superfamily entries, not families; the alpha-alpha toroid also covers squalene-hopene cyclase and the protein prenyltransferase alpha subunit. The Ocimeae (Lamiaceae) restriction reflects a real TPS expansion, so this is defensible, but two superfamily folds are weaker evidence than the family-level branches elsewhere in this rule.
ON-TARGET. IPR034741 (terpene cyclase-like 1, C-terminal) + IPR036965 (TPS N-terminal superfamily) + Pinus. Conifer terpene synthases are a genuine lineage-specific expansion and the two-domain requirement matches plant class-I TPS architecture.
ON-TARGET. PTHR31739:SF25 resolves to '(E,E)-GERANYLLINALOOL SYNTHASE' (panther.obo) - a diterpene synthase. The Eukaryota taxon adds nothing at subfamily resolution but is harmless.
ON-TARGET. IPR002060 (squalene/phytoene synthase) + IPR019845 (conserved site) + PTHR31480 = 'BIFUNCTIONAL LYCOPENE CYCLASE/PHYTOENE SYNTHASE'. Family plus conserved catalytic site plus named PANTHER family. GO:0006720 is the right level here precisely because the squalene/phytoene synthase fold cannot distinguish the sterol entry reaction from the carotenoid entry reaction without subfamily evidence.
HOLD / LIKELY REMOVE. IPR001128 is the generic cytochrome P450 domain; PTHR47955 resolves to 'CYTOCHROME P450 FAMILY 71 PROTEIN' (panther.obo). CYP71 is the largest and most promiscuous plant P450 family, acting on terpenoids, oximes, cyanogenic glucosides and glucosinolates. Family membership plus an Asterales restriction does not establish isoprenoid pathway participation. The deep research documents CYP706A3 with 29 substrates and CYP76M8 hydroxylating seven diterpene skeletons as evidence that plant P450 family membership is not diagnostic.
ON-TARGET, and empirically the single largest producer. IPR004294 (carotenoid oxygenase) + Mammalia captures BCO1, BCO2 and RPE65. Carotenoid cleavage is carotenoid catabolism and retinoid formation, both inside GO:0006720. NOTE: 62% of a random 300-protein sample of the rule's emitted set carries IPR004294, but 89% of those are NOT mammals - so either this taxon constraint is not enforced at annotation time or the published rule has diverged from the annotating release (see analysis section 3).
ON-TARGET. IPR036396 (P450 superfamily) + PTHR47950:SF4 + Nepetoideae. The PANTHER subfamily resolves to 'GERANIOL 8-HYDROXYLASE-LIKE' (panther.obo), i.e. the CYP76 monoterpenoid/iridoid- pathway enzymes. Unlike CS6 this branch is pinned to a named subfamily rather than a whole P450 family, which is what makes it acceptable.
ON-TARGET but fold-heavy. IPR002937 (amine oxidase) + IPR014105 (carotenoid/retinoid oxidoreductase) + IPR036188 (FAD/NAD(P)-binding superfamily). Only IPR014105 is diagnostic; the other two are broad cofactor-binding folds contributing little. Retain on the strength of IPR014105 alone, and consider dropping the two superfamily conditions rather than pretending they add specificity.
ON-TARGET. PTHR31225:SF98 resolves to 'TERPENE SYNTHASE 9-RELATED' (panther.obo). The deep research listed this among 'ambiguous, unnamed' branches to hold; the PANTHER lookup resolves it as a genuine TPS subfamily. Viridiplantae is appropriate.
ON-TARGET. IPR034686 (terpene cyclase-like 2) + Amoebozoa. Slime-mould terpene synthases are a documented lineage-specific family; the taxon restriction is biologically motivated rather than an annotation artifact.
ON-TARGET. PTHR31225:SF9 resolves to 'TERPENE SYNTHASE 10' (panther.obo). Elsholtzieae is a real Lamiaceae TPS expansion.
ON-TARGET. IPR017825 (lycopene cyclase domain) alone, no taxon constraint. Acceptable because the domain is family-level and functionally dedicated - one of the few single-condition branches in this rule that is safe.
OFF-TARGET - REMOVE. IPR002213 is the whole UDP-glucuronosyl/UDP-glucosyltransferase family: plant UGTs glycosylate flavonoids, phenolics, hormones, xenobiotics and terpenoids indiscriminately. A Caryophyllaceae restriction does not make the family diagnostic for isoprenoid metabolism. If the intent was triterpenoid saponin glycosyltransferases, the branch needs a specific subfamily, not a family plus a plant clade.
ON-TARGET. PTHR10543:SF57 resolves to 'RETINOID ISOMEROHYDROLASE' (panther.obo) = RPE65. The deep research could not audit this identifier and marked it for review; it is a named retinoid enzyme and retinoid metabolic process (GO:0001523) is a descendant of GO:0006720.
ON-TARGET. IPR000092 (polyprenyl synthetase-like) + IPR039702 (FPPS-like) + Ecdysozoa: farnesyl diphosphate synthase, a core isoprenoid chain-elongation enzyme. NOTE: 30% of a random 300-protein sample of the emitted set carries IPR000092 but only 28% of those also carry IPR039702, so the conjunction written here is not visible in the annotations the rule is credited with (analysis section 3).
ON-TARGET. IPR014102 (phytoene desaturase) + IPR050464 + PTHR42923:SF45 = '15-CIS-PHYTOENE DESATURASE, CHLOROPLASTIC_CHROMOPLASTIC' (panther.obo). Carotenoid backbone desaturation.
ON-TARGET. IPR010108 (lycopene cyclase, beta/epsilon) + Streptophyta. Family-level and pathway- dedicated.
ON-TARGET. PTHR31225:SF120 resolves to 'GERMACRENE-A SYNTHASE' (panther.obo), a sesquiterpene synthase. Embryophyta is appropriate.
ON-TARGET, and the deep research's objection to it is mistaken. It flagged the conjunction of FunFam 1.10.600.10:FF:000005 ('ent-kaur-16-ene synthase') AND 1.50.10.130:FF:000004 ('carene synthase') as possibly 'internally contradictory or yield[ing] no proteins'. The two FunFams sit in different CATH superfamilies - 1.10.600.10 is the isoprenoid-synthase alpha-helical fold, 1.50.10.130 the terpene-cyclase beta-gamma fold - and plant class-I diterpene synthases carry both. This is a domain-architecture requirement, and a well-constructed one.
ON-TARGET. FunFam 1.50.10.130:FF:000002 (ent-copalyl diphosphate synthase) + Nepetoideae. Class- II diterpene cyclase, the committed step to labdane-type diterpenoids.
ON-TARGET. FunFam 1.10.600.10:FF:000007 (isoprene synthase, chloroplastic) + Solanales. Named, narrow FunFam.
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000043 is labelled 'Cytochrome P450 99A2' but the branch is restricted to campanulids. CYP99A2 is a rice (Poales) momilactone-pathway diterpenoid P450; campanulids contain no rice. Either the FunFam label is being read as a function rather than as the name of its best-studied member, or the taxon is wrong. The branch cannot be signed off until the actual campanulid members are inspected.
ON-TARGET. FunFam 1.10.600.10:FF:000004 (phytoene synthase, chloroplastic). No taxon constraint, acceptable for a named pathway-dedicated FunFam.
ON-TARGET. FunFam 1.50.10.160:FF:000001 (ent-copalyl diphosphate synthase) + Viridiplantae. Overlaps CS21 in concept; see condition_overlap.
KEEP WITH REVIEW. FunFam 1.10.630.10:FF:000007 ('Cytochrome P450 76C4') + Mentheae. CYP76C-subfamily enzymes include monoterpenol oxidases in Lamiaceae, so the pairing of subfamily and clade is coherent - but CYP76C1 in Arabidopsis also acts outside isoprenoid metabolism, and this is still a P450 FunFam. Weaker than CS8, which uses a PANTHER subfamily named for the reaction.
ON-TARGET. FunFam 1.50.10.20:FF:000011 ('Terpene cyclase/mutase family member') + fabids. CATH 1.50.10.20 is the squalene-hopene cyclase / oxidosqualene cyclase superfamily - triterpene cyclization, squarely isoprenoid.
ON-TARGET. FunFam 2.60.120.330:FF:000003 (gibberellin 20-oxidase 2) + Streptophyta. GAs are diterpenoids; GO:0009686 gibberellin biosynthetic process is a descendant of GO:0006720.
ON-TARGET. FunFam 2.60.120.330:FF:000013 (gibberellin 3-beta-dioxygenase 1) + Embryophyta.
ON-TARGET. FunFam 1.10.600.10:FF:000020 (phytoene synthase). Duplicates CS24's concept with a different FunFam.
ON-TARGET. FunFam 2.60.120.330:FF:000014 (gibberellin 2-beta-dioxygenase 1) + Tracheophyta. GA deactivation is catabolism, still 'metabolic process'.
ON-TARGET. FunFam 3.50.50.60:FF:000091 (15-cis-phytoene desaturase, chloroplastic/chromoplastic) + Spermatophyta.
ON-TARGET. FunFam 3.50.50.60:FF:000101 (lycopene epsilon cyclase, chloroplastic) + Magnoliopsida.
ON-TARGET - one of the strongest branches. Three HMG-CoA reductase FunFams (1.10.3270.10:FF:000002, 3.30.70.420:FF:000001, 3.90.770.10:FF:000001) required together, matching HMGR's three-domain architecture. The rate-limiting enzyme of the mevalonate pathway. No taxon constraint, correctly.
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000008 is labelled 'Cytochrome P450 71D8' but the branch is restricted to Poales. CYP71D8 is a legume enzyme. As with CS23 and CS75, a P450 FunFam label is being treated as a function. CYP71D does contain terpenoid hydroxylases (e.g. CYP71D13/18 in mint) but also many non-isoprenoid enzymes; the Poales members must be inspected before this branch can be kept.
ON-TARGET. FunFam 1.10.630.10:FF:000041 (cytochrome P450 26A1) + Metazoa. CYP26A1 is the principal retinoic acid 4-hydroxylase; retinoid metabolic process is under GO:0006720. Metazoa is if anything narrower than needed.
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000299 ('Cytochrome P450 2C9') + Primates. CYP2C9 is a hepatic drug- and xenobiotic-metabolizing P450; its endogenous substrates are chiefly arachidonic acid and related eicosanoids, not isoprenoids. The deep research lists this among the branches to remove outright.
HOLD, but not for the reason the deep research gave. FunFam 1.50.10.20:FF:000064 is labelled 'Uncharacterized protein' + Saxifragales, and the deep research recommended deleting all unnamed FunFams. That blanket rule is wrong here: CATH 1.50.10.20 is the terpene cyclase/mutase (SHC/OSC) superfamily, which is functionally dedicated to isoprenoid cyclization, so the fold itself carries information even though the FunFam is unnamed. Inspect the Saxifragales members; this is a likely keep, not a likely delete.
MIXED - the rule's one systematic biological false positive. FunFam 3.40.50.720:FF:000808 ('Iridoid synthase') + asterids is the PRISE family (progesterone 5-beta-reductase / iridoid synthase enzymes). Its iridoid synthase members are correctly isoprenoid (monoterpenoid), but its progesterone 5-beta-reductase and 3-oxo-Delta(4,5)-steroid 5-beta-reductase members are steroid enzymes, and GO:0008202 steroid metabolic process is NOT a descendant of GO:0006720 (QuickGO ancestor closure, 2026-08-22). Empirically the steroid side outnumbers the iridoid side in this rule's output: 1.5% vs 0.9% of 8,974 annotations. Needs splitting to the iridoid synthase members only.
ON-TARGET. FunFam 1.50.10.130:FF:000006 ('Terpene synthase 7'). Named TPS FunFam, no taxon constraint.
ON-TARGET. Diphosphomevalonate decarboxylase FunFams (3.30.230.10:FF:000018 + 3.30.70.890:FF:000005) + Mammalia. Core mevalonate pathway. The Mammalia restriction is unnecessarily narrow - the MVA pathway is not mammal-specific.
ON-TARGET. Mevalonate kinase FunFams (3.30.230.10:FF:000119 + 3.30.70.890:FF:000003) + Chordata. Core mevalonate pathway; taxon again narrower than the biology.
REMOVE (as an unsatisfiable branch, not as an off-target one). Conjunction of FunFam 3.40.309.10:FF:000001 ('Mitochondrial aldehyde dehydrogenase 2'), 3.40.605.10:FF:000026 ('Aldehyde dehydrogenase, putative') and 3.40.605.10:FF:000054 ('Aldehyde dehydrogenase family 1 member A3'). The three FunFams are AND-ed, so - as for CS20 and CS61 - the broad conjuncts do not widen the match set and the most specific conjunct binds it: any protein this fires on is a subset of ALDH1A3-like proteins, whose retinaldehyde-dehydrogenase activity is legitimately within GO:0006720 via GO:0001523. The real defect is satisfiability rather than aim. 3.40.309.10 and 3.40.605.10 are the two ALDH-fold CATH superfamilies, and while requiring one FunFam from each is a sound domain-architecture requirement, this branch additionally requires two distinct FunFams (FF:000026 and FF:000054) from within the same superfamily 3.40.605.10 - which a single ALDH catalytic domain cannot satisfy. CS43 is therefore near-certainly a dead branch that matches nothing, which makes it further evidence for the section 3 finding that the emitted set cannot be reproduced from the published condition sets. No taxon constraint, so were it satisfiable it would fire across all of UniProt.
OFF-TARGET - REMOVE. Hormone-sensitive lipase FunFams (3.40.50.1820:FF:000110, :FF:000199) + Craniata. LIPE does have retinyl ester hydrolase activity in vitro, but its established biological role is neutral-lipid (triacylglycerol, cholesteryl ester) hydrolysis in adipocytes. A side activity on a retinyl ester is not grounds for a blanket isoprenoid metabolic process annotation.
ON-TARGET. FunFam 1.10.600.10:FF:000001 (geranylgeranyl diphosphate synthase) + Bacteria. Prenyl chain assembly.
ON-TARGET. FunFam 1.10.600.10:FF:000019 (2-methylisoborneol synthase) + Bacillati. A genuine actinobacterial monoterpene synthase; the taxon restriction is biologically motivated.
ON-TARGET. FunFam 1.10.600.10:FF:000047 ('Terpene synthase') + Amoebozoa. Overlaps CS70 and CS71 in concept.
ON-TARGET. FunFam 3.40.50.150:FF:000183 (geranyl diphosphate 2-C-methyltransferase) + Actinomycetota. Named enzyme acting directly on a prenyl diphosphate.
HOLD. FunFam 3.40.50.720:FF:000084 ('Short-chain dehydrogenase reductase') + Lamiaceae. CATH 3.40.50.720 is the Rossmann NAD(P)-binding superfamily, the single most promiscuous fold in this rule. The FunFam label is generic. If the intent was Lamiaceae monoterpene reductases (e.g. menthone reductases), that needs to be established from the members, not assumed from the clade.
ON-TARGET, taxon far too narrow. FunFam 3.40.50.720:FF:000145 (retinol dehydrogenase 12) + Haplorrhini. RDH12 is a genuine retinoid enzyme, but restricting it to Haplorrhini is an annotation-availability artifact - RDH12 orthologues are conserved across vertebrates.
OFF-TARGET - REMOVE. FunFam 3.40.50.720:FF:001857 ('Alcohol dehydrogenase class 4 mu/sigma chain') + Vertebrata = ADH7. ADH7 does oxidize retinol, but it is principally an ethanol/aliphatic-alcohol dehydrogenase of upper-aerodigestive mucosa, and a generic Rossmann FunFam label of 'alcohol dehydrogenase class 4' is not a proxy for retinoid metabolism.
ON-TARGET. 1-deoxy-D-xylulose-5-phosphate synthase FunFams (3.40.50.920:FF:000002 + 3.40.50.970:FF:000005). DXS is the entry enzyme of the MEP pathway - one of the most diagnostic possible matches for GO:0006720.
ON-TARGET. FunFam 3.50.50.60:FF:000378 (phytoene desaturase) + Pseudomonadati.
ON-TARGET. FunFam 3.90.1720.10:FF:000006 (lecithin retinol acyltransferase) + Euteleostomi. LRAT esterifies retinol for storage - retinoid metabolism.
OFF-TARGET - REMOVE, and the clearest single error in the rule. FunFam 3.90.180.10:FF:000001 ('S-(hydroxymethyl)glutathione dehydrogenase') restricted to Mus. This is ADH5/class-III ADH, the formaldehyde/S-nitrosoglutathione-detoxifying enzyme. It does not act on retinol and has no isoprenoid role. A genus-level (Mus) restriction on a universally conserved housekeeping enzyme is a textbook association-rule mining artifact.
ON-TARGET. FunFam 1.10.600.10:FF:000036 (cis-abienol synthase, chloroplastic). Named diterpene synthase.
ON-TARGET. FunFam 1.10.630.10:FF:000052 (ent-kaurenoic acid oxidase). KAO catalyses three sequential oxidations converting ent-kaurenoic acid to GA12 - a committed gibberellin-pathway step.
ON-TARGET. FunFam 1.10.630.10:FF:000062 (ent-kaurene oxidase 2). KO, the preceding gibberellin- pathway step.
ON-TARGET. FunFam 3.50.50.60:FF:000171 (zeta-carotene-forming phytoene desaturase).
ON-TARGET. FunFam 3.50.50.60:FF:000413 (phytoene desaturase, lycopene-forming).
ON-TARGET. FunFam 1.10.600.10:FF:000042 ('Probable terpene synthase 3') + 1.50.10.130:FF:000005 ('S-(+)-linalool synthase') + Poaceae. Same two-superfamily domain-architecture logic as CS20, and equally sound.
ON-TARGET. FunFams 1.50.10.20:FF:000002 + :FF:000022 ('Terpene cyclase/mutase family member') + Fagales. Triterpene cyclase.
OFF-TARGET - REMOVE, wrong on two independent grounds. FunFam 2.10.25.10:FF:000009 ('Low-density lipoprotein receptor') + 4.10.400.10:FF:000011 ('LRP1') + Catarrhini. (1) LDLR and LRP1 are endocytic receptors, not metabolic enzymes - this conflates transport of a lipid particle with metabolism of its contents. (2) Even the intended inference fails: the cargo is cholesterol, and GO:0008202 steroid metabolic process is not a descendant of GO:0006720. CATH 2.10.25.10 is also the EGF/laminin superfamily, one of the most promiscuous small structural domains in existence.
