View original ARBA rule on UniProt
Assigns GO:0046467 to proteins matching any of 51 alternative condition sets built from 14 InterPro entries, 51 CATH FunFams and 39 taxon constraints. GO obsoleted GO:0046467 ("membrane lipid biosynthetic process") on 2025-12-09, replacing it with GO:0008610 lipid biosynthetic process, so the rule now emits no annotations at all - yet UniProt still serves it, and its last modification date of 2025-12-15 is six days after the obsoletion. The antecedents are heterogeneous: at most 26 of the 51 sets describe enzymes that build a membrane lipid, while the rest capture sphingolipid hydrolases, lipid A and GPI remodelling enzymes, GPI/dolichol-donor assembly, non-diagnostic glycosyltransferase families, and four mechanistically unrelated proteins reached through promiscuous folds (a PH-domain protein kinase, phosphoglycerate kinase, 5-aminolevulinate synthase, and a lipid-binding lipoprotein).
Condition-set counts describe the sets recorded in this review, which may omit the full rule.
ARBA00028538 asserts a single consequent, GO:0046467, which the Gene Ontology obsoleted on 2025-12-09 (tracker geneontology/go-ontology#26698) as part of retiring the whole membrane-lipid grouping branch, on the ground that it grouped lipids by localization rather than by metabolic function. QuickGO now returns zero annotations to GO:0046467 from any source, so the rule is inert. It is not, however, retired: UniProt still serves it and its modifiedDate of 2025-12-15 falls six days after the obsoletion, which means routine maintenance touched the rule without noticing that its only output term had ceased to exist. That matters because the obvious repair - substituting the replaced_by term GO:0008610 lipid biosynthetic process - would silently reactivate the rule and restore the errors that a GO curator raised against it in go-annotation#5835. Auditing all 51 OR-ed condition sets against what their InterPro entries and CATH FunFams actually identify, at most 26 describe an enzyme that builds a membrane lipid, and even those would be better served by pathway-specific children (GO:0030148, GO:0046513, GO:0006688, GO:0009245). Five sets are directionally wrong: alkaline ceramidase, acid and neutral sphingomyelinase, neutral ceramidase and lysosomal glucosylceramidase all hydrolyse membrane sphingolipids. Six describe modification or interconversion rather than synthesis (PagP, lipid A 1-phosphatase, sphingoid-base phosphatase, sphingosine kinase 2, PNPLA1, VLCFA reductase). Five sets are non-diagnostic families whose acceptor could equally be a protein or a wall polymer, and five concern GPI-anchor or dolichyl-phosphate-mannose assembly, which is the specific complaint raised in go-annotation#5835 against S. pombe Dpm1 (O14466). Four sets are mechanistically unrelated and reveal how the rule was built: CS37 (5-aminolevulinate synthase) shares the Type I PLP fold with the serine palmitoyltransferases of CS18 and CS27; CS26 is phosphoglycerate kinase; CS12 is a PH-domain Ser/Thr protein kinase; CS43 is a lipid-BINDING lipoprotein, not a synthase. The apparent specificity of 51 named FunFams collapses onto only 26 CATH superfamilies, ten of which are reused across multiple condition sets, and in three cases same-superfamily neighbours disagree about whether a membrane lipid is being made at all. Two further conditions rest on the wrong domain entirely: CS16 and CS30 key on the cytochrome b5-like electron-donor module rather than a catalytic domain, and CS44 ("sphingomyelin synthase-related protein 1") is assigned to an Ets-1 winged-helix superfamily. Redundancy is substantial: KDSR is reached three ways (CS4, CS9, CS36) and DPM1 twice by two FunFams of one superfamily carrying identical labels (CS23, CS50). The 39 taxon conditions name 34 distinct taxa and 12 sets carry no taxon at all; scopes such as Homo for sphingosine kinase 2, Hominidae for SPT2 and Glires for B4GALT6 are annotation-bias fossils, while Bacteria and Eukaryota are wide open. The falcon deep research reached the same overall verdict independently, from the literature rather than from fold membership, and supplies the directional anchors for the hydrolases and for LpxC/LpxH; it was, however, told the pre-obsoletion label by this repository's own stale label cache and therefore reasoned throughout about a live term.
The rule's only consequent no longer exists in GO, so there is nothing to accept and no in-place edit that leaves the rule correct. The natural mechanical fix - remap GO:0046467 to its replaced_by term GO:0008610 - must be rejected: nine of the 51 condition sets (five catabolic, four mechanistically unrelated) would then produce annotations that are false rather than merely imprecise, and the remaining sound sets do not need a term as general as "lipid biosynthetic process" when pathway-specific children exist. Deprecation is also the honest response to go-annotation#5835. The obsoletion removed this rule's output without examining the reasoning that produced it, and the reporter's own follow-up ("oh no, some definitely are there now. Will reopen") is borne out: the same S. pombe Dpm1 protein still carries GO:0046474 from ARBA00028351 and GO:1901137 from ARBA00026302, making the same category error. The defensible core - sphingolipid, glycosphingolipid, lipid A, GPI-anchor and dolichyl-phosphate-mannose biosynthesis - should be re-derived as separate pathway-scoped rules with specific terms, not preserved by re-pointing this one.
BIOSYNTHETIC, TERM WRONG. IPR006634 (TLC domain) + IPR016439 (Lag1/Lac1) identifies chordate ceramide synthases CERS1-6, which acylate sphingoid bases in the ER. Direction is right; GO:0046513 ceramide biosynthetic process or GO:0030148 sphingolipid biosynthetic process is the term. The TLC/CLN8-homology family also contains functionally divergent members (CLN8-like), so the domain pair is suggestive rather than diagnostic.
CATABOLIC - WRONG DIRECTION. IPR008901 alkaline ceramidase (ACER1/2/3) hydrolyses ceramide to sphingosine. Ceramidases consume a membrane sphingolipid; feeding the salvage pathway does not make the hydrolase a participant in biosynthesis. Better: GO:0046514 ceramide catabolic process.
NON-DIAGNOSTIC. IPR001675 (GT29) + IPR051757 (beta-galactoside alpha2-3 sialyltransferases) is a family-level match. ST3GAL enzymes sialylate glycoprotein N-/O-glycans as well as glycolipid acceptors, so family membership alone does not establish a lipid acceptor. Only ganglioside-specific orthologues would justify GO:0001574/GO:0006688.
BIOSYNTHETIC, TERM WRONG. IPR002347 + IPR020904 + IPR045022 identifies KDSR (3-ketodihydrosphingosine reductase), the second step of de novo sphingolipid synthesis. Better: GO:0030148. Redundant with CS9 and CS36, which reach the same enzyme through CATH.
CATABOLIC - WRONG DIRECTION. IPR004843 + IPR008139 (saposin B) + IPR045473 is the acid sphingomyelinase (SMPD1) architecture, a lysosomal/secretory phospholipase C. Better: GO:0006685 sphingomyelin catabolic process. The deep research notes ASM cleaves more than 20 phospholipids in vitro, so the family is also less substrate-specific than its name suggests.
NON-DIAGNOSTIC. IPR007577 (DXD sugar-binding motif) + IPR051706 ('Glycosyltransferase domain-containing protein') across all Eukaryota. A DXD motif establishes a metal-dependent sugar transfer, not a lipid acceptor; the acceptor may equally be a protein, an N-/O-glycan or a wall polymer.
BIOSYNTHETIC, TERM WRONG. IPR004960 (LpxL/LpxM-type lipid A acyltransferases) acts in lipid A assembly, a genuine outer-membrane glycolipid. Better: GO:0009245 lipid A biosynthetic process. The 'Bacteria' scope is too broad: lipid A architecture and acyl-chain specificity vary widely across phyla and some taxa use non-homologous enzymes.
BIOSYNTHETIC, FRAGILE SIGNATURE. CATH names 1.10.10.60:FF:000020 'Ceramide synthase 5', but the superfamily is 'Homeodomain-like'. CerS2-6 do carry an N-terminal Hox-like domain, so the FunFam is plausible - but the condition keys on the homeobox rather than the TLC catalytic domain, which is a fragile basis for an enzymatic annotation. Better: GO:0046513.
BIOSYNTHETIC, TERM WRONG. KDSR FunFam in a NAD(P)-binding Rossmann superfamily (3.40.50.720). Correct enzyme, correct direction; better GO:0030148. Redundant with CS4 (InterPro route) and CS36 (Saccharomycotina subset of this set's Eukaryota scope).
MODIFICATION, NOT BIOSYNTHESIS. PagP transfers palmitate from a phospholipid onto lipid A in the outer membrane. This is lipid A tailoring in response to envelope stress, not core synthesis. Better: a lipid A modification term. 'Bacteria' scope is again too broad.
BIOSYNTHETIC, TERM WRONG. Fungal MIPC synthase subunit; mannosylinositol phosphorylceramide is the major complex sphingolipid of fungal membranes. Better: GO:0051999 mannosyl-inositol phosphorylceramide biosynthetic process, the exact biosynthetic term for this enzyme's product. Do not fall back on the metabolic parent GO:0006673 inositol phosphoceramide metabolic process, which repeats the over-generality this review criticises in the rule.
MECHANISTICALLY UNRELATED. 1.10.510.10 is the protein kinase catalytic domain and 2.30.29.30 is the PH/PTB domain, so this condition set describes a PH-domain-containing Ser/Thr protein kinase restricted to Primates. There is no route from that architecture to lipid biosynthesis. The taxon restriction cannot rescue a mechanistically unrelated signature.