OFF-TARGET - REMOVE. Patatin-like phospholipase domain-containing protein 2 FunFams (3.40.1090.10:FF:000003, :FF:000021) + Eutheria = PNPLA2/ATGL. ATGL's established role is triacylglycerol lipolysis; its reported retinyl ester hydrolase activity is a side activity and cannot support a blanket isoprenoid annotation. Same failure mode as CS44.
HOLD, but the case for retaining it is stronger than first stated. FunFam 3.40.309.10:FF:000021 ('Aldehyde dehydrogenase family 8 member A1') + 3.40.605.10:FF:000001 ('Aldehyde dehydrogenase 1') + Euarchontoglires. ALDH8A1 does have 9-cis-retinal dehydrogenase activity and ALDH1A1 is a retinal dehydrogenase, so the intent is legitimate. The earlier objection - that 'Aldehyde dehydrogenase 1' is too broad to exclude the wider ALDH superfamily - misread the conjunction as a disjunction. The two FunFams are AND-ed and sit in the two distinct ALDH-fold CATH superfamilies (3.40.309.10 and 3.40.605.10), so this is the same sound domain-architecture pattern accepted for CS20 and CS61; the broad conjunct cannot widen the match set, and the specific ALDH8A1 conjunct binds it. Unlike CS43 the branch is satisfiable, since it draws one FunFam from each superfamily rather than two from one. HOLD is retained only pending a member census to confirm the matched set is in fact ALDH8A1-like, not because the conjunction is unsound.
OFF-TARGET - REMOVE. FunFam 1.10.274.20:FF:000003 ('Phenylalanine aminomutase (L-beta- phenylalanine forming)') + Pinopsida. PAM is an MIO-dependent aminomutase of amino-acid specialized metabolism. It supplies the phenylisoserine side chain of paclitaxel, which is why it appears in a conifer context - but the enzyme's substrate and product are amino acids, not isoprenoids. Supplying a non-isoprenoid moiety to a taxane is not isoprenoid metabolism.
ON-TARGET. FunFam 1.10.600.10:FF:000008 (farnesyl pyrophosphate synthase) + rosids.
ON-TARGET. FunFam 1.10.600.10:FF:000018 ('Probable geranylgeranyl-diphosphate geranylgeranyltransferase (AL-2)') + Fungi. AL-2 is the Neurospora phytoene synthase/GGPP- transferase of the carotenoid pathway.
ON-TARGET. FunFam 1.10.600.10:FF:000021 (farnesyl pyrophosphate synthase) + Ecdysozoa. Largely redundant with CS16, which reaches the same enzymes in the same clade via InterPro.
ON-TARGET. FunFam 1.10.600.10:FF:000076 ('Terpene synthase') + Evosea. Redundant with CS47 and CS71: Evosea and Eumycetozoa are both inside Amoebozoa.
ON-TARGET. FunFam 1.10.600.10:FF:000078 ('Terpene synthase') + Eumycetozoa. See CS70.
ON-TARGET, taxon absurdly narrow. FunFam 1.10.630.10:FF:000009 (cytochrome P450 26B1) + Hominidae. CYP26B1 is a retinoic acid hydroxylase conserved across vertebrates; a Hominidae restriction is an annotation artifact, not biology.
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000011 ('Cytochrome P450 83B1') + Asterales. CYP83B1 (Arabidopsis SUR2) oxidizes indole-3-acetaldoxime in indole glucosinolate biosynthesis. Glucosinolates are amino-acid-derived, not isoprenoid. Asterales do not make glucosinolates, so the taxon is also inconsistent with the label - the same name/taxon contradiction seen in CS23, CS35 and CS75.
ON-TARGET, label needs fixing. FunFam 1.10.630.10:FF:000014 + lamiids. The label 'Abscisic acid 8' is a truncation of 'abscisic acid 8'-hydroxylase' (CYP707A), the principal ABA catabolic enzyme. ABA is an apocarotenoid, so ABA catabolism is inside GO:0006720. The truncated label should be corrected before the branch is relied on, since a label that stops mid-word is exactly the kind of thing that hides a wrong FunFam id.
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000022 is labelled 'Taxadiene 5-alpha hydroxylase' (CYP725A4, a Taxus/Pinopsida taxane P450) but the branch is restricted to Caryophyllales, which contains no Taxus. The named enzyme would be on-target; the members actually captured in Caryophyllales are unknown and must be inspected.
ON-TARGET. FunFam 1.10.630.10:FF:000080 ('Carotene epsilon-monooxygenase, chloroplastic') + eudicotyledons = LUT1/CYP97C, the carotenoid epsilon-ring hydroxylase of the lutein pathway.
HOLD. FunFam 1.10.630.10:FF:000097 ('Cytochrome P-450 19') + PACMAD clade. 'Cytochrome P-450 19' is ambiguous - in animals CYP19 is aromatase, in plants the numbering is unrelated. A P450 FunFam plus a grass clade is not diagnostic without inspecting members.
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000182, no label at all, restricted to Homo. A bare P450 FunFam pinned to a single genus is the weakest possible evidence for a biological process term, and the Homo restriction marks it as an annotation-availability artifact.
ON-TARGET. FunFam 2.40.400.10:FF:000003 ('Protein NEOXANTHIN-DEFICIENT 1') + Gunneridae. NSY converts violaxanthin to neoxanthin - xanthophyll (carotenoid) metabolism.
HOLD. FunFam 2.60.200.20:FF:000048 + 3.50.50.60:FF:000263, both unlabelled, no taxon constraint. 3.50.50.60 is the FAD/NAD(P)-binding superfamily that legitimately carries the phytoene desaturases in CS32/CS53/CS59/CS60, so the conjunction may well be a carotenoid enzyme - but with neither FunFam named and no taxon constraint, this branch fires across all of UniProt on unverifiable evidence.
ON-TARGET. FunFam 3.40.50.150:FF:000539 ('juvenile hormone acid O-methyltransferase') + Arthropoda. JH is a sesquiterpenoid; JHAMT catalyses a committed late step. Well-matched taxon.
HOLD / LIKELY REMOVE. FunFam 3.40.50.2000:FF:000019, unlabelled, + Apiales. CATH 3.40.50.2000 is the GT-B glycosyltransferase superfamily. Apiales include Panax (ginsenoside triterpenoid glycosides), so a terpenoid glycosyltransferase is plausible - but 'plausible given the clade' is precisely the reasoning that produces the errors in this rule. Inspect members.
OFF-TARGET - REMOVE. FunFam 3.40.50.2000:FF:000037 labelled only 'Glycosyltransferase' + Caryophyllaceae. A generic GT-B FunFam with a generic label. Same objection as CS14.
HOLD. FunFam 3.40.50.2000:FF:000040, unlabelled, + Gentianales. Gentianales include the monoterpene indole alkaloid producers, so a secologanin-pathway glycosyltransferase is possible. Unverifiable as written.
HOLD. FunFam 3.40.50.2000:FF:000047, unlabelled, + Caryophylleae. Triterpenoid saponin glycosylation is plausible in this clade; again unverifiable from the identifier.
ON-TARGET. FunFam 3.50.50.60:FF:000074 ('Squalene monooxygenase 2') + Araliaceae. SQE oxidizes squalene to 2,3-oxidosqualene, the branch point to triterpenoids and sterols; in Araliaceae this feeds ginsenoside biosynthesis. Note the caveat that the sterol branch downstream is outside GO:0006720, but the enzyme itself acts on squalene, an isoprenoid.
HOLD. FunFam 3.90.79.10:FF:000025, unlabelled, + Pentapetalae. CATH 3.90.79.10 is the Nudix hydrolase superfamily. Some plant Nudix enzymes are genuine prenyl-diphosphate phosphatases (e.g. RhNUDX1 in rose, which makes monoterpene alcohols), but the superfamily is overwhelmingly composed of unrelated nucleotide hydrolases. Not usable without member inspection.
ON-TARGET, and empirically the second-largest producer. FunFam 1.10.600.10:FF:000011 ('Decaprenyl diphosphate synthase subunit 1') = PDSS1. Together with the other prenyl-synthase branches this accounts for ~30% of the rule's emitted annotations. No taxon constraint, correctly.
KEEP WITH REVIEW. FunFam 1.50.10.130:FF:000003, unlabelled, no taxon constraint. Like CS38, the superfamily itself is informative: CATH 1.50.10.130 is the terpene-cyclase beta-gamma fold, used by plant class-II diterpene synthases (it is the partner domain in CS20, CS21, CS40 and CS61). Likely a genuine TPS, but confirm the members.
HOLD. FunFam 2.60.120.330:FF:000021, unlabelled, no taxon constraint. CATH 2.60.120.330 is the double-stranded beta-helix 2-oxoglutarate/Fe(II) dioxygenase fold. The gibberellin oxidases in CS28/CS29/CS31/CS91 live here, but so do hundreds of unrelated 2ODDs acting on flavonoids, alkaloids and amino acids. An unnamed FunFam in this superfamily with no taxon constraint is not safe.
ON-TARGET. FunFam 2.60.120.330:FF:000025 ('Gibberellin 2-beta-dioxygenase 2'). Named GA catabolic enzyme.
HOLD. FunFam 2.60.120.330:FF:000050, unlabelled, no taxon constraint. Same objection as CS90.
ON-TARGET. FunFam 3.40.50.11270:FF:000001 ('4-hydroxy-3-methylbut-2-enyl diphosphate reductase') = IspH/LytB, the final MEP-pathway enzyme producing IPP and DMAPP. Among the most diagnostic branches in the rule.
OFF-TARGET - REMOVE. FunFam 3.40.50.720:FF:000131, unlabelled, no taxon constraint, in the Rossmann NAD(P)-binding superfamily. Unnamed, unconstrained and in the most promiscuous fold present - this branch cannot support any biological process term.
94 condition sets and 122 distinct signature entries for a single GO term is far past the point where a rule can be reasoned about; the repository's own analysis tooling refuses to process it, capping post-enrichment analysis at 12 condition sets. Some of the complexity is legitimate - isoprenoid metabolism genuinely spans plant terpene synthases, bacterial 2-methylisoborneol synthases, mammalian carotenoid oxygenases, insect JHAMT and the MVA and MEP pathways, and no small set of signatures covers that. But much of it is accretion. The same enzyme concept is expressed repeatedly through different databases and taxa: farnesyl diphosphate synthase appears in CS16 (InterPro, Ecdysozoa), CS67 (FunFam, rosids) and CS69 (FunFam, Ecdysozoa, overlapping CS16 directly); ent-copalyl diphosphate synthase in CS21 and CS25; phytoene synthase in CS24 and CS30; slime-mould terpene synthases in CS47, CS70 and CS71 across three nested clades of Amoebozoa; phytoene desaturase in CS17, CS32, CS53, CS59 and CS60. The decisive parsimony argument, though, is empirical rather than structural: the emitted annotation set is dominated by carotenoid oxygenases and prenyl diphosphate synthases, so the great majority of the 94 branches contribute nothing observable. The ~19 terpene synthase branches - the largest single group - account for 0.4% of output.
The commissioned deep research (Falcon / Edison Scientific Literature, run 2026-08-22) supports the GO term for the rule's core families and contradicts a defined minority of its branches - a genuinely mixed verdict, unlike the flat contradiction found for ARBA00028655. Its positive case rests on pathway-level reviews placing carotenoids, gibberellins, ABA and related hormones downstream of IPP/DMAPP and prenyl diphosphates (Bajguz & Piotrowska- Niczyporuk 2023, Metabolites 13:884), and on terpene synthase biology (Karunanithi & Zerbe 2019, PMID:31632418). Its negative case rests on documented enzyme promiscuity: Werck- Reichhart 2023 (PMID:36830762) reports that CYP706A3 "oxidizes more than twenty different mono- and sesquiterpenes" and that CYP720B4 "catalyzes the three successive oxidations at C18 of 8 out of 24 different diterpenoid olefin skeletons" - which is what makes "generic P450 domain plus plant clade" non-diagnostic. It also cites quantitative limits on domain-based process transfer - DomFun Fmax 0.624 for molecular function versus 0.492 for biological process (Rojano et al. 2022, PMID:35033002); FunFam members agreeing on only 36.9 +/- 0.6% of binding-residue annotations (Scheibenreif et al. 2019, PMID:31319797). Four honest limits on this evidence. First, the report was given the API's false "zero annotated proteins" figure and built its "no present production set from which empirical precision can be measured" reasoning on it; the census in the companion analysis corrects that. Second, it does not mention the GO Consortium or issue #5835 anywhere, so it offers no independent read on the curators' complaint and no Consortium ruling should be claimed from it. Third, its blanket recommendation to hold or delete "opaque" PANTHER identifiers and unnamed FunFams is too coarse - nine of the PANTHER branches resolve to named isoprenoid families, and two unnamed FunFams sit in isoprenoid-dedicated cyclase superfamilies. Fourth, one figure it quotes needs a scope caveat rather than a correction: its "29 documented substrates" for CYP706A3 is accurate - PMID:36830762 states verbatim "A total of 29 different substrates are thus currently reported for this enzyme" - but that total is reached only by adding the dinitroaniline herbicides to the terpenoid substrates, so 29 is not a count of terpenoid substrates and should not be cited as one. Only one external provider was reachable in this environment: Perplexity returned an insufficient-quota error, Cyberian is not configured, and the OpenAI deep-research model returned a 404, so this review rests on one commissioned report plus the independent empirical census rather than on two providers.
Quantitative pairwise overlap could not be computed: the repository's analyse-rule step refuses rules above 12 condition sets ("Rule ARBA00027853 has 94 condition sets, which exceeds the maximum of 12"), so no pairwise_overlap blocks are populated below and the assessment here is structural rather than metric. Read at that level the redundancy is plain. Farnesyl diphosphate synthase is reached three times - CS16 (IPR000092 + IPR039702, Ecdysozoa), CS67 (FunFam 1.10.600.10:FF:000008, rosids) and CS69 (FunFam 1.10.600.10:FF:000021, Ecdysozoa) - with CS16 and CS69 targeting the same enzymes in the same clade through different databases. Phytoene desaturase appears in five sets (CS17, CS32, CS53, CS59, CS60), phytoene synthase in three (CS5, CS24, CS30), ent-copalyl diphosphate synthase in two (CS21, CS25), triterpene cyclases in three (CS27, CS38, CS62), and Amoebozoan terpene synthases in three nested clades (CS47 Amoebozoa, CS70 Evosea, CS71 Eumycetozoa - Evosea and Eumycetozoa both lie inside Amoebozoa, so CS70 and CS71 add coverage only if their FunFams differ from CS47's, which is not established). CATH superfamily 1.10.600.10 alone supplies conditions to 14 different sets and 1.10.630.10 to 13. Cross-database redundancy of this kind is not harmful in itself - it buys coverage when InterPro, PANTHER and CATH disagree about a protein - but at this scale it makes the rule unmaintainable and hides which branch is responsible for any given annotation, which is exactly the problem when a curator files an off-target report.
GO:0006720 is the right level for a rule whose branches span terpene, carotenoid, retinoid, gibberellin, ABA, triterpenoid and juvenile-hormone chemistry, and it is genuinely correct for the families that dominate the output. Its definition - "compounds containing or derived from linked isoprene residues" - covers apocarotenoids and retinoids, and QuickGO ancestor closures checked on 2026-08-22 confirm GO:0001523 retinoid metabolic process, GO:0006721 terpenoid metabolic process, GO:0016114 terpenoid biosynthetic process, GO:0016109 tetraterpenoid biosynthetic process, GO:0016102 diterpenoid biosynthetic process and GO:0009686 gibberellin biosynthetic process are all descendants. Breadth is also a virtue for specific branches: CS5 combines the squalene/phytoene synthase family with its conserved site, and that signature cannot distinguish the sterol entry reaction from the carotenoid entry reaction, so the shared parent is the honest term. The one place the term is wrong rather than broad is where a branch reaches steroid chemistry: GO:0008202 steroid metabolic process is not a descendant of GO:0006720, which condemns CS63 (LDLR/LRP1, cholesterol uptake) and the progesterone 5-beta-reductase half of CS39. Against APPROPRIATE it must be said that for the strongest branches - HMGCR, DXS, IspH, mevalonate kinase, the GA oxidases, CYP26, LRAT, JHAMT - a specific descendant would carry far more information than the shared parent, and the deep research argues the term is "too broad as the only annotation" for those. That is an argument for splitting the rule, which is recorded under action_rationale, not for calling the term itself mismatched.
The constraints are internally inconsistent, and the dominant failure is over-restriction on conserved families rather than over-inclusion. Six branches pin universally conserved enzymes to a single genus, family or primate clade: CS55 restricts ADH5 - a housekeeping formaldehyde dehydrogenase present in essentially every eukaryote - to Mus; CS78 restricts an unlabelled P450 FunFam to Homo; CS72 restricts CYP26B1 to Hominidae and CS36 restricts CYP26A1 to Metazoa though both are vertebrate-wide; CS50 restricts RDH12 to Haplorrhini; CS37 restricts CYP2C9 to Primates; CS63 restricts LDLR/LRP1 to Catarrhini. Restrictions of that shape are artifacts of where experimental annotation happens to exist, not statements about where the enzyme exists, and they guarantee the rule under-annotates real orthologues. Similarly CS41 (Mammalia) and CS42 (Chordata) restrict diphosphomevalonate decarboxylase and mevalonate kinase, which are not mammal- or chordate-specific. In the other direction, 24 branches carry no taxon constraint at all and fire across all of UniProt, including CS94 (bare Rossmann FunFam) and, nominally, CS43 (ALDH FunFams) - though CS43 appears unsatisfiable, so the missing constraint there is moot. Where such a branch does fire, a constraint would at least have limited the damage. Where the restrictions do track biology they are well chosen - the Lamiaceae tribes Ocimeae, Nepetoideae, Elsholtzieae and Mentheae for terpene synthase expansions, Pinus for conifer TPS, Amoebozoa for slime-mould TPS, Bacillati and Actinomycetota for 2-methylisoborneol synthase and geranyl diphosphate methyltransferase, Arthropoda for JHAMT - and those should be kept. Finally, the empirical audit shows the constraints may not be doing what they appear to: 167 of 187 sampled IPR004294 proteins are non-mammalian although the only IPR004294 branch requires Mammalia.
GO:0006720 is an appropriate broad parent for the rule's genuine pathway enzymes - terpene synthases, MVA/MEP precursor enzymes, prenyl diphosphate synthases, carotenoid enzymes, gibberellin and ABA enzymes, retinoid enzymes and JHAMT - but the rule is "unsafe as a single 94-branch OR rule".
Eight branch groups are off-target and should be removed: CYP2C9 (CS37), the mixed ALDH2 conjunction (CS43), hormone-sensitive lipase (CS44), PNPLA2 (CS64), alcohol and formaldehyde dehydrogenases (CS51, CS55), LDLR/LRP1 (CS63), phenylalanine aminomutase (CS66) and CYP83B1 (CS73).
Plant cytochrome P450 promiscuity is the rule rather than the exception - CYP706A3 has 29 documented substrates spanning more than 20 mono- and sesquiterpenes plus herbicides - so a generic P450 domain plus a plant taxon cannot predict isoprenoid metabolism.
Domain-based transfer is measurably weaker for biological process than for molecular function (DomFun Fmax 0.492 vs 0.624), and FunFam members agree on only 36.9% of binding-residue annotations, so a bare FunFam identifier is not self-validating for a process term.
The report was given the UniProt API's figure of zero annotated proteins and reasoned from it that no empirical precision estimate was possible. That premise is false; see the companion analysis.
QuickGO on 2026-08-22 reports 8,974 live GO:0006720 / ECO:0000256 / GO_REF:0000117 annotations citing ARBA:ARBA00027853 across 1,570 taxa, contradicting the rule's own statistics block of zero proteins.
A protein-name census of all 8,974 gives 47.6% carotenoid/retinoid cleavage enzymes, 30.4% prenyl and polyprenyl diphosphate synthases and 16.0% uncharacterized entries, with a demonstrable false-positive tail of a few percent; the 47 human annotations are 100% on-target.
In a random 300-protein sample, 187 carry IPR004294 but 167 of those are non-mammalian although the rule's only IPR004294 branch requires Mammalia, and only 25 of 89 IPR000092 proteins carry the co-required IPR039702 - so the emitted set cannot be reproduced from the published condition sets.
GO:0008202 steroid metabolic process is not a descendant of GO:0006720, which makes CS63 (LDLR/LRP1) and the progesterone 5-beta-reductase half of CS39 wrong-branch rather than merely broad.
Nine PANTHER conditions the deep research treated as unauditable resolve against interpro/panther/panther.obo to named isoprenoid families, and the CS20/CS61 multi-FunFam conjunctions it suspected of being contradictory are sound domain-architecture requirements spanning two different CATH superfamilies.
Re-derives the annotation count, protein-name census and signature/taxon audit from primary sources with nothing hard-coded; run on 2026-08-22 it reproduces the figures quoted in the analysis and this review.
Places gibberellins, abscisic acid, strigolactones and brassinosteroids downstream of IPP/DMAPP and the prenyl diphosphates, which is the pathway-level basis for treating GA20ox/GA3ox/GA2ox, ent-kaurene oxidase, ent-kaurenoic acid oxidase and the ABA 8'-hydroxylases (CS28, CS29, CS31, CS57, CS58, CS74, CS91) as isoprenoid metabolism.
CYP706A3 "oxidizes more than twenty different mono- and sesquiterpenes", and CYP720B4 "catalyzes the three successive oxidations at C18 of 8 out of 24 different diterpenoid olefin skeletons". Plant P450 substrate breadth of this kind is why a generic P450 domain plus a plant clade (CS6, CS23, CS26, CS35, CS73, CS77) cannot predict isoprenoid metabolism.
Plant terpene synthase families diversify by duplication, domain loss/gain and small active-site changes that redirect carbocation cascades and product profiles. This supports annotating TPS branches at the level of GO:0006720 rather than transferring product-specific terms from a broad TPS domain.
Domain-based function transfer benchmarks measurably worse for biological process than for molecular function (Fmax 0.492 vs 0.624 in the no-knowledge partial evaluation), which is the general reason a FunFam or InterPro signature is weaker evidence for GO:0006720 than it would be for a corresponding molecular function term.
CATH FunFam-based function prediction reaches competitive but bounded accuracy, with union models increasing false-positive risk and intersection models trading coverage for precision - the trade-off this rule makes implicitly by OR-ing 94 branches.
FunFam members agree on only 36.9 +/- 0.6% of binding-residue annotations - better than random grouping but far from uniform. A named, well-populated FunFam can be strong evidence, but a bare FunFam identifier is not self-validating, which is the basis for holding rather than accepting the 16 unlabelled-FunFam branches.
id: ARBA00027853
description: >-
Assigns GO:0006720 "isoprenoid metabolic process" to proteins matching any of 94
alternative condition sets built from InterPro entries, CATH FunFams, PANTHER families
and taxon constraints. The GO term is in the right branch for most of the rule: 64 of the
94 sets identify genuine isoprenoid enzymes - terpene synthases and cyclases, the MVA and
MEP precursor pathways, prenyl diphosphate synthases, carotenoid backbone and cleavage
enzymes, gibberellin and ABA oxidases, named retinoid enzymes and JHAMT. Thirteen sets
should be removed - twelve off-target (LDLR/LRP1, CYP2C9, CYP83B1, ADH5, ADH7,
hormone-sensitive lipase, PNPLA2, phenylalanine aminomutase, bare UGT and Rossmann
FunFams) plus one ALDH FunFam conjunction that appears unsatisfiable - one set is mixed
and needs splitting, and sixteen cannot be audited from their identifiers. The rule's most
serious defect is not biological: the 8,974 annotations it currently emits cannot be
reproduced from its published condition sets.
status: COMPLETE
rule_type: ARBA
rule:
rule_id: ARBA00027853
condition_sets:
- number: 1
conditions:
- condition_type: INTERPRO
value: IPR001906
curie: InterPro:IPR001906
label: "Terpene synthase, N-terminal domain"
interpro_type: DOMAIN
negated: false
- condition_type: INTERPRO
value: IPR005630
curie: InterPro:IPR005630
label: "Terpene synthase-like, metal-binding domain"
interpro_type: DOMAIN
negated: false
- condition_type: PANTHER
value: PTHR31225:SF93
curie: PTHR31225:SF93
negated: false
notes: >-
ON-TARGET. IPR001906 (TPS N-terminal) + IPR005630 (TPS metal-binding C-terminal) + PANTHER
PTHR31225:SF93. The PANTHER subfamily resolves to 'ALPHA-HUMULENE_(-)-(E)-BETA-CARYOPHYLLENE
SYNTHASE' (panther.obo), a sesquiterpene synthase - the deep research treated bare PANTHER ids
as unauditable, but the repository PANTHER build settles this one. Requiring both TPS domains
plus a named subfamily is one of the better-constructed branches. No taxon constraint, which is
fine for a subfamily this narrow.
- number: 2
conditions:
- condition_type: INTERPRO
value: IPR008930
curie: InterPro:IPR008930
label: "Terpenoid cyclases/protein prenyltransferase alpha-alpha toroid"
interpro_type: HOMOLOGOUS_SUPERFAMILY
negated: false
- condition_type: INTERPRO
value: IPR008949
curie: InterPro:IPR008949
label: "Isoprenoid synthase domain superfamily"
interpro_type: HOMOLOGOUS_SUPERFAMILY
negated: false
- condition_type: TAXON
value: Ocimeae
curie: NCBITaxon:216719
label: "Ocimeae"
negated: false
notes: >-
KEEP WITH CAUTION. IPR008930 (terpenoid cyclase/protein prenyltransferase alpha-alpha toroid) +
IPR008949 (isoprenoid synthase domain superfamily) + Ocimeae. Both conditions are homologous-
superfamily entries, not families; the alpha-alpha toroid also covers squalene-hopene cyclase
and the protein prenyltransferase alpha subunit. The Ocimeae (Lamiaceae) restriction reflects a
real TPS expansion, so this is defensible, but two superfamily folds are weaker evidence than
the family-level branches elsewhere in this rule.
- number: 3
conditions:
- condition_type: INTERPRO
value: IPR034741
curie: InterPro:IPR034741
label: "Terpene cyclase-like 1, C-terminal domain"
interpro_type: DOMAIN
negated: false
- condition_type: INTERPRO
value: IPR036965
curie: InterPro:IPR036965
label: "Terpene synthase, N-terminal domain superfamily"
interpro_type: HOMOLOGOUS_SUPERFAMILY
negated: false
- condition_type: TAXON
value: Pinus
curie: NCBITaxon:3337
label: "Pinus"
negated: false
notes: >-
ON-TARGET. IPR034741 (terpene cyclase-like 1, C-terminal) + IPR036965 (TPS N-terminal
superfamily) + Pinus. Conifer terpene synthases are a genuine lineage-specific expansion and the
two-domain requirement matches plant class-I TPS architecture.
- number: 4
conditions:
- condition_type: PANTHER
value: PTHR31739:SF25
curie: PTHR31739:SF25
negated: false
- condition_type: TAXON
value: Eukaryota
curie: NCBITaxon:2759
label: "Eukaryota"
negated: false
notes: >-
ON-TARGET. PTHR31739:SF25 resolves to '(E,E)-GERANYLLINALOOL SYNTHASE' (panther.obo) - a
diterpene synthase. The Eukaryota taxon adds nothing at subfamily resolution but is harmless.
- number: 5
conditions:
- condition_type: INTERPRO
value: IPR002060
curie: InterPro:IPR002060
label: "Squalene/phytoene synthase"
interpro_type: FAMILY
negated: false
- condition_type: INTERPRO
value: IPR019845
curie: InterPro:IPR019845
label: "Squalene/phytoene synthase, conserved site"
interpro_type: CONSERVED_SITE
negated: false
- condition_type: PANTHER
value: PTHR31480
curie: PTHR31480
negated: false
notes: >-
ON-TARGET. IPR002060 (squalene/phytoene synthase) + IPR019845 (conserved site) + PTHR31480 =
'BIFUNCTIONAL LYCOPENE CYCLASE/PHYTOENE SYNTHASE'. Family plus conserved catalytic site plus
named PANTHER family. GO:0006720 is the right level here precisely because the squalene/phytoene
synthase fold cannot distinguish the sterol entry reaction from the carotenoid entry reaction
without subfamily evidence.
- number: 6
conditions:
- condition_type: INTERPRO
value: IPR001128
curie: InterPro:IPR001128
label: "Cytochrome P450"
interpro_type: FAMILY
negated: false
- condition_type: PANTHER
value: PTHR47955
curie: PTHR47955
negated: false
- condition_type: TAXON
value: Asterales
curie: NCBITaxon:4209
label: "Asterales"
negated: false
notes: >-
HOLD / LIKELY REMOVE. IPR001128 is the generic cytochrome P450 domain; PTHR47955 resolves to
'CYTOCHROME P450 FAMILY 71 PROTEIN' (panther.obo). CYP71 is the largest and most promiscuous
plant P450 family, acting on terpenoids, oximes, cyanogenic glucosides and glucosinolates.
Family membership plus an Asterales restriction does not establish isoprenoid pathway
participation. The deep research documents CYP706A3 with 29 substrates and CYP76M8 hydroxylating
seven diterpene skeletons as evidence that plant P450 family membership is not diagnostic.
- number: 7
conditions:
- condition_type: INTERPRO
value: IPR004294
curie: InterPro:IPR004294
label: "Carotenoid oxygenase"
interpro_type: FAMILY
negated: false
- condition_type: TAXON
value: Mammalia
curie: NCBITaxon:40674
label: "Mammalia"
negated: false
notes: >-
ON-TARGET, and empirically the single largest producer. IPR004294 (carotenoid oxygenase) +
Mammalia captures BCO1, BCO2 and RPE65. Carotenoid cleavage is carotenoid catabolism and
retinoid formation, both inside GO:0006720. NOTE: 62% of a random 300-protein sample of the
rule's emitted set carries IPR004294, but 89% of those are NOT mammals - so either this taxon
constraint is not enforced at annotation time or the published rule has diverged from the
annotating release (see analysis section 3).
- number: 8
conditions:
- condition_type: INTERPRO
value: IPR036396
curie: InterPro:IPR036396
label: "Cytochrome P450 superfamily"
interpro_type: HOMOLOGOUS_SUPERFAMILY
negated: false
- condition_type: PANTHER
value: PTHR47950:SF4
curie: PTHR47950:SF4
negated: false
- condition_type: TAXON
value: Nepetoideae
curie: NCBITaxon:216706
label: "Nepetoideae"
negated: false
notes: >-
ON-TARGET. IPR036396 (P450 superfamily) + PTHR47950:SF4 + Nepetoideae. The PANTHER subfamily
resolves to 'GERANIOL 8-HYDROXYLASE-LIKE' (panther.obo), i.e. the CYP76 monoterpenoid/iridoid-
pathway enzymes. Unlike CS6 this branch is pinned to a named subfamily rather than a whole P450
family, which is what makes it acceptable.
- number: 9
conditions:
- condition_type: INTERPRO
value: IPR002937
curie: InterPro:IPR002937
label: "Amine oxidase"
interpro_type: DOMAIN
negated: false
- condition_type: INTERPRO
value: IPR014105
curie: InterPro:IPR014105
label: "Carotenoid/retinoid oxidoreductase"
interpro_type: FAMILY
negated: false
- condition_type: INTERPRO
value: IPR036188
curie: InterPro:IPR036188
label: "FAD/NAD(P)-binding domain superfamily"
interpro_type: HOMOLOGOUS_SUPERFAMILY
negated: false
notes: >-
ON-TARGET but fold-heavy. IPR002937 (amine oxidase) + IPR014105 (carotenoid/retinoid
oxidoreductase) + IPR036188 (FAD/NAD(P)-binding superfamily). Only IPR014105 is diagnostic; the
other two are broad cofactor-binding folds contributing little. Retain on the strength of
IPR014105 alone, and consider dropping the two superfamily conditions rather than pretending
they add specificity.
- number: 10
conditions:
- condition_type: PANTHER
value: PTHR31225:SF98
curie: PTHR31225:SF98
negated: false
- condition_type: TAXON
value: Viridiplantae
curie: NCBITaxon:33090
label: "Viridiplantae"
negated: false
notes: >-
ON-TARGET. PTHR31225:SF98 resolves to 'TERPENE SYNTHASE 9-RELATED' (panther.obo). The deep
research listed this among 'ambiguous, unnamed' branches to hold; the PANTHER lookup resolves it
as a genuine TPS subfamily. Viridiplantae is appropriate.
- number: 11
conditions:
- condition_type: INTERPRO
value: IPR034686
curie: InterPro:IPR034686
label: "Terpene cyclase-like 2"
interpro_type: FAMILY
negated: false
- condition_type: TAXON
value: Amoebozoa
curie: NCBITaxon:554915
label: "Amoebozoa"
negated: false
notes: >-
ON-TARGET. IPR034686 (terpene cyclase-like 2) + Amoebozoa. Slime-mould terpene synthases are a
documented lineage-specific family; the taxon restriction is biologically motivated rather than
an annotation artifact.
- number: 12
conditions:
- condition_type: PANTHER
value: PTHR31225:SF9
curie: PTHR31225:SF9
negated: false
- condition_type: TAXON
value: Elsholtzieae
curie: NCBITaxon:216720
label: "Elsholtzieae"
negated: false
notes: >-
ON-TARGET. PTHR31225:SF9 resolves to 'TERPENE SYNTHASE 10' (panther.obo). Elsholtzieae is a real
Lamiaceae TPS expansion.
- number: 13
conditions:
- condition_type: INTERPRO
value: IPR017825
curie: InterPro:IPR017825
label: "Lycopene cyclase domain"
interpro_type: DOMAIN
negated: false
notes: >-
ON-TARGET. IPR017825 (lycopene cyclase domain) alone, no taxon constraint. Acceptable because
the domain is family-level and functionally dedicated - one of the few single-condition branches
in this rule that is safe.
- number: 14
conditions:
- condition_type: INTERPRO
value: IPR002213
curie: InterPro:IPR002213
label: "UDP-glucuronosyl/UDP-glucosyltransferase"
interpro_type: FAMILY
negated: false
- condition_type: TAXON
value: Caryophyllaceae
curie: NCBITaxon:3568
label: "Caryophyllaceae"
negated: false
notes: >-
OFF-TARGET - REMOVE. IPR002213 is the whole UDP-glucuronosyl/UDP-glucosyltransferase family:
plant UGTs glycosylate flavonoids, phenolics, hormones, xenobiotics and terpenoids
indiscriminately. A Caryophyllaceae restriction does not make the family diagnostic for
isoprenoid metabolism. If the intent was triterpenoid saponin glycosyltransferases, the branch
needs a specific subfamily, not a family plus a plant clade.
- number: 15
conditions:
- condition_type: PANTHER
value: PTHR10543:SF57
curie: PTHR10543:SF57
negated: false
- condition_type: TAXON
value: Metazoa
curie: NCBITaxon:33208
label: "Metazoa"
negated: false
notes: >-
ON-TARGET. PTHR10543:SF57 resolves to 'RETINOID ISOMEROHYDROLASE' (panther.obo) = RPE65. The
deep research could not audit this identifier and marked it for review; it is a named retinoid
enzyme and retinoid metabolic process (GO:0001523) is a descendant of GO:0006720.
- number: 16
conditions:
- condition_type: INTERPRO
value: IPR000092
curie: InterPro:IPR000092
label: "Polyprenyl synthetase-like"
interpro_type: FAMILY
negated: false
- condition_type: INTERPRO
value: IPR039702
curie: InterPro:IPR039702
label: "Farnesyl pyrophosphate synthase-like"
interpro_type: FAMILY
negated: false
- condition_type: TAXON
value: Ecdysozoa
curie: NCBITaxon:1206794
label: "Ecdysozoa"
negated: false
notes: >-
ON-TARGET. IPR000092 (polyprenyl synthetase-like) + IPR039702 (FPPS-like) + Ecdysozoa: farnesyl
diphosphate synthase, a core isoprenoid chain-elongation enzyme. NOTE: 30% of a random
300-protein sample of the emitted set carries IPR000092 but only 28% of those also carry
IPR039702, so the conjunction written here is not visible in the annotations the rule is
credited with (analysis section 3).