BIOSYNTHETIC, TERM WRONG. 3.40.50.2000:FF:000033 is UGT8/CGT, which makes galactosylceramide. Better: GO:0006682 galactosylceramide biosynthetic process (not GO:0006679, which is the glucosylceramide term and belongs to CS40). The conjoined FF:000001 'UDP-glucuronosyltransferase' is the generic UGT FunFam of the same GT-B superfamily and contributes drug-glucuronidation enzymes, weakening the set.
BIOSYNTHETIC, TERM WRONG. Two FunFams of the GT-B superfamily both named for PimA, the alpha-(1-6)-phosphatidylinositol monomannoside mannosyltransferase that starts the PIM/LM/LAM branch of the actinobacterial envelope. Better: GO:0009247 glycolipid biosynthetic process. GO currently has no phosphatidylinositol mannoside biosynthesis child (the chemistry is represented only as CHEBI:59466), so a pathway-specific term would have to be requested rather than cited. Overlaps CS34.
CATABOLIC - WRONG DIRECTION. ASAH2 neutral ceramidase; the CATH superfamily is explicitly 'Neutral/alkaline non-lysosomal ceramidase, C-terminal domain'. Hydrolysis of ceramide opposes the biosynthetic assertion. Better: GO:0046514.
SIGNATURE IS THE WRONG DOMAIN. FA2H genuinely 2-hydroxylates sphingolipid acyl chains, but 3.10.120.10 is the cytochrome b5-like heme/steroid binding superfamily - the fused electron-donor module, not the di-iron catalytic domain. The condition therefore keys on a module shared with desaturases and many unrelated b5-domain proteins.
BIOSYNTHETIC, TERM WRONG. MGD3 makes monogalactosyldiacylglycerol, the dominant thylakoid membrane lipid. This is one of the clearest positives in the rule. Better: GO:0009247 glycolipid biosynthetic process or a galactolipid child.
BIOSYNTHETIC, TERM WRONG. SPT1 catalyses the committed first step of de novo sphingolipid synthesis. Better: GO:0030148. Note the superfamily, 3.40.640.10 'Type I PLP-dependent aspartate aminotransferase-like', is the same fold the rule follows into 5-aminolevulinate synthase at CS37.
CATABOLIC - WRONG DIRECTION. Sphingomyelin phosphodiesterase FunFam. Better: GO:0006685. Its superfamily 3.60.21.10 also supplies CS38 (LpxH, biosynthetic) and CS48 (unknown), so one metallophosphoesterase fold yields three different answers to whether a membrane lipid is being made.
PLAUSIBLE BUT UNVERIFIED. ST3GAL2 does contribute to ganglioside synthesis (GD1a/GM1b), but GT29 members act on glycoproteins too. Retain only with evidence that the orthologue's physiological acceptor is a glycolipid; then use GO:0001574 ganglioside biosynthetic process.
PLAUSIBLE BUT UNVERIFIED. B4GALT6 is implicated in lactosylceramide synthesis, but the B4GALT family has many non-lipid acceptors. If retained, GO:0006688 glycosphingolipid biosynthetic process.
WRONG PATHWAY. PIGN transfers ethanolamine phosphate onto the GPI precursor. GPI is a membrane glycolipid, so the annotation is not absurd, but the informative term is GO:0006506 GPI anchor biosynthetic process. Overlaps CS47 (same superfamily 3.40.720.10).
WRONG PATHWAY - THIS IS THE CURATOR'S COMPLAINT. DPM1 synthesises dolichyl-phosphate mannose, a lipid-linked sugar DONOR consumed by N-glycosylation, O- and C-mannosylation and GPI assembly; it is not a structural membrane lipid. go-annotation#5835 states 'Dolichol phosphate mannose (DPM) is not directly involved in membrane lipid biosynthesis, but rather in protein glycosylation processes'. UniProt already annotates the flagged protein O14466 to GO:0180047 dolichol phosphate mannose biosynthetic process, which is the correct term. Duplicated by CS50.
BIOSYNTHETIC, TERM WRONG. LpxC (both CATH domains required) performs the first committed step of lipid A biosynthesis. Strong, well-evidenced positive. Better: GO:0009245.
MODIFICATION / SPECIALISED PRODUCT. PNPLA1 is a neofunctionalised patatin-family member that makes epidermal omega-O-acylceramides. If retained, use an acylceramide biosynthesis term, never a generic lipid term propagated from the patatin fold - most PNPLA members are lipases. Euarchontoglires scope is an annotation-bias artefact.
MECHANISTICALLY UNRELATED. CATH names both FunFams 'Phosphoglycerate kinase' and the superfamily is 'Phosphoglycerate kinase, N-terminal domain'. PGK is a glycolytic enzyme; there is no basis for a lipid biosynthetic annotation in Streptophyta or anywhere else.