- number: 17
conditions:
- condition_type: INTERPRO
value: IPR014102
curie: InterPro:IPR014102
label: "Phytoene desaturase"
interpro_type: FAMILY
negated: false
- condition_type: INTERPRO
value: IPR050464
curie: InterPro:IPR050464
label: "Zeta Carotene Desaturase and Related Oxidoreductases"
interpro_type: FAMILY
negated: false
- condition_type: PANTHER
value: PTHR42923:SF45
curie: PTHR42923:SF45
negated: false
notes: >-
ON-TARGET. IPR014102 (phytoene desaturase) + IPR050464 + PTHR42923:SF45 = '15-CIS-PHYTOENE
DESATURASE, CHLOROPLASTIC_CHROMOPLASTIC' (panther.obo). Carotenoid backbone desaturation.
- number: 18
conditions:
- condition_type: INTERPRO
value: IPR010108
curie: InterPro:IPR010108
label: "Lycopene cyclase, beta/epsilon"
interpro_type: FAMILY
negated: false
- condition_type: TAXON
value: Streptophyta
curie: NCBITaxon:35493
label: "Streptophyta"
negated: false
notes: >-
ON-TARGET. IPR010108 (lycopene cyclase, beta/epsilon) + Streptophyta. Family-level and pathway-
dedicated.
- number: 19
conditions:
- condition_type: PANTHER
value: PTHR31225:SF120
curie: PTHR31225:SF120
negated: false
- condition_type: TAXON
value: Embryophyta
curie: NCBITaxon:3193
label: "Embryophyta"
negated: false
notes: >-
ON-TARGET. PTHR31225:SF120 resolves to 'GERMACRENE-A SYNTHASE' (panther.obo), a sesquiterpene
synthase. Embryophyta is appropriate.
- number: 20
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000005
curie: CATH.FunFam:1.10.600.10:FF:000005
label: "Ent-kaur-16-ene synthase, chloroplastic"
negated: false
- condition_type: FUNFAM
value: 1.50.10.130:FF:000004
curie: CATH.FunFam:1.50.10.130:FF:000004
label: "Carene synthase, chloroplastic"
negated: false
- condition_type: TAXON
value: Eukaryota
curie: NCBITaxon:2759
label: "Eukaryota"
negated: false
notes: >-
ON-TARGET, and the deep research's objection to it is mistaken. It flagged the conjunction of
FunFam 1.10.600.10:FF:000005 ('ent-kaur-16-ene synthase') AND 1.50.10.130:FF:000004 ('carene
synthase') as possibly 'internally contradictory or yield[ing] no proteins'. The two FunFams sit
in different CATH superfamilies - 1.10.600.10 is the isoprenoid-synthase alpha-helical fold,
1.50.10.130 the terpene-cyclase beta-gamma fold - and plant class-I diterpene synthases carry
both. This is a domain-architecture requirement, and a well-constructed one.
- number: 21
conditions:
- condition_type: FUNFAM
value: 1.50.10.130:FF:000002
curie: CATH.FunFam:1.50.10.130:FF:000002
label: "Ent-copalyl diphosphate synthase, chloroplastic"
negated: false
- condition_type: TAXON
value: Nepetoideae
curie: NCBITaxon:216706
label: "Nepetoideae"
negated: false
notes: >-
ON-TARGET. FunFam 1.50.10.130:FF:000002 (ent-copalyl diphosphate synthase) + Nepetoideae. Class-
II diterpene cyclase, the committed step to labdane-type diterpenoids.
- number: 22
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000007
curie: CATH.FunFam:1.10.600.10:FF:000007
label: "Isoprene synthase, chloroplastic"
negated: false
- condition_type: TAXON
value: Solanales
curie: NCBITaxon:4069
label: "Solanales"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000007 (isoprene synthase, chloroplastic) + Solanales. Named,
narrow FunFam.
- number: 23
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000043
curie: CATH.FunFam:1.10.630.10:FF:000043
label: "Cytochrome P450 99A2"
negated: false
- condition_type: TAXON
value: campanulids
curie: NCBITaxon:91882
label: "campanulids"
negated: false
notes: >-
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000043 is labelled 'Cytochrome P450 99A2'
but the branch is restricted to campanulids. CYP99A2 is a rice (Poales) momilactone-pathway
diterpenoid P450; campanulids contain no rice. Either the FunFam label is being read as a
function rather than as the name of its best-studied member, or the taxon is wrong. The branch
cannot be signed off until the actual campanulid members are inspected.
- number: 24
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000004
curie: CATH.FunFam:1.10.600.10:FF:000004
label: "Phytoene synthase chloroplastic"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000004 (phytoene synthase, chloroplastic). No taxon constraint,
acceptable for a named pathway-dedicated FunFam.
- number: 25
conditions:
- condition_type: FUNFAM
value: 1.50.10.160:FF:000001
curie: CATH.FunFam:1.50.10.160:FF:000001
label: "Ent-copalyl diphosphate synthase"
negated: false
- condition_type: TAXON
value: Viridiplantae
curie: NCBITaxon:33090
label: "Viridiplantae"
negated: false
notes: >-
ON-TARGET. FunFam 1.50.10.160:FF:000001 (ent-copalyl diphosphate synthase) + Viridiplantae.
Overlaps CS21 in concept; see condition_overlap.
- number: 26
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000007
curie: CATH.FunFam:1.10.630.10:FF:000007
label: "Cytochrome P450 76C4"
negated: false
- condition_type: TAXON
value: Mentheae
curie: NCBITaxon:216718
label: "Mentheae"
negated: false
notes: >-
KEEP WITH REVIEW. FunFam 1.10.630.10:FF:000007 ('Cytochrome P450 76C4') + Mentheae.
CYP76C-subfamily enzymes include monoterpenol oxidases in Lamiaceae, so the pairing of subfamily
and clade is coherent - but CYP76C1 in Arabidopsis also acts outside isoprenoid metabolism, and
this is still a P450 FunFam. Weaker than CS8, which uses a PANTHER subfamily named for the
reaction.
- number: 27
conditions:
- condition_type: FUNFAM
value: 1.50.10.20:FF:000011
curie: CATH.FunFam:1.50.10.20:FF:000011
label: "Terpene cyclase/mutase family member"
negated: false
- condition_type: TAXON
value: fabids
curie: NCBITaxon:91835
label: "fabids"
negated: false
notes: >-
ON-TARGET. FunFam 1.50.10.20:FF:000011 ('Terpene cyclase/mutase family member') + fabids. CATH
1.50.10.20 is the squalene-hopene cyclase / oxidosqualene cyclase superfamily - triterpene
cyclization, squarely isoprenoid.
- number: 28
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000003
curie: CATH.FunFam:2.60.120.330:FF:000003
label: "Gibberellin 20 oxidase 2"
negated: false
- condition_type: TAXON
value: Streptophyta
curie: NCBITaxon:35493
label: "Streptophyta"
negated: false
notes: >-
ON-TARGET. FunFam 2.60.120.330:FF:000003 (gibberellin 20-oxidase 2) + Streptophyta. GAs are
diterpenoids; GO:0009686 gibberellin biosynthetic process is a descendant of GO:0006720.
- number: 29
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000013
curie: CATH.FunFam:2.60.120.330:FF:000013
label: "Gibberellin 3-beta-dioxygenase 1"
negated: false
- condition_type: TAXON
value: Embryophyta
curie: NCBITaxon:3193
label: "Embryophyta"
negated: false
notes: >-
ON-TARGET. FunFam 2.60.120.330:FF:000013 (gibberellin 3-beta-dioxygenase 1) + Embryophyta.
- number: 30
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000020
curie: CATH.FunFam:1.10.600.10:FF:000020
label: "Phytoene synthase"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000020 (phytoene synthase). Duplicates CS24's concept with a
different FunFam.
- number: 31
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000014
curie: CATH.FunFam:2.60.120.330:FF:000014
label: "Gibberellin 2-beta-dioxygenase 1"
negated: false
- condition_type: TAXON
value: Tracheophyta
curie: NCBITaxon:58023
label: "Tracheophyta"
negated: false
notes: >-
ON-TARGET. FunFam 2.60.120.330:FF:000014 (gibberellin 2-beta-dioxygenase 1) + Tracheophyta. GA
deactivation is catabolism, still 'metabolic process'.
- number: 32
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000091
curie: CATH.FunFam:3.50.50.60:FF:000091
label: "15-cis-phytoene desaturase, chloroplastic/chromoplastic"
negated: false
- condition_type: TAXON
value: Spermatophyta
curie: NCBITaxon:58024
label: "Spermatophyta"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000091 (15-cis-phytoene desaturase, chloroplastic/chromoplastic)
+ Spermatophyta.
- number: 33
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000101
curie: CATH.FunFam:3.50.50.60:FF:000101
label: "lycopene epsilon cyclase, chloroplastic"
negated: false
- condition_type: TAXON
value: Magnoliopsida
curie: NCBITaxon:3398
label: "Magnoliopsida"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000101 (lycopene epsilon cyclase, chloroplastic) +
Magnoliopsida.
- number: 34
conditions:
- condition_type: FUNFAM
value: 1.10.3270.10:FF:000002
curie: CATH.FunFam:1.10.3270.10:FF:000002
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
negated: false
- condition_type: FUNFAM
value: 3.30.70.420:FF:000001
curie: CATH.FunFam:3.30.70.420:FF:000001
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
negated: false
- condition_type: FUNFAM
value: 3.90.770.10:FF:000001
curie: CATH.FunFam:3.90.770.10:FF:000001
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
negated: false
notes: >-
ON-TARGET - one of the strongest branches. Three HMG-CoA reductase FunFams
(1.10.3270.10:FF:000002, 3.30.70.420:FF:000001, 3.90.770.10:FF:000001) required together,
matching HMGR's three-domain architecture. The rate-limiting enzyme of the mevalonate pathway.
No taxon constraint, correctly.
- number: 35
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000008
curie: CATH.FunFam:1.10.630.10:FF:000008
label: "Cytochrome P450 71D8"
negated: false
- condition_type: TAXON
value: Poales
curie: NCBITaxon:38820
label: "Poales"
negated: false
notes: >-
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000008 is labelled 'Cytochrome P450 71D8'
but the branch is restricted to Poales. CYP71D8 is a legume enzyme. As with CS23 and CS75, a
P450 FunFam label is being treated as a function. CYP71D does contain terpenoid hydroxylases
(e.g. CYP71D13/18 in mint) but also many non-isoprenoid enzymes; the Poales members must be
inspected before this branch can be kept.
- number: 36
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000041
curie: CATH.FunFam:1.10.630.10:FF:000041
label: "Cytochrome P450 26A1 isoform 1"
negated: false
- condition_type: TAXON
value: Metazoa
curie: NCBITaxon:33208
label: "Metazoa"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.630.10:FF:000041 (cytochrome P450 26A1) + Metazoa. CYP26A1 is the
principal retinoic acid 4-hydroxylase; retinoid metabolic process is under GO:0006720. Metazoa
is if anything narrower than needed.
- number: 37
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000299
curie: CATH.FunFam:1.10.630.10:FF:000299
label: "Cytochrome P450 2C9"
negated: false
- condition_type: TAXON
value: Primates
curie: NCBITaxon:9443
label: "Primates"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000299 ('Cytochrome P450 2C9') + Primates. CYP2C9 is
a hepatic drug- and xenobiotic-metabolizing P450; its endogenous substrates are chiefly
arachidonic acid and related eicosanoids, not isoprenoids. The deep research lists this among
the branches to remove outright.
- number: 38
conditions:
- condition_type: FUNFAM
value: 1.50.10.20:FF:000064
curie: CATH.FunFam:1.50.10.20:FF:000064
label: "Uncharacterized protein"
negated: false
- condition_type: TAXON
value: Saxifragales
curie: NCBITaxon:41946
label: "Saxifragales"
negated: false
notes: >-
HOLD, but not for the reason the deep research gave. FunFam 1.50.10.20:FF:000064 is labelled
'Uncharacterized protein' + Saxifragales, and the deep research recommended deleting all unnamed
FunFams. That blanket rule is wrong here: CATH 1.50.10.20 is the terpene cyclase/mutase
(SHC/OSC) superfamily, which is functionally dedicated to isoprenoid cyclization, so the fold
itself carries information even though the FunFam is unnamed. Inspect the Saxifragales members;
this is a likely keep, not a likely delete.
- number: 39
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000808
curie: CATH.FunFam:3.40.50.720:FF:000808
label: "Iridoid synthase"
negated: false
- condition_type: TAXON
value: asterids
curie: NCBITaxon:71274
label: "asterids"
negated: false
notes: >-
MIXED - the rule's one systematic biological false positive. FunFam 3.40.50.720:FF:000808
('Iridoid synthase') + asterids is the PRISE family (progesterone 5-beta-reductase / iridoid
synthase enzymes). Its iridoid synthase members are correctly isoprenoid (monoterpenoid), but
its progesterone 5-beta-reductase and 3-oxo-Delta(4,5)-steroid 5-beta-reductase members are
steroid enzymes, and GO:0008202 steroid metabolic process is NOT a descendant of GO:0006720
(QuickGO ancestor closure, 2026-08-22). Empirically the steroid side outnumbers the iridoid side
in this rule's output: 1.5% vs 0.9% of 8,974 annotations. Needs splitting to the iridoid
synthase members only.
- number: 40
conditions:
- condition_type: FUNFAM
value: 1.50.10.130:FF:000006
curie: CATH.FunFam:1.50.10.130:FF:000006
label: "Terpene synthase 7"
negated: false
notes: >-
ON-TARGET. FunFam 1.50.10.130:FF:000006 ('Terpene synthase 7'). Named TPS FunFam, no taxon
constraint.
- number: 41
conditions:
- condition_type: FUNFAM
value: 3.30.230.10:FF:000018
curie: CATH.FunFam:3.30.230.10:FF:000018
label: "Diphosphomevalonate decarboxylase"
negated: false
- condition_type: FUNFAM
value: 3.30.70.890:FF:000005
curie: CATH.FunFam:3.30.70.890:FF:000005
label: "Diphosphomevalonate decarboxylase"
negated: false
- condition_type: TAXON
value: Mammalia
curie: NCBITaxon:40674
label: "Mammalia"
negated: false
notes: >-
ON-TARGET. Diphosphomevalonate decarboxylase FunFams (3.30.230.10:FF:000018 +
3.30.70.890:FF:000005) + Mammalia. Core mevalonate pathway. The Mammalia restriction is
unnecessarily narrow - the MVA pathway is not mammal-specific.
- number: 42
conditions:
- condition_type: FUNFAM
value: 3.30.230.10:FF:000119
curie: CATH.FunFam:3.30.230.10:FF:000119
label: "Mevalonate kinase"
negated: false
- condition_type: FUNFAM
value: 3.30.70.890:FF:000003
curie: CATH.FunFam:3.30.70.890:FF:000003
label: "Mevalonate kinase"
negated: false
- condition_type: TAXON
value: Chordata
curie: NCBITaxon:7711
label: "Chordata"
negated: false
notes: >-
ON-TARGET. Mevalonate kinase FunFams (3.30.230.10:FF:000119 + 3.30.70.890:FF:000003) + Chordata.
Core mevalonate pathway; taxon again narrower than the biology.
- number: 43
conditions:
- condition_type: FUNFAM
value: 3.40.309.10:FF:000001
curie: CATH.FunFam:3.40.309.10:FF:000001
label: "Mitochondrial aldehyde dehydrogenase 2"
negated: false
- condition_type: FUNFAM
value: 3.40.605.10:FF:000026
curie: CATH.FunFam:3.40.605.10:FF:000026
label: "Aldehyde dehydrogenase, putative"
negated: false
- condition_type: FUNFAM
value: 3.40.605.10:FF:000054
curie: CATH.FunFam:3.40.605.10:FF:000054
label: "Aldehyde dehydrogenase family 1 member A3"
negated: false
notes: >-
REMOVE (as an unsatisfiable branch, not as an off-target one). Conjunction of FunFam
3.40.309.10:FF:000001 ('Mitochondrial aldehyde dehydrogenase 2'), 3.40.605.10:FF:000026
('Aldehyde dehydrogenase, putative') and 3.40.605.10:FF:000054 ('Aldehyde dehydrogenase
family 1 member A3'). The three FunFams are AND-ed, so - as for CS20 and CS61 - the broad
conjuncts do not widen the match set and the most specific conjunct binds it: any protein
this fires on is a subset of ALDH1A3-like proteins, whose retinaldehyde-dehydrogenase
activity is legitimately within GO:0006720 via GO:0001523. The real defect is
satisfiability rather than aim. 3.40.309.10 and 3.40.605.10 are the two ALDH-fold CATH
superfamilies, and while requiring one FunFam from each is a sound domain-architecture
requirement, this branch additionally requires two distinct FunFams (FF:000026 and
FF:000054) from within the same superfamily 3.40.605.10 - which a single ALDH catalytic
domain cannot satisfy. CS43 is therefore near-certainly a dead branch that matches nothing,
which makes it further evidence for the section 3 finding that the emitted set cannot be
reproduced from the published condition sets. No taxon constraint, so were it satisfiable
it would fire across all of UniProt.
- number: 44
conditions:
- condition_type: FUNFAM
value: 3.40.50.1820:FF:000110
curie: CATH.FunFam:3.40.50.1820:FF:000110
label: "Hormone-sensitive lipase"
negated: false
- condition_type: FUNFAM
value: 3.40.50.1820:FF:000199
curie: CATH.FunFam:3.40.50.1820:FF:000199
label: "Hormone-sensitive lipase"
negated: false
- condition_type: TAXON
value: Craniata
curie: NCBITaxon:89593
label: "Craniata"
negated: false
notes: >-
OFF-TARGET - REMOVE. Hormone-sensitive lipase FunFams (3.40.50.1820:FF:000110, :FF:000199) +
Craniata. LIPE does have retinyl ester hydrolase activity in vitro, but its established
biological role is neutral-lipid (triacylglycerol, cholesteryl ester) hydrolysis in adipocytes.
A side activity on a retinyl ester is not grounds for a blanket isoprenoid metabolic process
annotation.
- number: 45
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000001
curie: CATH.FunFam:1.10.600.10:FF:000001
label: "Geranylgeranyl diphosphate synthase"
negated: false
- condition_type: TAXON
value: Bacteria
curie: NCBITaxon:2
label: "Bacteria"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000001 (geranylgeranyl diphosphate synthase) + Bacteria. Prenyl
chain assembly.