BIOSYNTHETIC BUT INCOHERENT LOGIC. 3.40.640.10:FF:000047 is SPT2, a genuine sphingolipid enzyme. Conjoining it with 3.90.1150.10:FF:000004 ('2-amino-3-ketobutyrate coenzyme A ligase') most likely reflects domain segmentation: 3.90.1150.10 is the small domain of the same Type I PLP fold, mislabelled after a different family member. The Hominidae restriction is meaningless for a pan-eukaryotic enzyme.
INTERCONVERSION, NOT BIOSYNTHESIS. Fungal dihydrosphingosine-1-phosphate phosphatase (LCB3/YSR2) dephosphorylates sphingoid base phosphates. Better: GO:0006665 sphingolipid metabolic process. Shares superfamily 1.20.144.10 with CS29 despite acting in a different kingdom and pathway.
MODIFICATION, NOT BIOSYNTHESIS. LpxE-type lipid A 1-phosphatase removes a phosphate from assembled lipid A. This is remodelling, typically for cationic-peptide resistance, not synthesis.
PLAUSIBLE, TERM WRONG. Plant Delta(8) desaturase acts on sphingolipid long-chain bases. As with CS16 the signature is the cytochrome b5-like module (3.10.120.10), not the catalytic domain, so substrate specificity is not established by the condition.
INTERCONVERSION / SIGNALLING. SPHK2 phosphorylates sphingosine to S1P, a signalling metabolite that is exported or degraded, not incorporated into membranes. The superfamily is 'Probable inorganic polyphosphate/ATP-NAD kinase; domain 1'. The 'Homo' taxon restriction on a pan-eukaryotic enzyme is an annotation-bias artefact rather than a biological claim.
PLAUSIBLE BUT UNVERIFIED. A Mycobacteriales FunFam of the GT-B superfamily named only 'Glycosyl transferase'. It may well be a PIM/LM/LAM enzyme, but the name does not establish it. Retain only with validated orthology.
NON-DIAGNOSTIC. 'Probable glycosyl transferase' across all Actinomycetota. A FunFam whose own name hedges cannot support a specific process assertion.
BIOSYNTHETIC, TERM WRONG. PimB-type phosphatidyl-myo-inositol mannosyltransferase in Actinomycetes; a real envelope glycolipid enzyme. Better: GO:0009247 glycolipid biosynthetic process. As noted for CS14, GO currently has no phosphatidylinositol mannoside biosynthesis child, so a pathway-specific term would have to be requested rather than cited. Overlaps CS14 and CS32 in the same superfamily.
PRECURSOR SUPPLY, NOT MEMBRANE LIPID SYNTHESIS. Fungal very-long-chain 3-oxoacyl-CoA reductase (KAR/Ifa38) acts in the VLCFA elongation cycle. VLCFAs feed sphingolipids among other fates. Better: GO:0042761 very long-chain fatty acid biosynthetic process.
BIOSYNTHETIC, TERM WRONG, AND REDUNDANT. Saccharomycotina KDSR (TSC10). Correct enzyme, but its taxon scope is a subset of CS9's Eukaryota scope and the enzyme is already reached by CS4. Third route to one enzyme.
MECHANISTICALLY UNRELATED - FOLD-LEVEL FALSE POSITIVE. 5-aminolevulinate synthase is a heme/tetrapyrrole biosynthesis enzyme. It shares the Type I PLP fold (3.40.640.10) with SPT1 (CS18) and SPT2 (CS27), which is exactly how the rule reached it. The FunFam is named 'mitochondrial' while the condition restricts to Bacteroidota - the taxon constraint conceals the error rather than catching it.
BIOSYNTHETIC DESPITE HYDROLYTIC CHEMISTRY. LpxH hydrolyses UDP-2,3-diacylglucosamine to lipid X, an obligate step inside the anabolic Raetz pathway. This is the counter-example that shows reaction chemistry does not settle direction - pathway position does. Better: GO:0009245.
PLAUSIBLE BUT UNVERIFIED. A cyanobacterial beta-(1-3)-glucosyltransferase FunFam; likely glycolipid synthesis but the acceptor is not established by the name.
BIOSYNTHETIC, TERM WRONG. UGCG/ceramide glucosyltransferase makes glucosylceramide, the root of most glycosphingolipids. Better: GO:0006679 glucosylceramide biosynthetic process, which is the exact term for this activity, rather than the parent GO:0006688.
BIOSYNTHETIC, TERM WRONG. Processive diacylglycerol beta-glycosyltransferase makes glycosyldiacylglycerols, the principal membrane glycolipids of Mycoplasmatota. Better: GO:0009247 glycolipid biosynthetic process.
PLAUSIBLE BUT UNVERIFIED. An arthropod alpha1,4-GalNAc transferase; may extend glycolipid glycans, but the GT-A family is not lipid-directed as a whole.
NOT A BIOSYNTHETIC ENZYME. The condition's own label says 'binding protein', and the superfamily is 'Dipeptide-binding Protein; Domain 1' - a periplasmic-binding-protein-like fold. LprG-type mycobacterial lipoproteins carry and traffic PIMs; ligand binding is not catalysis and cannot support a biosynthetic process term.
SIGNATURE IS SUSPECT. SMSr/SAMD8 makes ceramide phosphoethanolamine, so the biology could support a sphingolipid term - but CATH places 1.10.150.50 in the 'Transcription Factor, Ets-1' winged-helix superfamily, whereas the SMS catalytic domain is a lipid-phosphate-phosphatase-like fold. This looks like a partial or spurious domain assignment and should not be used as a condition without checking.
CATABOLIC - WRONG DIRECTION. GBA lysosomal acid glucosylceramidase (a TIM-barrel glycosidase) removes glucose from glucosylceramide. Better: GO:0006680 glucosylceramide catabolic process.
WRONG PATHWAY. PIGL de-N-acetylates GlcNAc-PI, an early committed step of GPI assembly. Better: GO:0006506 GPI anchor biosynthetic process.
WRONG PATHWAY. PIGO/GPI7-type GPI ethanolamine phosphate transferase. Better: GO:0006506. Overlaps CS22 in superfamily 3.40.720.10.
NON-DIAGNOSTIC. 'Putative metallophosphoesterase 1' in the calcineurin-like superfamily 3.60.21.10, which also contains the catabolic SMPD (CS19) and the biosynthetic LpxH (CS38). If this is MPPE1/PGAP5 it is a GPI-remodelling enzyme, but the FunFam name does not establish that, and the superfamily spans many non-lipid functions.
PLAUSIBLE BUT UNVERIFIED. ST6GALNAC6-like GT29 sialyltransferase. Same caveat as CS3 and CS20: sialyltransferase family membership does not establish a glycolipid acceptor.
WRONG PATHWAY AND REDUNDANT WITH CS23. A second DPM1 FunFam of the same superfamily, carrying the identical CATH label. Both should collapse into one condition pointing at GO:0180047 dolichol phosphate mannose biosynthetic process.
NON-DIAGNOSTIC. An unqualified 'Hexosyltransferase' FunFam with no taxon constraint. A generic sugar-transfer family cannot establish that the acceptor is a lipid.
51 OR-ed condition sets for one consequent, against a working guideline of 12. The repository's own analysis tooling refuses to process it: 'just analyze-rule ARBA00028538' aborts with 'Rule ARBA00028538 has 51 condition sets, which exceeds the maximum of 12', so no pairwise overlap or heatmap data could be generated and the condition-level audit had to be done by hand. Complexity here is not the price of biological breadth; it is what allows five catabolic and four unrelated condition sets to sit unnoticed alongside genuine sphingolipid and lipid A enzymes, because in an OR the weakest condition sets the rule's real specificity.
Support is bimodal and the rule as a whole is governed by its weakest sets. The literature is strong and specific for a core: serine palmitoyltransferase, KDSR and the ceramide synthases as the de novo sphingolipid pathway; LpxC and LpxH as committed steps of lipid A biosynthesis; the GPI enzymes and Dol-P-Man donor chemistry. For five condition sets the literature is not merely thin but directly contradictory - acid sphingomyelinase, alkaline and neutral ceramidase and lysosomal glucosylceramidase are characterised as catabolic enzymes. For CS12, CS26, CS37 and CS43 no literature connects the family to membrane lipid synthesis at all, and their presence is explained by shared folds rather than shared function. The overall grade is therefore WEAK rather than STRONG or CONTRADICTED: a substantial well-evidenced core exists, but it cannot be inferred from the rule, only from auditing it set by set.
The 51 named FunFams occupy only 26 distinct CATH superfamilies, and ten superfamilies are reused across two or more condition sets: 3.40.50.2000 across six (CS13, 14, 17, 32, 33, 34), 3.90.550.10 across six (CS21, 23, 39, 40, 41, 50), and 3.40.50.720, 3.40.640.10 and 3.60.21.10 across three each. Outright redundancy: KDSR is matched three times, by InterPro (CS4) and by two FunFams of one Rossmann superfamily (CS9 Eukaryota, CS36 Saccharomycotina - a taxon subset of CS9); DPM1 is matched twice by CS23 and CS50, two FunFams of 3.90.550.10 carrying the identical CATH label; GPI ethanolamine phosphate transferases appear as CS22 and CS47 in 3.40.720.10. Worse than the duplication is the disagreement: 3.60.21.10 supplies the catabolic sphingomyelin phosphodiesterase (CS19), the biosynthetic LpxH (CS38) and an unidentified metallophosphoesterase (CS48), and 3.40.640.10 supplies both serine palmitoyltransferases and 5-aminolevulinate synthase. Quantitative pairwise overlap could not be computed because the analysis tool declines rules with more than 12 condition sets; the superfamily census in scripts/census_arba00028538.py --cath is the substitute.