- number: 46
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000019
curie: CATH.FunFam:1.10.600.10:FF:000019
label: "2-methylisoborneol synthase"
negated: false
- condition_type: TAXON
value: Bacillati
curie: NCBITaxon:1783272
label: "Bacillati"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000019 (2-methylisoborneol synthase) + Bacillati. A genuine
actinobacterial monoterpene synthase; the taxon restriction is biologically motivated.
- number: 47
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000047
curie: CATH.FunFam:1.10.600.10:FF:000047
label: "Terpene synthase"
negated: false
- condition_type: TAXON
value: Amoebozoa
curie: NCBITaxon:554915
label: "Amoebozoa"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000047 ('Terpene synthase') + Amoebozoa. Overlaps CS70 and CS71
in concept.
- number: 48
conditions:
- condition_type: FUNFAM
value: 3.40.50.150:FF:000183
curie: CATH.FunFam:3.40.50.150:FF:000183
label: "Geranyl diphosphate 2-C-methyltransferase"
negated: false
- condition_type: TAXON
value: Actinomycetota
curie: NCBITaxon:201174
label: "Actinomycetota"
negated: false
notes: >-
ON-TARGET. FunFam 3.40.50.150:FF:000183 (geranyl diphosphate 2-C-methyltransferase) +
Actinomycetota. Named enzyme acting directly on a prenyl diphosphate.
- number: 49
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000084
curie: CATH.FunFam:3.40.50.720:FF:000084
label: "Short-chain dehydrogenase reductase"
negated: false
- condition_type: TAXON
value: Lamiaceae
curie: NCBITaxon:4136
label: "Lamiaceae"
negated: false
notes: >-
HOLD. FunFam 3.40.50.720:FF:000084 ('Short-chain dehydrogenase reductase') + Lamiaceae. CATH
3.40.50.720 is the Rossmann NAD(P)-binding superfamily, the single most promiscuous fold in this
rule. The FunFam label is generic. If the intent was Lamiaceae monoterpene reductases (e.g.
menthone reductases), that needs to be established from the members, not assumed from the clade.
- number: 50
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000145
curie: CATH.FunFam:3.40.50.720:FF:000145
label: "Retinol dehydrogenase 12"
negated: false
- condition_type: TAXON
value: Haplorrhini
curie: NCBITaxon:376913
label: "Haplorrhini"
negated: false
notes: >-
ON-TARGET, taxon far too narrow. FunFam 3.40.50.720:FF:000145 (retinol dehydrogenase 12) +
Haplorrhini. RDH12 is a genuine retinoid enzyme, but restricting it to Haplorrhini is an
annotation-availability artifact - RDH12 orthologues are conserved across vertebrates.
- number: 51
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:001857
curie: CATH.FunFam:3.40.50.720:FF:001857
label: "Alcohol dehydrogenase class 4 mu/sigma chain"
negated: false
- condition_type: TAXON
value: Vertebrata
curie: NCBITaxon:7742
label: "Vertebrata"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 3.40.50.720:FF:001857 ('Alcohol dehydrogenase class 4 mu/sigma
chain') + Vertebrata = ADH7. ADH7 does oxidize retinol, but it is principally an
ethanol/aliphatic-alcohol dehydrogenase of upper-aerodigestive mucosa, and a generic Rossmann
FunFam label of 'alcohol dehydrogenase class 4' is not a proxy for retinoid metabolism.
- number: 52
conditions:
- condition_type: FUNFAM
value: 3.40.50.920:FF:000002
curie: CATH.FunFam:3.40.50.920:FF:000002
label: "1-deoxy-D-xylulose-5-phosphate synthase"
negated: false
- condition_type: FUNFAM
value: 3.40.50.970:FF:000005
curie: CATH.FunFam:3.40.50.970:FF:000005
label: "1-deoxy-D-xylulose-5-phosphate synthase"
negated: false
notes: >-
ON-TARGET. 1-deoxy-D-xylulose-5-phosphate synthase FunFams (3.40.50.920:FF:000002 +
3.40.50.970:FF:000005). DXS is the entry enzyme of the MEP pathway - one of the most diagnostic
possible matches for GO:0006720.
- number: 53
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000378
curie: CATH.FunFam:3.50.50.60:FF:000378
label: "Phytoene desaturase"
negated: false
- condition_type: TAXON
value: Pseudomonadati
curie: NCBITaxon:3379134
label: "Pseudomonadati"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000378 (phytoene desaturase) + Pseudomonadati.
- number: 54
conditions:
- condition_type: FUNFAM
value: 3.90.1720.10:FF:000006
curie: CATH.FunFam:3.90.1720.10:FF:000006
label: "Lecithin retinol acyltransferase"
negated: false
- condition_type: TAXON
value: Euteleostomi
curie: NCBITaxon:117571
label: "Euteleostomi"
negated: false
notes: >-
ON-TARGET. FunFam 3.90.1720.10:FF:000006 (lecithin retinol acyltransferase) + Euteleostomi. LRAT
esterifies retinol for storage - retinoid metabolism.
- number: 55
conditions:
- condition_type: FUNFAM
value: 3.90.180.10:FF:000001
curie: CATH.FunFam:3.90.180.10:FF:000001
label: "S-(hydroxymethyl)glutathione dehydrogenase"
negated: false
- condition_type: TAXON
value: Mus
curie: NCBITaxon:10088
label: "Mus"
negated: false
notes: >-
OFF-TARGET - REMOVE, and the clearest single error in the rule. FunFam 3.90.180.10:FF:000001
('S-(hydroxymethyl)glutathione dehydrogenase') restricted to Mus. This is ADH5/class-III ADH,
the formaldehyde/S-nitrosoglutathione-detoxifying enzyme. It does not act on retinol and has no
isoprenoid role. A genus-level (Mus) restriction on a universally conserved housekeeping enzyme
is a textbook association-rule mining artifact.
- number: 56
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000036
curie: CATH.FunFam:1.10.600.10:FF:000036
label: "cis-abienol synthase, chloroplastic"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000036 (cis-abienol synthase, chloroplastic). Named diterpene
synthase.
- number: 57
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000052
curie: CATH.FunFam:1.10.630.10:FF:000052
label: "Ent-kaurenoic acid oxidase"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.630.10:FF:000052 (ent-kaurenoic acid oxidase). KAO catalyses three
sequential oxidations converting ent-kaurenoic acid to GA12 - a committed gibberellin-pathway
step.
- number: 58
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000062
curie: CATH.FunFam:1.10.630.10:FF:000062
label: "Ent-kaurene oxidase 2"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.630.10:FF:000062 (ent-kaurene oxidase 2). KO, the preceding gibberellin-
pathway step.
- number: 59
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000171
curie: CATH.FunFam:3.50.50.60:FF:000171
label: "zeta-carotene-forming phytoene desaturase"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000171 (zeta-carotene-forming phytoene desaturase).
- number: 60
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000413
curie: CATH.FunFam:3.50.50.60:FF:000413
label: "Phytoene desaturase (lycopene-forming)"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000413 (phytoene desaturase, lycopene-forming).
- number: 61
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000042
curie: CATH.FunFam:1.10.600.10:FF:000042
label: "Probable terpene synthase 3, chloroplastic"
negated: false
- condition_type: FUNFAM
value: 1.50.10.130:FF:000005
curie: CATH.FunFam:1.50.10.130:FF:000005
label: "S-(+)-linalool synthase, chloroplastic"
negated: false
- condition_type: TAXON
value: Poaceae
curie: NCBITaxon:4479
label: "Poaceae"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000042 ('Probable terpene synthase 3') + 1.50.10.130:FF:000005
('S-(+)-linalool synthase') + Poaceae. Same two-superfamily domain-architecture logic as CS20,
and equally sound.
- number: 62
conditions:
- condition_type: FUNFAM
value: 1.50.10.20:FF:000002
curie: CATH.FunFam:1.50.10.20:FF:000002
label: "Terpene cyclase/mutase family member"
negated: false
- condition_type: FUNFAM
value: 1.50.10.20:FF:000022
curie: CATH.FunFam:1.50.10.20:FF:000022
label: "Terpene cyclase/mutase family member"
negated: false
- condition_type: TAXON
value: Fagales
curie: NCBITaxon:3502
label: "Fagales"
negated: false
notes: >-
ON-TARGET. FunFams 1.50.10.20:FF:000002 + :FF:000022 ('Terpene cyclase/mutase family member') +
Fagales. Triterpene cyclase.
- number: 63
conditions:
- condition_type: FUNFAM
value: 2.10.25.10:FF:000009
curie: CATH.FunFam:2.10.25.10:FF:000009
label: "Low-density lipoprotein receptor isoform 1"
negated: false
- condition_type: FUNFAM
value: 4.10.400.10:FF:000011
curie: CATH.FunFam:4.10.400.10:FF:000011
label: "Low-density lipoprotein receptor-related protein 1"
negated: false
- condition_type: TAXON
value: Catarrhini
curie: NCBITaxon:9526
label: "Catarrhini"
negated: false
notes: >-
OFF-TARGET - REMOVE, wrong on two independent grounds. FunFam 2.10.25.10:FF:000009 ('Low-density
lipoprotein receptor') + 4.10.400.10:FF:000011 ('LRP1') + Catarrhini. (1) LDLR and LRP1 are
endocytic receptors, not metabolic enzymes - this conflates transport of a lipid particle with
metabolism of its contents. (2) Even the intended inference fails: the cargo is cholesterol, and
GO:0008202 steroid metabolic process is not a descendant of GO:0006720. CATH 2.10.25.10 is also
the EGF/laminin superfamily, one of the most promiscuous small structural domains in existence.
- number: 64
conditions:
- condition_type: FUNFAM
value: 3.40.1090.10:FF:000003
curie: CATH.FunFam:3.40.1090.10:FF:000003
label: "Patatin-like phospholipase domain-containing protein 2"
negated: false
- condition_type: FUNFAM
value: 3.40.1090.10:FF:000021
curie: CATH.FunFam:3.40.1090.10:FF:000021
label: "Patatin-like phospholipase domain containing 2"
negated: false
- condition_type: TAXON
value: Eutheria
curie: NCBITaxon:9347
label: "Eutheria"
negated: false
notes: >-
OFF-TARGET - REMOVE. Patatin-like phospholipase domain-containing protein 2 FunFams
(3.40.1090.10:FF:000003, :FF:000021) + Eutheria = PNPLA2/ATGL. ATGL's established role is
triacylglycerol lipolysis; its reported retinyl ester hydrolase activity is a side activity and
cannot support a blanket isoprenoid annotation. Same failure mode as CS44.
- number: 65
conditions:
- condition_type: FUNFAM
value: 3.40.309.10:FF:000021
curie: CATH.FunFam:3.40.309.10:FF:000021
label: "Aldehyde dehydrogenase family 8 member A1"
negated: false
- condition_type: FUNFAM
value: 3.40.605.10:FF:000001
curie: CATH.FunFam:3.40.605.10:FF:000001
label: "Aldehyde dehydrogenase 1"
negated: false
- condition_type: TAXON
value: Euarchontoglires
curie: NCBITaxon:314146
label: "Euarchontoglires"
negated: false
notes: >-
HOLD, but the case for retaining it is stronger than first stated. FunFam
3.40.309.10:FF:000021 ('Aldehyde dehydrogenase family 8 member A1') + 3.40.605.10:FF:000001
('Aldehyde dehydrogenase 1') + Euarchontoglires. ALDH8A1 does have 9-cis-retinal
dehydrogenase activity and ALDH1A1 is a retinal dehydrogenase, so the intent is legitimate.
The earlier objection - that 'Aldehyde dehydrogenase 1' is too broad to exclude the wider
ALDH superfamily - misread the conjunction as a disjunction. The two FunFams are AND-ed and
sit in the two distinct ALDH-fold CATH superfamilies (3.40.309.10 and 3.40.605.10), so this
is the same sound domain-architecture pattern accepted for CS20 and CS61; the broad conjunct
cannot widen the match set, and the specific ALDH8A1 conjunct binds it. Unlike CS43 the
branch is satisfiable, since it draws one FunFam from each superfamily rather than two from
one. HOLD is retained only pending a member census to confirm the matched set is in fact
ALDH8A1-like, not because the conjunction is unsound.
- number: 66
conditions:
- condition_type: FUNFAM
value: 1.10.274.20:FF:000003
curie: CATH.FunFam:1.10.274.20:FF:000003
label: "Phenylalanine aminomutase (L-beta-phenylalanine forming)"
negated: false
- condition_type: TAXON
value: Pinopsida
curie: NCBITaxon:58019
label: "Pinopsida"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 1.10.274.20:FF:000003 ('Phenylalanine aminomutase (L-beta-
phenylalanine forming)') + Pinopsida. PAM is an MIO-dependent aminomutase of amino-acid
specialized metabolism. It supplies the phenylisoserine side chain of paclitaxel, which is why
it appears in a conifer context - but the enzyme's substrate and product are amino acids, not
isoprenoids. Supplying a non-isoprenoid moiety to a taxane is not isoprenoid metabolism.
- number: 67
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000008
curie: CATH.FunFam:1.10.600.10:FF:000008
label: "Farnesyl pyrophosphate synthase"
negated: false
- condition_type: TAXON
value: rosids
curie: NCBITaxon:71275
label: "rosids"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000008 (farnesyl pyrophosphate synthase) + rosids.
- number: 68
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000018
curie: CATH.FunFam:1.10.600.10:FF:000018
label: "Probable geranylgeranyl-diphosphate geranylgeranyltransferase (AL-2)"
negated: false
- condition_type: TAXON
value: Fungi
curie: NCBITaxon:4751
label: "Fungi"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000018 ('Probable geranylgeranyl-diphosphate
geranylgeranyltransferase (AL-2)') + Fungi. AL-2 is the Neurospora phytoene synthase/GGPP-
transferase of the carotenoid pathway.
- number: 69
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000021
curie: CATH.FunFam:1.10.600.10:FF:000021
label: "Farnesyl pyrophosphate synthase"
negated: false
- condition_type: TAXON
value: Ecdysozoa
curie: NCBITaxon:1206794
label: "Ecdysozoa"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000021 (farnesyl pyrophosphate synthase) + Ecdysozoa. Largely
redundant with CS16, which reaches the same enzymes in the same clade via InterPro.
- number: 70
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000076
curie: CATH.FunFam:1.10.600.10:FF:000076
label: "Terpene synthase"
negated: false
- condition_type: TAXON
value: Evosea
curie: NCBITaxon:2605435
label: "Evosea"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000076 ('Terpene synthase') + Evosea. Redundant with CS47 and
CS71: Evosea and Eumycetozoa are both inside Amoebozoa.
- number: 71
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000078
curie: CATH.FunFam:1.10.600.10:FF:000078
label: "Terpene synthase"
negated: false
- condition_type: TAXON
value: Eumycetozoa
curie: NCBITaxon:142796
label: "Eumycetozoa"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.600.10:FF:000078 ('Terpene synthase') + Eumycetozoa. See CS70.
- number: 72
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000009
curie: CATH.FunFam:1.10.630.10:FF:000009
label: "Cytochrome P450 26B1 isoform 1"
negated: false
- condition_type: TAXON
value: Hominidae
curie: NCBITaxon:9604
label: "Hominidae"
negated: false
notes: >-
ON-TARGET, taxon absurdly narrow. FunFam 1.10.630.10:FF:000009 (cytochrome P450 26B1) +
Hominidae. CYP26B1 is a retinoic acid hydroxylase conserved across vertebrates; a Hominidae
restriction is an annotation artifact, not biology.
- number: 73
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000011
curie: CATH.FunFam:1.10.630.10:FF:000011
label: "Cytochrome P450 83B1"
negated: false
- condition_type: TAXON
value: Asterales
curie: NCBITaxon:4209
label: "Asterales"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000011 ('Cytochrome P450 83B1') + Asterales. CYP83B1
(Arabidopsis SUR2) oxidizes indole-3-acetaldoxime in indole glucosinolate biosynthesis.
Glucosinolates are amino-acid-derived, not isoprenoid. Asterales do not make glucosinolates, so
the taxon is also inconsistent with the label - the same name/taxon contradiction seen in CS23,
CS35 and CS75.
- number: 74
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000014
curie: CATH.FunFam:1.10.630.10:FF:000014
label: "Abscisic acid 8"
negated: false
- condition_type: TAXON
value: lamiids
curie: NCBITaxon:91888
label: "lamiids"
negated: false
notes: >-
ON-TARGET, label needs fixing. FunFam 1.10.630.10:FF:000014 + lamiids. The label 'Abscisic acid
8' is a truncation of 'abscisic acid 8'-hydroxylase' (CYP707A), the principal ABA catabolic
enzyme. ABA is an apocarotenoid, so ABA catabolism is inside GO:0006720. The truncated label
should be corrected before the branch is relied on, since a label that stops mid-word is exactly
the kind of thing that hides a wrong FunFam id.
- number: 75
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000022
curie: CATH.FunFam:1.10.630.10:FF:000022
label: "Taxadiene 5-alpha hydroxylase"
negated: false
- condition_type: TAXON
value: Caryophyllales
curie: NCBITaxon:3524
label: "Caryophyllales"
negated: false
notes: >-
HOLD - NAME/TAXON CONTRADICTION. FunFam 1.10.630.10:FF:000022 is labelled 'Taxadiene 5-alpha
hydroxylase' (CYP725A4, a Taxus/Pinopsida taxane P450) but the branch is restricted to
Caryophyllales, which contains no Taxus. The named enzyme would be on-target; the members
actually captured in Caryophyllales are unknown and must be inspected.
- number: 76
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000080
curie: CATH.FunFam:1.10.630.10:FF:000080
label: "Carotene epsilon-monooxygenase, chloroplastic"
negated: false
- condition_type: TAXON
value: eudicotyledons
curie: NCBITaxon:71240
label: "eudicotyledons"
negated: false
notes: >-
ON-TARGET. FunFam 1.10.630.10:FF:000080 ('Carotene epsilon-monooxygenase, chloroplastic') +
eudicotyledons = LUT1/CYP97C, the carotenoid epsilon-ring hydroxylase of the lutein pathway.