The consequent is obsolete. GO:0046467 was obsoleted on 2025-12-09 and renamed "obsolete membrane lipid biosynthetic process", with replaced_by GO:0008610 lipid biosynthetic process; the stated reason is that the term "groups lipids by their localization rather than by their metabolic function, which is problematic since most lipids can function in multiple locations". QuickGO returns 0 annotations to it from any source. Beyond the obsoletion the term choice was already wrong in two distinct ways: directionally, for the five hydrolase condition sets, where the correct terms lie in the catabolic subtree; and by pathway, for the six GPI/dolichol sets, where GPI anchor biosynthesis (GO:0006506) and dolichol phosphate mannose biosynthesis (GO:0180047) are the right terms - the latter being exactly the point made in go-annotation#5835 about Dpm1, and the term UniProt already assigns to the flagged protein O14466. For the ~26 defensible sets the term was merely far too broad; GO:0030148, GO:0046513, GO:0006688 and GO:0009245 are available. Because the errors are of three different kinds, no single replacement term fixes the rule.
The 39 taxon conditions name 34 distinct taxa, and 12 of the 51 sets carry no taxon constraint at all - a distribution that reflects where training annotations happened to sit rather than any phylogenetic judgment. Most scopes are far narrower than the biology: Homo for sphingosine kinase 2 (CS31) and Hominidae for serine palmitoyltransferase 2 (CS27), both pan-eukaryotic enzymes; Glires for B4GALT6 (CS21); Euarchontoglires for ST3GAL (CS3) and PNPLA1 (CS25); Primates for the protein kinase (CS12). A handful run the other way and are too broad: Bacteria for the lipid A acyltransferases (CS7) and PagP (CS10), despite well-documented plasticity of lipid A architecture across bacterial phyla, and Eukaryota for the generic glycosyltransferase set (CS6). In CS37 the taxon constraint is actively misleading rather than merely arbitrary: a FunFam named "5-aminolevulinate synthase, mitochondrial" is restricted to Bacteroidota, and the restriction hides the fold-level false positive instead of catching it. TOO_NARROW is recorded because arbitrary narrow clades are the dominant pattern, but the underlying problem is that taxon conditions here are doing filtering work, not encoding conservation.
GO:0046467 was obsoleted on 2025-12-09 with replaced_by GO:0008610; QuickGO returns 0 annotations to it from any source, so the rule is inert.
The rule's modifiedDate of 2025-12-15 is six days after the obsoletion, so it is inert but not retired.
The 51 FunFams occupy only 26 CATH superfamilies; 3.40.640.10 supplies both serine palmitoyltransferases and 5-aminolevulinate synthase, and 3.60.21.10 supplies both the catabolic SMPD and the biosynthetic LpxH.
At most 26 of 51 condition sets describe an enzyme that builds a membrane lipid; 9 would become false rather than merely vague if the consequent were remapped to GO:0008610.
Independent per-condition audit of all 51 sets concludes the rule is not valid in its present OR-ed form and should be split by pathway and reaction direction.
Flags CS12, CS26, CS37 and CS43 as mechanistically unrelated, and CS2, CS5, CS15, CS19, CS45 as catabolic - matching the fold-based analysis derived independently from CATH superfamily membership.
Did not know GO:0046467 is obsolete: the prompt supplied the pre-obsoletion label from this repository's stale rules/_labels.json cache, so its recommendation to rely on GO ancestor propagation to GO:0046467 is no longer available.
Its executive summary enumerates 14 removal candidates while claiming 12, and cites Grabner et al. 2022 (adipocyte lipolysis inhibitors) for PNPLA1's role in epidermal acylceramide synthesis; the per-set verdicts are sound but the roll-up counts and that one citation should not be quoted.
Fetches the rule from rest.uniprot.org/arba, the term from QuickGO, and the CATH superfamily names live; all numbers in the analysis are reproducible from this script.
S. pombe Dpm1 (O14466) no longer carries ARBA00028538, but still carries GO:0046474 from ARBA00028351 and GO:1901137 from ARBA00026302, which make the same category error.
Frames acid sphingomyelinase as a catabolic enzyme: "sphingolipids of mammalian cells are degraded in a stepwise manner without an effective bypass in human lysosomal catabolism", and "The recessively inherited ASMD mainly leads to impaired cellular SM catabolism causing progressive SM accumulation in the cellular membranes." This is the direct basis for calling CS5 (SMPD1 architecture) and CS19 (sphingomyelin phosphodiesterase FunFam) directionally wrong.
"The essential enzyme LpxC catalyzes the first committed step of LPS synthesis and thus represents a promising antibiotic target", and "Instead of the first enzyme LpxA, which catalyzes a reversible reaction, the second enzyme, LpxC, is the main driver in this process as it catalyzes the first committed step". This supports CS24 (LpxC) as a genuine biosynthetic condition and, with it, the general point that hydrolytic chemistry inside an anabolic pathway (CS38, LpxH) is still biosynthesis.