- number: 77
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000097
curie: CATH.FunFam:1.10.630.10:FF:000097
label: "Cytochrome P-450 19"
negated: false
- condition_type: TAXON
value: PACMAD clade
curie: NCBITaxon:147370
label: "PACMAD clade"
negated: false
notes: >-
HOLD. FunFam 1.10.630.10:FF:000097 ('Cytochrome P-450 19') + PACMAD clade. 'Cytochrome P-450 19'
is ambiguous - in animals CYP19 is aromatase, in plants the numbering is unrelated. A P450
FunFam plus a grass clade is not diagnostic without inspecting members.
- number: 78
conditions:
- condition_type: FUNFAM
value: 1.10.630.10:FF:000182
curie: CATH.FunFam:1.10.630.10:FF:000182
negated: false
- condition_type: TAXON
value: Homo
curie: NCBITaxon:9605
label: "Homo"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 1.10.630.10:FF:000182, no label at all, restricted to Homo. A bare
P450 FunFam pinned to a single genus is the weakest possible evidence for a biological process
term, and the Homo restriction marks it as an annotation-availability artifact.
- number: 79
conditions:
- condition_type: FUNFAM
value: 2.40.400.10:FF:000003
curie: CATH.FunFam:2.40.400.10:FF:000003
label: "Protein NEOXANTHIN-DEFICIENT 1"
negated: false
- condition_type: TAXON
value: Gunneridae
curie: NCBITaxon:91827
label: "Gunneridae"
negated: false
notes: >-
ON-TARGET. FunFam 2.40.400.10:FF:000003 ('Protein NEOXANTHIN-DEFICIENT 1') + Gunneridae. NSY
converts violaxanthin to neoxanthin - xanthophyll (carotenoid) metabolism.
- number: 80
conditions:
- condition_type: FUNFAM
value: 2.60.200.20:FF:000048
curie: CATH.FunFam:2.60.200.20:FF:000048
negated: false
- condition_type: FUNFAM
value: 3.50.50.60:FF:000263
curie: CATH.FunFam:3.50.50.60:FF:000263
negated: false
notes: >-
HOLD. FunFam 2.60.200.20:FF:000048 + 3.50.50.60:FF:000263, both unlabelled, no taxon constraint.
3.50.50.60 is the FAD/NAD(P)-binding superfamily that legitimately carries the phytoene
desaturases in CS32/CS53/CS59/CS60, so the conjunction may well be a carotenoid enzyme - but
with neither FunFam named and no taxon constraint, this branch fires across all of UniProt on
unverifiable evidence.
- number: 81
conditions:
- condition_type: FUNFAM
value: 3.40.50.150:FF:000539
curie: CATH.FunFam:3.40.50.150:FF:000539
label: "juvenile hormone acid O-methyltransferase"
negated: false
- condition_type: TAXON
value: Arthropoda
curie: NCBITaxon:6656
label: "Arthropoda"
negated: false
notes: >-
ON-TARGET. FunFam 3.40.50.150:FF:000539 ('juvenile hormone acid O-methyltransferase') +
Arthropoda. JH is a sesquiterpenoid; JHAMT catalyses a committed late step. Well-matched taxon.
- number: 82
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000019
curie: CATH.FunFam:3.40.50.2000:FF:000019
negated: false
- condition_type: TAXON
value: Apiales
curie: NCBITaxon:4036
label: "Apiales"
negated: false
notes: >-
HOLD / LIKELY REMOVE. FunFam 3.40.50.2000:FF:000019, unlabelled, + Apiales. CATH 3.40.50.2000 is
the GT-B glycosyltransferase superfamily. Apiales include Panax (ginsenoside triterpenoid
glycosides), so a terpenoid glycosyltransferase is plausible - but 'plausible given the clade'
is precisely the reasoning that produces the errors in this rule. Inspect members.
- number: 83
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000037
curie: CATH.FunFam:3.40.50.2000:FF:000037
label: "Glycosyltransferase"
negated: false
- condition_type: TAXON
value: Caryophyllaceae
curie: NCBITaxon:3568
label: "Caryophyllaceae"
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 3.40.50.2000:FF:000037 labelled only 'Glycosyltransferase' +
Caryophyllaceae. A generic GT-B FunFam with a generic label. Same objection as CS14.
- number: 84
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000040
curie: CATH.FunFam:3.40.50.2000:FF:000040
negated: false
- condition_type: TAXON
value: Gentianales
curie: NCBITaxon:4055
label: "Gentianales"
negated: false
notes: >-
HOLD. FunFam 3.40.50.2000:FF:000040, unlabelled, + Gentianales. Gentianales include the
monoterpene indole alkaloid producers, so a secologanin-pathway glycosyltransferase is possible.
Unverifiable as written.
- number: 85
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000047
curie: CATH.FunFam:3.40.50.2000:FF:000047
negated: false
- condition_type: TAXON
value: Caryophylleae
curie: NCBITaxon:1141493
label: "Caryophylleae"
negated: false
notes: >-
HOLD. FunFam 3.40.50.2000:FF:000047, unlabelled, + Caryophylleae. Triterpenoid saponin
glycosylation is plausible in this clade; again unverifiable from the identifier.
- number: 86
conditions:
- condition_type: FUNFAM
value: 3.50.50.60:FF:000074
curie: CATH.FunFam:3.50.50.60:FF:000074
label: "Squalene monooxygenase 2"
negated: false
- condition_type: TAXON
value: Araliaceae
curie: NCBITaxon:4050
label: "Araliaceae"
negated: false
notes: >-
ON-TARGET. FunFam 3.50.50.60:FF:000074 ('Squalene monooxygenase 2') + Araliaceae. SQE oxidizes
squalene to 2,3-oxidosqualene, the branch point to triterpenoids and sterols; in Araliaceae this
feeds ginsenoside biosynthesis. Note the caveat that the sterol branch downstream is outside
GO:0006720, but the enzyme itself acts on squalene, an isoprenoid.
- number: 87
conditions:
- condition_type: FUNFAM
value: 3.90.79.10:FF:000025
curie: CATH.FunFam:3.90.79.10:FF:000025
negated: false
- condition_type: TAXON
value: Pentapetalae
curie: NCBITaxon:1437201
label: "Pentapetalae"
negated: false
notes: >-
HOLD. FunFam 3.90.79.10:FF:000025, unlabelled, + Pentapetalae. CATH 3.90.79.10 is the Nudix
hydrolase superfamily. Some plant Nudix enzymes are genuine prenyl-diphosphate phosphatases
(e.g. RhNUDX1 in rose, which makes monoterpene alcohols), but the superfamily is overwhelmingly
composed of unrelated nucleotide hydrolases. Not usable without member inspection.
- number: 88
conditions:
- condition_type: FUNFAM
value: 1.10.600.10:FF:000011
curie: CATH.FunFam:1.10.600.10:FF:000011
label: "Decaprenyl diphosphate synthase subunit 1"
negated: false
notes: >-
ON-TARGET, and empirically the second-largest producer. FunFam 1.10.600.10:FF:000011
('Decaprenyl diphosphate synthase subunit 1') = PDSS1. Together with the other prenyl-synthase
branches this accounts for ~30% of the rule's emitted annotations. No taxon constraint,
correctly.
- number: 89
conditions:
- condition_type: FUNFAM
value: 1.50.10.130:FF:000003
curie: CATH.FunFam:1.50.10.130:FF:000003
negated: false
notes: >-
KEEP WITH REVIEW. FunFam 1.50.10.130:FF:000003, unlabelled, no taxon constraint. Like CS38, the
superfamily itself is informative: CATH 1.50.10.130 is the terpene-cyclase beta-gamma fold, used
by plant class-II diterpene synthases (it is the partner domain in CS20, CS21, CS40 and CS61).
Likely a genuine TPS, but confirm the members.
- number: 90
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000021
curie: CATH.FunFam:2.60.120.330:FF:000021
negated: false
notes: >-
HOLD. FunFam 2.60.120.330:FF:000021, unlabelled, no taxon constraint. CATH 2.60.120.330 is the
double-stranded beta-helix 2-oxoglutarate/Fe(II) dioxygenase fold. The gibberellin oxidases in
CS28/CS29/CS31/CS91 live here, but so do hundreds of unrelated 2ODDs acting on flavonoids,
alkaloids and amino acids. An unnamed FunFam in this superfamily with no taxon constraint is not
safe.
- number: 91
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000025
curie: CATH.FunFam:2.60.120.330:FF:000025
label: "Gibberellin 2-beta-dioxygenase 2"
negated: false
notes: >-
ON-TARGET. FunFam 2.60.120.330:FF:000025 ('Gibberellin 2-beta-dioxygenase 2'). Named GA
catabolic enzyme.
- number: 92
conditions:
- condition_type: FUNFAM
value: 2.60.120.330:FF:000050
curie: CATH.FunFam:2.60.120.330:FF:000050
negated: false
notes: >-
HOLD. FunFam 2.60.120.330:FF:000050, unlabelled, no taxon constraint. Same objection as CS90.
- number: 93
conditions:
- condition_type: FUNFAM
value: 3.40.50.11270:FF:000001
curie: CATH.FunFam:3.40.50.11270:FF:000001
label: "4-hydroxy-3-methylbut-2-enyl diphosphate reductase"
negated: false
notes: >-
ON-TARGET. FunFam 3.40.50.11270:FF:000001 ('4-hydroxy-3-methylbut-2-enyl diphosphate reductase')
= IspH/LytB, the final MEP-pathway enzyme producing IPP and DMAPP. Among the most diagnostic
branches in the rule.
- number: 94
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000131
curie: CATH.FunFam:3.40.50.720:FF:000131
negated: false
notes: >-
OFF-TARGET - REMOVE. FunFam 3.40.50.720:FF:000131, unlabelled, no taxon constraint, in the
Rossmann NAD(P)-binding superfamily. Unnamed, unconstrained and in the most promiscuous fold
present - this branch cannot support any biological process term.
go_annotations:
- go_id: GO:0006720
go_label: isoprenoid metabolic process
aspect: P
reviewed_protein_count: 0
unreviewed_protein_count: 0
created_date: '2021-10-20'
modified_date: '2025-12-15'
entries:
- id: IPR001906
label: "Terpene synthase, N-terminal domain"
type: INTERPRO
appears_in_condition_sets:
- 1
- id: IPR005630
label: "Terpene synthase-like, metal-binding domain"
type: INTERPRO
appears_in_condition_sets:
- 1
- id: PTHR31225:SF93
type: PANTHER
appears_in_condition_sets:
- 1
- id: IPR008930
label: "Terpenoid cyclases/protein prenyltransferase alpha-alpha toroid"
type: INTERPRO
appears_in_condition_sets:
- 2
- id: IPR008949
label: "Isoprenoid synthase domain superfamily"
type: INTERPRO
appears_in_condition_sets:
- 2
- id: IPR034741
label: "Terpene cyclase-like 1, C-terminal domain"
type: INTERPRO
appears_in_condition_sets:
- 3
- id: IPR036965
label: "Terpene synthase, N-terminal domain superfamily"
type: INTERPRO
appears_in_condition_sets:
- 3
- id: PTHR31739:SF25
type: PANTHER
appears_in_condition_sets:
- 4
- id: IPR002060
label: "Squalene/phytoene synthase"
type: INTERPRO
appears_in_condition_sets:
- 5
- id: IPR019845
label: "Squalene/phytoene synthase, conserved site"
type: INTERPRO
appears_in_condition_sets:
- 5
- id: PTHR31480
type: PANTHER
appears_in_condition_sets:
- 5
- id: IPR001128
label: "Cytochrome P450"
type: INTERPRO
appears_in_condition_sets:
- 6
- id: PTHR47955
type: PANTHER
appears_in_condition_sets:
- 6
- id: IPR004294
label: "Carotenoid oxygenase"
type: INTERPRO
appears_in_condition_sets:
- 7
- id: IPR036396
label: "Cytochrome P450 superfamily"
type: INTERPRO
appears_in_condition_sets:
- 8
- id: PTHR47950:SF4
type: PANTHER
appears_in_condition_sets:
- 8
- id: IPR002937
label: "Amine oxidase"
type: INTERPRO
appears_in_condition_sets:
- 9
- id: IPR014105
label: "Carotenoid/retinoid oxidoreductase"
type: INTERPRO
appears_in_condition_sets:
- 9
- id: IPR036188
label: "FAD/NAD(P)-binding domain superfamily"
type: INTERPRO
appears_in_condition_sets:
- 9
- id: PTHR31225:SF98
type: PANTHER
appears_in_condition_sets:
- 10
- id: IPR034686
label: "Terpene cyclase-like 2"
type: INTERPRO
appears_in_condition_sets:
- 11
- id: PTHR31225:SF9
type: PANTHER
appears_in_condition_sets:
- 12
- id: IPR017825
label: "Lycopene cyclase domain"
type: INTERPRO
appears_in_condition_sets:
- 13
- id: IPR002213
label: "UDP-glucuronosyl/UDP-glucosyltransferase"
type: INTERPRO
appears_in_condition_sets:
- 14
- id: PTHR10543:SF57
type: PANTHER
appears_in_condition_sets:
- 15
- id: IPR000092
label: "Polyprenyl synthetase-like"
type: INTERPRO
appears_in_condition_sets:
- 16
- id: IPR039702
label: "Farnesyl pyrophosphate synthase-like"
type: INTERPRO
appears_in_condition_sets:
- 16
- id: IPR014102
label: "Phytoene desaturase"
type: INTERPRO
appears_in_condition_sets:
- 17
- id: IPR050464
label: "Zeta Carotene Desaturase and Related Oxidoreductases"
type: INTERPRO
appears_in_condition_sets:
- 17
- id: PTHR42923:SF45
type: PANTHER
appears_in_condition_sets:
- 17
- id: IPR010108
label: "Lycopene cyclase, beta/epsilon"
type: INTERPRO
appears_in_condition_sets:
- 18
- id: PTHR31225:SF120
type: PANTHER
appears_in_condition_sets:
- 19
- id: 1.10.600.10:FF:000005
label: "Ent-kaur-16-ene synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 20
- id: 1.50.10.130:FF:000004
label: "Carene synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 20
- id: 1.50.10.130:FF:000002
label: "Ent-copalyl diphosphate synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 21
- id: 1.10.600.10:FF:000007
label: "Isoprene synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 22
- id: 1.10.630.10:FF:000043
label: "Cytochrome P450 99A2"
type: FUNFAM
appears_in_condition_sets:
- 23
- id: 1.10.600.10:FF:000004
label: "Phytoene synthase chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 24
- id: 1.50.10.160:FF:000001
label: "Ent-copalyl diphosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 25
- id: 1.10.630.10:FF:000007
label: "Cytochrome P450 76C4"
type: FUNFAM
appears_in_condition_sets:
- 26
- id: 1.50.10.20:FF:000011
label: "Terpene cyclase/mutase family member"
type: FUNFAM
appears_in_condition_sets:
- 27
- id: 2.60.120.330:FF:000003
label: "Gibberellin 20 oxidase 2"
type: FUNFAM
appears_in_condition_sets:
- 28
- id: 2.60.120.330:FF:000013
label: "Gibberellin 3-beta-dioxygenase 1"
type: FUNFAM
appears_in_condition_sets:
- 29
- id: 1.10.600.10:FF:000020
label: "Phytoene synthase"
type: FUNFAM
appears_in_condition_sets:
- 30
- id: 2.60.120.330:FF:000014
label: "Gibberellin 2-beta-dioxygenase 1"
type: FUNFAM
appears_in_condition_sets:
- 31
- id: 3.50.50.60:FF:000091
label: "15-cis-phytoene desaturase, chloroplastic/chromoplastic"
type: FUNFAM
appears_in_condition_sets:
- 32
- id: 3.50.50.60:FF:000101
label: "lycopene epsilon cyclase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 33
- id: 1.10.3270.10:FF:000002
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
type: FUNFAM
appears_in_condition_sets:
- 34
- id: 3.30.70.420:FF:000001
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
type: FUNFAM
appears_in_condition_sets:
- 34
- id: 3.90.770.10:FF:000001
label: "3-hydroxy-3-methylglutaryl coenzyme A reductase"
type: FUNFAM
appears_in_condition_sets:
- 34
- id: 1.10.630.10:FF:000008
label: "Cytochrome P450 71D8"
type: FUNFAM
appears_in_condition_sets:
- 35
- id: 1.10.630.10:FF:000041
label: "Cytochrome P450 26A1 isoform 1"
type: FUNFAM
appears_in_condition_sets:
- 36
- id: 1.10.630.10:FF:000299
label: "Cytochrome P450 2C9"
type: FUNFAM
appears_in_condition_sets:
- 37
- id: 1.50.10.20:FF:000064
label: "Uncharacterized protein"
type: FUNFAM
appears_in_condition_sets:
- 38
- id: 3.40.50.720:FF:000808
label: "Iridoid synthase"
type: FUNFAM
appears_in_condition_sets:
- 39
- id: 1.50.10.130:FF:000006
label: "Terpene synthase 7"
type: FUNFAM
appears_in_condition_sets:
- 40
- id: 3.30.230.10:FF:000018
label: "Diphosphomevalonate decarboxylase"
type: FUNFAM
appears_in_condition_sets:
- 41
- id: 3.30.70.890:FF:000005
label: "Diphosphomevalonate decarboxylase"
type: FUNFAM
appears_in_condition_sets:
- 41
- id: 3.30.230.10:FF:000119
label: "Mevalonate kinase"
type: FUNFAM
appears_in_condition_sets:
- 42
- id: 3.30.70.890:FF:000003
label: "Mevalonate kinase"
type: FUNFAM
appears_in_condition_sets:
- 42
- id: 3.40.309.10:FF:000001
label: "Mitochondrial aldehyde dehydrogenase 2"
type: FUNFAM
appears_in_condition_sets:
- 43
- id: 3.40.605.10:FF:000026
label: "Aldehyde dehydrogenase, putative"
type: FUNFAM
appears_in_condition_sets:
- 43
- id: 3.40.605.10:FF:000054
label: "Aldehyde dehydrogenase family 1 member A3"
type: FUNFAM
appears_in_condition_sets:
- 43
- id: 3.40.50.1820:FF:000110
label: "Hormone-sensitive lipase"
type: FUNFAM
appears_in_condition_sets:
- 44
- id: 3.40.50.1820:FF:000199
label: "Hormone-sensitive lipase"
type: FUNFAM
appears_in_condition_sets:
- 44
- id: 1.10.600.10:FF:000001
label: "Geranylgeranyl diphosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 45