Describes the GPI backbone as "ethanolamine phosphate (EtNP), three mannoses (Mans), one non-N-acetylated glucosamine, and inositol phospholipid" whose "Fatty chains of inositol phospholipids are inserted into the outer leaflet of the plasma membrane". GPI is therefore a membrane glycolipid, which is why CS22/CS46/CS47 are not absurd - but the informative term is GO:0006506, not a generic lipid-biosynthesis term.
The FunFam method paper states that "the functional annotations provided by FunFams are found to be more precise compared with those generated by other domain-based resources". FunFam membership is nonetheless a prediction of functional coherence within a superfamily, not a measurement of a specific reaction - which is what CS8 (a ceramide synthase FunFam inside a Homeodomain-like superfamily), CS16/CS30 (cytochrome b5-like superfamily) and CS44 (Ets-1 winged-helix superfamily) illustrate.
States plainly that "ceramidases catalyze the degradation of ceramide to sphingosine", naming "alkaline ceramidases (encoded by ACER1, ACER2 and ACER3), acidic ceramidase (encoded by ASAH1) and neutral ceramidase (encoded by ASAH2)". This covers CS2 (ACER InterPro entry) and CS15 (ASAH2 FunFam), both of which the rule assigns to a biosynthetic term.
"Lipid A, the conserved, lipid component of LPS, anchors LPS to the outer membrane", and while "E. coli acyl transferases are chain length specific, orthologs from other Gram-negative bacteria are more tolerant, yielding a heterogeneous Lipid A composition of the outer membrane." The first clause supports CS7/CS24/CS38 as membrane-lipid biosynthesis; the second is why the "Bacteria" taxon scope on CS7 and CS10 is too broad.
id: ARBA00028538
description: >-
Assigns GO:0046467 to proteins matching any of 51 alternative condition sets built
from 14 InterPro entries, 51 CATH FunFams and 39 taxon constraints. GO obsoleted
GO:0046467 ("membrane lipid biosynthetic process") on 2025-12-09, replacing it with
GO:0008610 lipid biosynthetic process, so the rule now emits no annotations at all -
yet UniProt still serves it, and its last modification date of 2025-12-15 is six days
after the obsoletion. The antecedents are heterogeneous: at most 26 of the 51 sets
describe enzymes that build a membrane lipid, while the rest capture sphingolipid
hydrolases, lipid A and GPI remodelling enzymes, GPI/dolichol-donor assembly,
non-diagnostic glycosyltransferase families, and four mechanistically unrelated
proteins reached through promiscuous folds (a PH-domain protein kinase,
phosphoglycerate kinase, 5-aminolevulinate synthase, and a lipid-binding lipoprotein).
status: COMPLETE
rule_type: ARBA
rule:
rule_id: ARBA00028538
condition_sets:
- number: 1
conditions:
- condition_type: INTERPRO
value: IPR006634
curie: InterPro:IPR006634
label: "TRAM/LAG1/CLN8 homology domain"
negated: false
- condition_type: INTERPRO
value: IPR016439
curie: InterPro:IPR016439
label: "Sphingosine N-acyltransferase Lag1/Lac1-like"
negated: false
- condition_type: TAXON
value: Chordata
curie: NCBITaxon:7711
label: "Chordata"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. IPR006634 (TLC domain) + IPR016439 (Lag1/Lac1) identifies chordate ceramide synthases CERS1-6, which acylate sphingoid bases in the ER. Direction is right; GO:0046513 ceramide biosynthetic process or GO:0030148 sphingolipid biosynthetic process is the term. The TLC/CLN8-homology family also contains functionally divergent members (CLN8-like), so the domain pair is suggestive rather than diagnostic."
- number: 2
conditions:
- condition_type: INTERPRO
value: IPR008901
curie: InterPro:IPR008901
label: "Alkaline ceramidase"
negated: false
- condition_type: TAXON
value: Craniata
curie: NCBITaxon:89593
label: "Craniata"
negated: false
notes: "CATABOLIC - WRONG DIRECTION. IPR008901 alkaline ceramidase (ACER1/2/3) hydrolyses ceramide to sphingosine. Ceramidases consume a membrane sphingolipid; feeding the salvage pathway does not make the hydrolase a participant in biosynthesis. Better: GO:0046514 ceramide catabolic process."