- id: 1.10.600.10:FF:000019
label: "2-methylisoborneol synthase"
type: FUNFAM
appears_in_condition_sets:
- 46
- id: 1.10.600.10:FF:000047
label: "Terpene synthase"
type: FUNFAM
appears_in_condition_sets:
- 47
- id: 3.40.50.150:FF:000183
label: "Geranyl diphosphate 2-C-methyltransferase"
type: FUNFAM
appears_in_condition_sets:
- 48
- id: 3.40.50.720:FF:000084
label: "Short-chain dehydrogenase reductase"
type: FUNFAM
appears_in_condition_sets:
- 49
- id: 3.40.50.720:FF:000145
label: "Retinol dehydrogenase 12"
type: FUNFAM
appears_in_condition_sets:
- 50
- id: 3.40.50.720:FF:001857
label: "Alcohol dehydrogenase class 4 mu/sigma chain"
type: FUNFAM
appears_in_condition_sets:
- 51
- id: 3.40.50.920:FF:000002
label: "1-deoxy-D-xylulose-5-phosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 52
- id: 3.40.50.970:FF:000005
label: "1-deoxy-D-xylulose-5-phosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 52
- id: 3.50.50.60:FF:000378
label: "Phytoene desaturase"
type: FUNFAM
appears_in_condition_sets:
- 53
- id: 3.90.1720.10:FF:000006
label: "Lecithin retinol acyltransferase"
type: FUNFAM
appears_in_condition_sets:
- 54
- id: 3.90.180.10:FF:000001
label: "S-(hydroxymethyl)glutathione dehydrogenase"
type: FUNFAM
appears_in_condition_sets:
- 55
- id: 1.10.600.10:FF:000036
label: "cis-abienol synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 56
- id: 1.10.630.10:FF:000052
label: "Ent-kaurenoic acid oxidase"
type: FUNFAM
appears_in_condition_sets:
- 57
- id: 1.10.630.10:FF:000062
label: "Ent-kaurene oxidase 2"
type: FUNFAM
appears_in_condition_sets:
- 58
- id: 3.50.50.60:FF:000171
label: "zeta-carotene-forming phytoene desaturase"
type: FUNFAM
appears_in_condition_sets:
- 59
- id: 3.50.50.60:FF:000413
label: "Phytoene desaturase (lycopene-forming)"
type: FUNFAM
appears_in_condition_sets:
- 60
- id: 1.10.600.10:FF:000042
label: "Probable terpene synthase 3, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 61
- id: 1.50.10.130:FF:000005
label: "S-(+)-linalool synthase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 61
- id: 1.50.10.20:FF:000002
label: "Terpene cyclase/mutase family member"
type: FUNFAM
appears_in_condition_sets:
- 62
- id: 1.50.10.20:FF:000022
label: "Terpene cyclase/mutase family member"
type: FUNFAM
appears_in_condition_sets:
- 62
- id: 2.10.25.10:FF:000009
label: "Low-density lipoprotein receptor isoform 1"
type: FUNFAM
appears_in_condition_sets:
- 63
- id: 4.10.400.10:FF:000011
label: "Low-density lipoprotein receptor-related protein 1"
type: FUNFAM
appears_in_condition_sets:
- 63
- id: 3.40.1090.10:FF:000003
label: "Patatin-like phospholipase domain-containing protein 2"
type: FUNFAM
appears_in_condition_sets:
- 64
- id: 3.40.1090.10:FF:000021
label: "Patatin-like phospholipase domain containing 2"
type: FUNFAM
appears_in_condition_sets:
- 64
- id: 3.40.309.10:FF:000021
label: "Aldehyde dehydrogenase family 8 member A1"
type: FUNFAM
appears_in_condition_sets:
- 65
- id: 3.40.605.10:FF:000001
label: "Aldehyde dehydrogenase 1"
type: FUNFAM
appears_in_condition_sets:
- 65
- id: 1.10.274.20:FF:000003
label: "Phenylalanine aminomutase (L-beta-phenylalanine forming)"
type: FUNFAM
appears_in_condition_sets:
- 66
- id: 1.10.600.10:FF:000008
label: "Farnesyl pyrophosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 67
- id: 1.10.600.10:FF:000018
label: "Probable geranylgeranyl-diphosphate geranylgeranyltransferase (AL-2)"
type: FUNFAM
appears_in_condition_sets:
- 68
- id: 1.10.600.10:FF:000021
label: "Farnesyl pyrophosphate synthase"
type: FUNFAM
appears_in_condition_sets:
- 69
- id: 1.10.600.10:FF:000076
label: "Terpene synthase"
type: FUNFAM
appears_in_condition_sets:
- 70
- id: 1.10.600.10:FF:000078
label: "Terpene synthase"
type: FUNFAM
appears_in_condition_sets:
- 71
- id: 1.10.630.10:FF:000009
label: "Cytochrome P450 26B1 isoform 1"
type: FUNFAM
appears_in_condition_sets:
- 72
- id: 1.10.630.10:FF:000011
label: "Cytochrome P450 83B1"
type: FUNFAM
appears_in_condition_sets:
- 73
- id: 1.10.630.10:FF:000014
label: "Abscisic acid 8"
type: FUNFAM
appears_in_condition_sets:
- 74
- id: 1.10.630.10:FF:000022
label: "Taxadiene 5-alpha hydroxylase"
type: FUNFAM
appears_in_condition_sets:
- 75
- id: 1.10.630.10:FF:000080
label: "Carotene epsilon-monooxygenase, chloroplastic"
type: FUNFAM
appears_in_condition_sets:
- 76
- id: 1.10.630.10:FF:000097
label: "Cytochrome P-450 19"
type: FUNFAM
appears_in_condition_sets:
- 77
- id: 1.10.630.10:FF:000182
type: FUNFAM
appears_in_condition_sets:
- 78
- id: 2.40.400.10:FF:000003
label: "Protein NEOXANTHIN-DEFICIENT 1"
type: FUNFAM
appears_in_condition_sets:
- 79
- id: 2.60.200.20:FF:000048
type: FUNFAM
appears_in_condition_sets:
- 80
- id: 3.50.50.60:FF:000263
type: FUNFAM
appears_in_condition_sets:
- 80
- id: 3.40.50.150:FF:000539
label: "juvenile hormone acid O-methyltransferase"
type: FUNFAM
appears_in_condition_sets:
- 81
- id: 3.40.50.2000:FF:000019
type: FUNFAM
appears_in_condition_sets:
- 82
- id: 3.40.50.2000:FF:000037
label: "Glycosyltransferase"
type: FUNFAM
appears_in_condition_sets:
- 83
- id: 3.40.50.2000:FF:000040
type: FUNFAM
appears_in_condition_sets:
- 84
- id: 3.40.50.2000:FF:000047
type: FUNFAM
appears_in_condition_sets:
- 85
- id: 3.50.50.60:FF:000074
label: "Squalene monooxygenase 2"
type: FUNFAM
appears_in_condition_sets:
- 86
- id: 3.90.79.10:FF:000025
type: FUNFAM
appears_in_condition_sets:
- 87
- id: 1.10.600.10:FF:000011
label: "Decaprenyl diphosphate synthase subunit 1"
type: FUNFAM
appears_in_condition_sets:
- 88
- id: 1.50.10.130:FF:000003
type: FUNFAM
appears_in_condition_sets:
- 89
- id: 2.60.120.330:FF:000021
type: FUNFAM
appears_in_condition_sets:
- 90
- id: 2.60.120.330:FF:000025
label: "Gibberellin 2-beta-dioxygenase 2"
type: FUNFAM
appears_in_condition_sets:
- 91
- id: 2.60.120.330:FF:000050
type: FUNFAM
appears_in_condition_sets:
- 92
- id: 3.40.50.11270:FF:000001
label: "4-hydroxy-3-methylbut-2-enyl diphosphate reductase"
type: FUNFAM
appears_in_condition_sets:
- 93
- id: 3.40.50.720:FF:000131
type: FUNFAM
appears_in_condition_sets:
- 94
review_summary: >-
ARBA00027853 is a 94-branch omnibus rule whose GO term is correct and whose biological
core is sound, but which is not auditable as published. GO:0006720 is a genuinely
appropriate broad parent for the families the rule mostly targets: its definition covers
compounds "containing or derived from linked isoprene residues", and QuickGO ancestor
closures checked on 2026-08-22 confirm that retinoid metabolic process (GO:0001523),
terpenoid metabolic process (GO:0006721), gibberellin biosynthetic process (GO:0009686)
and the tetraterpenoid/diterpenoid biosynthetic terms are all descendants of it. That
definition also answers the objection recorded in geneontology/go-annotation#5835 that a
metabolite "is derived from an isoprenoid, but not an isoprenoid" - derived-from is
explicitly in scope. The boundary that does bite is steroids: GO:0008202 steroid metabolic
process is NOT a descendant of GO:0006720 (its QuickGO ancestor closure is GO:0008150,
GO:0008152, GO:0006629, GO:0008202, GO:0044238, GO:0009987), which makes two branches
wrong-branch rather than merely broad. Note also that the specific protein cited in #5835,
S. pombe SPAC31G5.16c = dpm1 (O14466), is not reachable from any of this rule's condition
sets and carries no GO:0006720 today; that part of the issue belongs to ARBA00028538 and
ARBA00028655. This puts ARBA00027853 in a different position from its sibling rules in
#5835: the GO term itself is in the right branch here, so the rule's problems are ones of
branch construction and reproducibility rather than of a mis-chosen term. Branch triage
gives 64 of 94 condition sets on-target, 13 to remove, 16 unresolvable from their
identifiers, and one (CS39) mixed and needing a split. The empirical picture is better than the branch
count suggests. The UniProt API reports zero annotated proteins for this rule - the figure
the commissioned deep research was given, and on which it built its "no present production
set" reasoning - but a QuickGO census on 2026-08-22 finds 8,974 live GO:0006720 /
ECO:0000256 / GO_REF:0000117 annotations citing ARBA:ARBA00027853, across 1,570 taxa. A
protein-name census of all 8,974 shows 47.6% carotenoid/retinoid cleavage enzymes (BCO1,
BCO2, RPE65, NinaB), 30.4% prenyl and polyprenyl diphosphate synthases (PDSS1, PDSS2,
FPPS, GGPPS), 16.0% uncharacterized entries, and a demonstrable false-positive tail of only
a few percent. The 47 human annotations, enumerated exhaustively, are 100% on-target. The
serious finding is a reproducibility failure: in a random 300-protein sample, 187 (62%)
carry InterPro IPR004294 "Carotenoid oxygenase" but 167 of those 187 are not mammals, while
the rule's only IPR004294 branch (CS7) requires taxon Mammalia; and 89 (30%) carry
IPR000092 while only 25 of them also carry the IPR039702 that the rule's only IPR000092
branch (CS16) requires. Either the published conjunctions are not enforced at annotation
time or the rule content served today has diverged from the annotating release. Either way,
the constraints that make this rule look safe on paper are invisible in the data it is
credited with, and that is a plausible mechanism for the "lots of off target inferences"
reported by curators.
action: MODIFY
action_rationale: >-
MODIFY rather than DEPRECATE, because the rule is currently producing roughly 8,900
annotations that are in the large majority correct and useful - carotenoid oxygenases and
polyprenyl diphosphate synthases across 1,570 taxa, with a 100%-correct human subset.
Retiring the rule would discard far more good annotation than bad. This is the opposite
situation from ARBA00028655 in the same GO issue, where under 1% of emitted annotations
were defensible and DEPRECATE/SPLIT was the right call. MODIFY rather than ACCEPT, because
13 branches are indefensible on their own terms and because the rule cannot be audited
against its own output. The commissioned deep research recommends SPLIT into pathway-specific
rules (precursor/backbone, terpene synthases, carotenoids, gibberellin/ABA, retinoids,
sterol/triterpenoid, juvenile hormone), which is a reasonable longer-term target and would
also let each fragment carry a more specific descendant term than GO:0006720; but the
immediate, minimal fix that answers the curators' complaint is branch surgery plus
reconciliation of the published rule with its emitted annotations, not restructuring. The
single highest-priority action is not biological at all: UniProt should reconcile the
condition sets with the annotation set and fix the statistics block that reports zero
proteins for a rule with 8,974 live annotations - that field caused an entire commissioned
literature review to be conducted on a false premise. The confidence value below is
confidence in this recommendation, not in the rule.
suggested_modifications:
- 'RECONCILE the published condition sets with the emitted annotation set, and fix the
statistics block. The API reports reviewedProteinCount 0 / unreviewedProteinCount 0 while
QuickGO reports 8,974 live annotations attributed to this rule (2026-08-22). Separately,
62% of a random 300-protein sample carries IPR004294 with 89% of those non-mammalian,
although the only IPR004294 branch requires Mammalia. Until this is explained, no
per-branch curation decision below can be verified against real output.'
- 'REMOVE the 13 unusable condition sets - 12 off-target plus CS43, which is removable because
it is unsatisfiable rather than because it is off-target: CS14 (IPR002213 bare UGT family +
Caryophyllaceae), CS37 (CYP2C9 FunFam, Primates), CS43 (requires two distinct FunFams from
within the same CATH superfamily 3.40.605.10, which one ALDH catalytic domain cannot satisfy),
CS44 (hormone-sensitive lipase FunFams, Craniata), CS51 (ADH class 4 / ADH7, Vertebrata),
CS55 (ADH5 / S-(hydroxymethyl)glutathione dehydrogenase, Mus), CS63 (LDLR + LRP1 FunFams,
Catarrhini), CS64 (PNPLA2/ATGL FunFams, Eutheria), CS66 (phenylalanine aminomutase,
Pinopsida), CS73 (CYP83B1, Asterales), CS78 (unlabelled P450 FunFam, Homo), CS83 (FunFam
labelled only "Glycosyltransferase", Caryophyllaceae), CS94 (unlabelled Rossmann/SDR FunFam,
no taxon).'
- 'SPLIT CS39 (CATH FunFam 3.40.50.720:FF:000808, "Iridoid synthase", asterids). This is the
PRISE family, which contains both iridoid synthases (isoprenoid, correct) and progesterone
5-beta-reductase / 3-oxo-Delta(4,5)-steroid 5-beta-reductases (steroid, and GO:0008202 is
not under GO:0006720). Empirically the steroid members outnumber the iridoid members in this
rule''s output: 132 vs 84 of 8,974 annotations. Restrict the branch to the iridoid synthase
members, or drop it.'
- 'RESOLVE or drop the 16 unauditable branches: CS6 (generic P450 + PANTHER CYP71 family +
Asterales), CS23, CS35, CS49, CS65, CS75, CS77, CS80, CS82, CS84, CS85, CS87, CS90, CS92 and
- with a note that these two are likely keeps rather than deletes - CS38 and CS89. A bare
FunFam identifier is not evidence, but the superfamily it sits in still carries information:
CS38 (1.50.10.20) and CS89 (1.50.10.130) are in isoprenoid-dedicated cyclase folds, whereas
CS90/CS92 (2.60.120.330 2OG dioxygenase), CS94 (3.40.50.720 Rossmann) and CS82/CS84/CS85
(3.40.50.2000 GT-B) are in folds that carry mostly unrelated chemistry.'
- 'FIX three FunFam branches whose label names an enzyme that cannot occur in the clade the
branch is restricted to - a reliable sign that a FunFam label is being read as a function
rather than as the name of its best-studied member. CS23: "Cytochrome P450 99A2" (a rice,
Poales, momilactone P450) restricted to campanulids. CS35: "Cytochrome P450 71D8" (a legume
enzyme) restricted to Poales. CS75: "Taxadiene 5-alpha hydroxylase" (Taxus, Pinopsida)
restricted to Caryophyllales.'
- 'CORRECT the truncated label on CS74, "Abscisic acid 8", to "abscisic acid 8''-hydroxylase"
(CYP707A). The branch is biologically on-target - ABA is an apocarotenoid and its catabolism
is inside GO:0006720 - but a label that stops mid-word is exactly the kind of thing that
conceals a wrong FunFam id.'
- 'REVIEW the taxon constraints as a whole. Twenty-four of 94 sets have no taxon constraint at
all (CS1, CS5, CS9, CS13, CS17, CS24, CS30, CS34, CS40, CS43, CS52, CS56-CS60, CS80,
CS88-CS94) and therefore fire across all of UniProt, while other branches on universally
conserved enzymes are pinned to a single genus or family: CS55 (Mus, on ADH5), CS78 (Homo,
on an unlabelled P450 FunFam), CS72 (Hominidae, on CYP26B1), CS37 (Primates), CS50
(Haplorrhini, on RDH12), CS63 (Catarrhini). These read as annotation-availability artifacts
of the association-rule mining. By contrast the Lamiaceae tribe restrictions (Ocimeae,
Nepetoideae, Elsholtzieae, Mentheae), Pinus, Amoebozoa/Evosea/Eumycetozoa, Actinomycetota
and Arthropoda restrictions track real lineage-specific family expansions and should stay.'
- 'CONSIDER the deep research''s SPLIT recommendation as the longer-term target: separate rules
for MVA/MEP precursor metabolism, prenyl diphosphate synthases, terpene synthases,
carotenoid metabolism, gibberellin/ABA metabolism, retinoid metabolism, triterpenoid/squalene
metabolism and juvenile hormone metabolism. Each fragment could then carry a more specific
descendant of GO:0006720 - GO:0016114 terpenoid biosynthetic process, GO:0016109
tetraterpenoid biosynthetic process, GO:0009686 gibberellin biosynthetic process, GO:0001523
retinoid metabolic process - instead of the shared broad parent.'
- 'DO NOT act on the deep research''s recommendation to hold or delete the PANTHER-based
branches CS1, CS4, CS5, CS8, CS10, CS12, CS15, CS17 and CS19 as "opaque identifiers". All
nine resolve against interpro/panther/panther.obo to named isoprenoid families (see
condition-set notes) and are on-target. Likewise CS20 and CS61 are not, as the report
speculated, internally contradictory multi-FunFam conjunctions: the two FunFams in each are
in different CATH superfamilies (1.10.600.10 isoprenoid-synthase alpha fold and 1.50.10.130
terpene-cyclase beta-gamma fold) and plant class-I diterpene synthases carry both, so the
conjunction is a sound domain-architecture requirement.'
parsimony:
assessment: OVERLY_COMPLEX
notes: >-
94 condition sets and 122 distinct signature entries for a single GO term is far past the
point where a rule can be reasoned about; the repository's own analysis tooling refuses to
process it, capping post-enrichment analysis at 12 condition sets. Some of the complexity
is legitimate - isoprenoid metabolism genuinely spans plant terpene synthases, bacterial
2-methylisoborneol synthases, mammalian carotenoid oxygenases, insect JHAMT and the MVA and
MEP pathways, and no small set of signatures covers that. But much of it is accretion. The
same enzyme concept is expressed repeatedly through different databases and taxa: farnesyl
diphosphate synthase appears in CS16 (InterPro, Ecdysozoa), CS67 (FunFam, rosids) and CS69
(FunFam, Ecdysozoa, overlapping CS16 directly); ent-copalyl diphosphate synthase in CS21
and CS25; phytoene synthase in CS24 and CS30; slime-mould terpene synthases in CS47, CS70
and CS71 across three nested clades of Amoebozoa; phytoene desaturase in CS17, CS32, CS53,
CS59 and CS60. The decisive parsimony argument, though, is empirical rather than structural:
the emitted annotation set is dominated by carotenoid oxygenases and prenyl diphosphate
synthases, so the great majority of the 94 branches contribute nothing observable. The ~19
terpene synthase branches - the largest single group - account for 0.4% of output.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-analysis.md
supporting_text: "condition sets contribute nothing observable"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "biologically coherent at the GO-term level, but unsafe as a single 94-branch OR rule"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "The top-level OR, however, means that the weakest single branch determines the rule’s overall false-positive liability."
literature_support:
assessment: MODERATE
notes: >-
The commissioned deep research (Falcon / Edison Scientific Literature, run 2026-08-22)
supports the GO term for the rule's core families and contradicts a defined minority of its
branches - a genuinely mixed verdict, unlike the flat contradiction found for ARBA00028655.
Its positive case rests on pathway-level reviews placing carotenoids, gibberellins, ABA and
related hormones downstream of IPP/DMAPP and prenyl diphosphates (Bajguz & Piotrowska-
Niczyporuk 2023, Metabolites 13:884), and on terpene synthase biology (Karunanithi & Zerbe
2019, PMID:31632418). Its negative case rests on documented enzyme promiscuity: Werck-
Reichhart 2023 (PMID:36830762) reports that CYP706A3 "oxidizes more than twenty different
mono- and sesquiterpenes" and that CYP720B4 "catalyzes the three successive oxidations at
C18 of 8 out of 24 different diterpenoid olefin skeletons" - which is what makes "generic
P450 domain plus plant clade" non-diagnostic. It also cites quantitative limits on
domain-based process transfer - DomFun Fmax 0.624 for molecular function versus 0.492 for
biological process (Rojano et al. 2022, PMID:35033002); FunFam members agreeing on only
36.9 +/- 0.6% of binding-residue annotations (Scheibenreif et al. 2019, PMID:31319797).
Four honest limits on this evidence. First, the report was given the API's
false "zero annotated proteins" figure and built its "no present production set from which
empirical precision can be measured" reasoning on it; the census in the companion analysis
corrects that. Second, it does not mention the GO Consortium or issue #5835 anywhere, so it
offers no independent read on the curators' complaint and no Consortium ruling should be
claimed from it. Third, its blanket recommendation to hold
or delete "opaque" PANTHER identifiers and unnamed FunFams is too coarse - nine of the
PANTHER branches resolve to named isoprenoid families, and two unnamed FunFams sit in
isoprenoid-dedicated cyclase superfamilies. Fourth, one figure it quotes needs a
scope caveat rather than a correction: its "29 documented substrates" for CYP706A3 is
accurate - PMID:36830762 states verbatim "A total of 29 different substrates are thus
currently reported for this enzyme" - but that total is reached only by adding the
dinitroaniline herbicides to the terpenoid substrates, so 29 is not a count of terpenoid
substrates and should not be cited as one. Only one external provider was reachable in this
environment: Perplexity returned an insufficient-quota error, Cyberian is not configured,
and the OpenAI deep-research model returned a 404, so this review rests on one commissioned
report plus the independent empirical census rather than on two providers.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "GO:0006720 is **neither too narrow nor intrinsically incorrect** for the genuine pathway enzymes in this rule."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "Several branches appear clearly off-target—notably CYP2C9, CYP83B1, hormone-sensitive lipase, PNPLA2, LDLR/LRP1, phenylalanine aminomutase, class-4 alcohol dehydrogenase and formaldehyde dehydrogenase."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "one experimentally studied CYP706A3 accepts more than 20 mono- and sesquiterpenes plus herbicides, with 29 documented substrates"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "These branches should not be retained merely because their substrates or transported particles can contain lipid-soluble isoprenoids."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "a named, well-populated FunFam can be strong evidence, but a bare FunFam ID is not self-validating, particularly for a biological-process term"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "This concern is amplified by the stated outcome of **zero currently annotated proteins**: there is no present production set from which empirical precision can be measured."
condition_overlap:
assessment: SIGNIFICANT
notes: >-
Quantitative pairwise overlap could not be computed: the repository's analyse-rule step
refuses rules above 12 condition sets ("Rule ARBA00027853 has 94 condition sets, which
exceeds the maximum of 12"), so no pairwise_overlap blocks are populated below and the
assessment here is structural rather than metric. Read at that level the redundancy is
plain. Farnesyl diphosphate synthase is reached three times - CS16 (IPR000092 + IPR039702,
Ecdysozoa), CS67 (FunFam 1.10.600.10:FF:000008, rosids) and CS69 (FunFam
1.10.600.10:FF:000021, Ecdysozoa) - with CS16 and CS69 targeting the same enzymes in the
same clade through different databases. Phytoene desaturase appears in five sets (CS17,
CS32, CS53, CS59, CS60), phytoene synthase in three (CS5, CS24, CS30), ent-copalyl
diphosphate synthase in two (CS21, CS25), triterpene cyclases in three (CS27, CS38, CS62),
and Amoebozoan terpene synthases in three nested clades (CS47 Amoebozoa, CS70 Evosea, CS71
Eumycetozoa - Evosea and Eumycetozoa both lie inside Amoebozoa, so CS70 and CS71 add
coverage only if their FunFams differ from CS47's, which is not established). CATH
superfamily 1.10.600.10 alone supplies conditions to 14 different sets and 1.10.630.10 to
13. Cross-database redundancy of this kind is not harmful in itself - it buys coverage when
InterPro, PANTHER and CATH disagree about a protein - but at this scale it makes the rule
unmaintainable and hides which branch is responsible for any given annotation, which is
exactly the problem when a curator files an off-target report.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "Redundancy is extensive among TPS InterPro/PANTHER/FunFam conditions, carotenoid synthase/desaturase/cyclase conditions, prenyl synthases and GA enzymes."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "it complicates maintenance and can conceal inconsistent thresholds or obsolete family labels"
go_specificity:
assessment: APPROPRIATE
notes: >-
GO:0006720 is the right level for a rule whose branches span terpene, carotenoid, retinoid,
gibberellin, ABA, triterpenoid and juvenile-hormone chemistry, and it is genuinely correct
for the families that dominate the output. Its definition - "compounds containing or
derived from linked isoprene residues" - covers apocarotenoids and retinoids, and QuickGO
ancestor closures checked on 2026-08-22 confirm GO:0001523 retinoid metabolic process,
GO:0006721 terpenoid metabolic process, GO:0016114 terpenoid biosynthetic process,
GO:0016109 tetraterpenoid biosynthetic process, GO:0016102 diterpenoid biosynthetic process
and GO:0009686 gibberellin biosynthetic process are all descendants. Breadth is also a
virtue for specific branches: CS5 combines the squalene/phytoene synthase family with its
conserved site, and that signature cannot distinguish the sterol entry reaction from the
carotenoid entry reaction, so the shared parent is the honest term. The one place the term
is wrong rather than broad is where a branch reaches steroid chemistry: GO:0008202 steroid
metabolic process is not a descendant of GO:0006720, which condemns CS63 (LDLR/LRP1,
cholesterol uptake) and the progesterone 5-beta-reductase half of CS39. Against APPROPRIATE
it must be said that for the strongest branches - HMGCR, DXS, IspH, mevalonate kinase, the
GA oxidases, CYP26, LRAT, JHAMT - a specific descendant would carry far more information
than the shared parent, and the deep research argues the term is "too broad as the only
annotation" for those. That is an argument for splitting the rule, which is recorded under
action_rationale, not for calling the term itself mismatched.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-analysis.md
supporting_text: "**steroids are not in the\nisoprenoid branch**"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "GO:0006720 is therefore safer than either a squalene- or phytoene-specific process term."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "For many high-confidence branches, GO:0006720 is **too broad as the only annotation**, although it remains a valid ancestor."
taxonomic_scope:
assessment: TOO_NARROW
notes: >-
The constraints are internally inconsistent, and the dominant failure is over-restriction
on conserved families rather than over-inclusion. Six branches pin universally conserved
enzymes to a single genus, family or primate clade: CS55 restricts ADH5 - a housekeeping
formaldehyde dehydrogenase present in essentially every eukaryote - to Mus; CS78 restricts
an unlabelled P450 FunFam to Homo; CS72 restricts CYP26B1 to Hominidae and CS36 restricts
CYP26A1 to Metazoa though both are vertebrate-wide; CS50 restricts RDH12 to Haplorrhini;
CS37 restricts CYP2C9 to Primates; CS63 restricts LDLR/LRP1 to Catarrhini. Restrictions of
that shape are artifacts of where experimental annotation happens to exist, not statements
about where the enzyme exists, and they guarantee the rule under-annotates real orthologues.
Similarly CS41 (Mammalia) and CS42 (Chordata) restrict diphosphomevalonate decarboxylase and
mevalonate kinase, which are not mammal- or chordate-specific. In the other direction, 24
branches carry no taxon constraint at all and fire across all of UniProt, including CS94
(bare Rossmann FunFam) and, nominally, CS43 (ALDH FunFams) - though CS43 appears
unsatisfiable, so the missing constraint there is moot. Where such a branch does fire, a
constraint would at least have limited the damage. Where the restrictions do track biology they are well chosen
- the Lamiaceae tribes Ocimeae, Nepetoideae, Elsholtzieae and Mentheae for terpene synthase
expansions, Pinus for conifer TPS, Amoebozoa for slime-mould TPS, Bacillati and
Actinomycetota for 2-methylisoborneol synthase and geranyl diphosphate methyltransferase,
Arthropoda for JHAMT - and those should be kept. Finally, the empirical audit shows the
constraints may not be doing what they appear to: 167 of 187 sampled IPR004294 proteins are
non-mammalian although the only IPR004294 branch requires Mammalia.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-analysis.md
supporting_text: "these read as annotation-availability artifacts of\nthe association-rule mining, not as biology"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "Several restrictions may be unnecessarily narrow: CYP26, RDH12, LRAT, GA enzymes and many prenyl synthases are conserved beyond the listed crown taxa."
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "They become problematic when used to rescue a generic fold. A Caryophyllaceae glycosyltransferase, Asterales P450 or mammalian ALDH is not automatically an isoprenoid enzyme."
confidence: 0.8
references:
- id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
title: "Deep research analysis via Falcon (Edison Scientific Literature), run 2026-08-22"
findings:
- statement: >-
GO:0006720 is an appropriate broad parent for the rule's genuine pathway enzymes -
terpene synthases, MVA/MEP precursor enzymes, prenyl diphosphate synthases, carotenoid
enzymes, gibberellin and ABA enzymes, retinoid enzymes and JHAMT - but the rule is
"unsafe as a single 94-branch OR rule".
- statement: >-
Eight branch groups are off-target and should be removed: CYP2C9 (CS37), the mixed ALDH2
conjunction (CS43), hormone-sensitive lipase (CS44), PNPLA2 (CS64), alcohol and
formaldehyde dehydrogenases (CS51, CS55), LDLR/LRP1 (CS63), phenylalanine aminomutase
(CS66) and CYP83B1 (CS73).
- statement: >-
Plant cytochrome P450 promiscuity is the rule rather than the exception - CYP706A3 has 29
documented substrates spanning more than 20 mono- and sesquiterpenes plus herbicides - so
a generic P450 domain plus a plant taxon cannot predict isoprenoid metabolism.
- statement: >-
Domain-based transfer is measurably weaker for biological process than for molecular
function (DomFun Fmax 0.492 vs 0.624), and FunFam members agree on only 36.9% of
binding-residue annotations, so a bare FunFam identifier is not self-validating for a
process term.
- statement: >-
The report was given the UniProt API's figure of zero annotated proteins and reasoned
from it that no empirical precision estimate was possible. That premise is false; see the
companion analysis.
- id: file:rules/arba/ARBA00027853/ARBA00027853-analysis.md
title: "ARBA00027853 analysis - condition-set triage, ontology-branch check and annotation census"
findings:
- statement: >-
QuickGO on 2026-08-22 reports 8,974 live GO:0006720 / ECO:0000256 / GO_REF:0000117
annotations citing ARBA:ARBA00027853 across 1,570 taxa, contradicting the rule's own
statistics block of zero proteins.
- statement: >-
A protein-name census of all 8,974 gives 47.6% carotenoid/retinoid cleavage enzymes,
30.4% prenyl and polyprenyl diphosphate synthases and 16.0% uncharacterized entries, with
a demonstrable false-positive tail of a few percent; the 47 human annotations are 100%
on-target.
- statement: >-
In a random 300-protein sample, 187 carry IPR004294 but 167 of those are non-mammalian
although the rule's only IPR004294 branch requires Mammalia, and only 25 of 89 IPR000092
proteins carry the co-required IPR039702 - so the emitted set cannot be reproduced from
the published condition sets.
- statement: >-
GO:0008202 steroid metabolic process is not a descendant of GO:0006720, which makes CS63
(LDLR/LRP1) and the progesterone 5-beta-reductase half of CS39 wrong-branch rather than
merely broad.
- statement: >-
Nine PANTHER conditions the deep research treated as unauditable resolve against
interpro/panther/panther.obo to named isoprenoid families, and the CS20/CS61 multi-FunFam
conjunctions it suspected of being contradictory are sound domain-architecture
requirements spanning two different CATH superfamilies.
- id: file:rules/arba/ARBA00027853/scripts/census_arba00027853.py
title: "Reproducible QuickGO/UniProt census script for ARBA00027853"
findings:
- statement: >-
Re-derives the annotation count, protein-name census and signature/taxon audit from
primary sources with nothing hard-coded; run on 2026-08-22 it reproduces the figures
quoted in the analysis and this review.
- id: PMID:37623827
title: "Biosynthetic Pathways of Hormones in Plants."
findings:
- statement: >-
Places gibberellins, abscisic acid, strigolactones and brassinosteroids downstream of
IPP/DMAPP and the prenyl diphosphates, which is the pathway-level basis for treating
GA20ox/GA3ox/GA2ox, ent-kaurene oxidase, ent-kaurenoic acid oxidase and the ABA
8'-hydroxylases (CS28, CS29, CS31, CS57, CS58, CS74, CS91) as isoprenoid metabolism.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.3390/metabo13080884 via NCBI esearch and cached with full text
on 2026-08-22; this is the review the Falcon report leans on for its positive case.
- id: PMID:36830762
title: "Promiscuity, a Driver of Plant Cytochrome P450 Evolution?"
findings:
- statement: >-
CYP706A3 "oxidizes more than twenty different mono- and sesquiterpenes", and CYP720B4
"catalyzes the three successive oxidations at C18 of 8 out of 24 different diterpenoid
olefin skeletons". Plant P450 substrate breadth of this kind is why a generic P450 domain
plus a plant clade (CS6, CS23, CS26, CS35, CS73, CS77) cannot predict isoprenoid
metabolism.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.3390/biom13020394 and cached with full text on 2026-08-22. The
two quoted figures were checked verbatim against the cached text, as was the Falcon
report's "29 documented substrates" for CYP706A3: the article states "A total of 29
different substrates are thus currently reported for this enzyme". The caveat on that
number is one of scope, not accuracy - it counts the dinitroaniline herbicides alongside
the mono- and sesquiterpenes, so it is a count of total substrate breadth and not of
terpenoid substrates.
- id: PMID:31632418
title: "Terpene Synthases as Metabolic Gatekeepers in the Evolution of Plant Terpenoid"
findings:
- statement: >-
Plant terpene synthase families diversify by duplication, domain loss/gain and small
active-site changes that redirect carbocation cascades and product profiles. This supports
annotating TPS branches at the level of GO:0006720 rather than transferring
product-specific terms from a broad TPS domain.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.3389/fpls.2019.01166 and cached with full text on 2026-08-22.
- id: PMID:35033002
title: "Assigning protein function from domain-function associations using DomFun."
findings:
- statement: >-
Domain-based function transfer benchmarks measurably worse for biological process than
for molecular function (Fmax 0.492 vs 0.624 in the no-knowledge partial evaluation),
which is the general reason a FunFam or InterPro signature is weaker evidence for
GO:0006720 than it would be for a corresponding molecular function term.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1186/s12859-022-04565-6 and cached on 2026-08-22. Methodological
background, not evidence about isoprenoid biology.
- id: PMID:38122964
title: "FunPredCATH: An ensemble method for predicting protein function using CATH."
full_text_unavailable: true
findings:
- statement: >-
CATH FunFam-based function prediction reaches competitive but bounded accuracy, with
union models increasing false-positive risk and intersection models trading coverage for
precision - the trade-off this rule makes implicitly by OR-ing 94 branches.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1016/j.bbapap.2023.140985 and cached on 2026-08-22.
Abstract-only (no PMC record), so the specific Fmax figures quoted by the Falcon report
were not independently checked against full text.
- id: PMID:31319797
title: "FunFam protein families improve residue level molecular function prediction."
findings:
- statement: >-
FunFam members agree on only 36.9 +/- 0.6% of binding-residue annotations - better than
random grouping but far from uniform. A named, well-populated FunFam can be strong
evidence, but a bare FunFam identifier is not self-validating, which is the basis for
holding rather than accepting the 16 unlabelled-FunFam branches.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1186/s12859-019-2988-x and cached with full text on 2026-08-22.
supported_by:
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-deep-research-falcon.md
supporting_text: "the rule contains a substantial, well-supported biological core, but its current omnibus logic is not curatorially defensible"
- reference_id: file:rules/arba/ARBA00027853/ARBA00027853-analysis.md
supporting_text: "the constraints that make this rule look safe on\npaper are not visible in the data it is credited with"