- number: 3
conditions:
- condition_type: INTERPRO
value: IPR001675
curie: InterPro:IPR001675
label: "Glycosyl transferase family 29"
negated: false
- condition_type: INTERPRO
value: IPR051757
curie: InterPro:IPR051757
label: "Beta-galactoside alpha2-3 sialyltransferases"
negated: false
- condition_type: TAXON
value: Euarchontoglires
curie: NCBITaxon:314146
label: "Euarchontoglires"
negated: false
notes: "NON-DIAGNOSTIC. IPR001675 (GT29) + IPR051757 (beta-galactoside alpha2-3 sialyltransferases) is a family-level match. ST3GAL enzymes sialylate glycoprotein N-/O-glycans as well as glycolipid acceptors, so family membership alone does not establish a lipid acceptor. Only ganglioside-specific orthologues would justify GO:0001574/GO:0006688."
- number: 4
conditions:
- condition_type: INTERPRO
value: IPR002347
curie: InterPro:IPR002347
label: "Short-chain dehydrogenase/reductase SDR"
negated: false
- condition_type: INTERPRO
value: IPR020904
curie: InterPro:IPR020904
label: "Short-chain dehydrogenase/reductase, conserved site"
negated: false
- condition_type: INTERPRO
value: IPR045022
curie: InterPro:IPR045022
label: "3-ketodihydrosphingosine reductase KDSR-like"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. IPR002347 + IPR020904 + IPR045022 identifies KDSR (3-ketodihydrosphingosine reductase), the second step of de novo sphingolipid synthesis. Better: GO:0030148. Redundant with CS9 and CS36, which reach the same enzyme through CATH."
- number: 5
conditions:
- condition_type: INTERPRO
value: IPR004843
curie: InterPro:IPR004843
label: "Calcineurin-like, phosphoesterase domain"
negated: false
- condition_type: INTERPRO
value: IPR008139
curie: InterPro:IPR008139
label: "Saposin B type domain"
negated: false
- condition_type: INTERPRO
value: IPR045473
curie: InterPro:IPR045473
label: "Sphingomyelin phosphodiesterase, C-terminal domain"
negated: false
notes: "CATABOLIC - WRONG DIRECTION. IPR004843 + IPR008139 (saposin B) + IPR045473 is the acid sphingomyelinase (SMPD1) architecture, a lysosomal/secretory phospholipase C. Better: GO:0006685 sphingomyelin catabolic process. The deep research notes ASM cleaves more than 20 phospholipids in vitro, so the family is also less substrate-specific than its name suggests."
- number: 6
conditions:
- condition_type: INTERPRO
value: IPR007577
curie: InterPro:IPR007577
label: "Glycosyltransferase, DXD sugar-binding motif"
negated: false
- condition_type: INTERPRO
value: IPR051706
curie: InterPro:IPR051706
label: "Glycosyltransferase domain-containing protein"
negated: false
- condition_type: TAXON
value: Eukaryota
curie: NCBITaxon:2759
label: "Eukaryota"
negated: false
notes: "NON-DIAGNOSTIC. IPR007577 (DXD sugar-binding motif) + IPR051706 ('Glycosyltransferase domain-containing protein') across all Eukaryota. A DXD motif establishes a metal-dependent sugar transfer, not a lipid acceptor; the acceptor may equally be a protein, an N-/O-glycan or a wall polymer."
- number: 7
conditions:
- condition_type: INTERPRO
value: IPR004960
curie: InterPro:IPR004960
label: "Bacterial lipid A biosynthesis acyltransferase"
negated: false
- condition_type: TAXON
value: Bacteria
curie: NCBITaxon:2
label: "Bacteria"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. IPR004960 (LpxL/LpxM-type lipid A acyltransferases) acts in lipid A assembly, a genuine outer-membrane glycolipid. Better: GO:0009245 lipid A biosynthetic process. The 'Bacteria' scope is too broad: lipid A architecture and acyl-chain specificity vary widely across phyla and some taxa use non-homologous enzymes."
- number: 8
conditions:
- condition_type: FUNFAM
value: 1.10.10.60:FF:000020
curie: CATH.FunFam:1.10.10.60:FF:000020
label: "Ceramide synthase 5"
negated: false
- condition_type: TAXON
value: Metazoa
curie: NCBITaxon:33208
label: "Metazoa"
negated: false
notes: "BIOSYNTHETIC, FRAGILE SIGNATURE. CATH names 1.10.10.60:FF:000020 'Ceramide synthase 5', but the superfamily is 'Homeodomain-like'. CerS2-6 do carry an N-terminal Hox-like domain, so the FunFam is plausible - but the condition keys on the homeobox rather than the TLC catalytic domain, which is a fragile basis for an enzymatic annotation. Better: GO:0046513."
- number: 9
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000165
curie: CATH.FunFam:3.40.50.720:FF:000165
label: "3-ketodihydrosphingosine reductase"
negated: false
- condition_type: TAXON
value: Eukaryota
curie: NCBITaxon:2759
label: "Eukaryota"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. KDSR FunFam in a NAD(P)-binding Rossmann superfamily (3.40.50.720). Correct enzyme, correct direction; better GO:0030148. Redundant with CS4 (InterPro route) and CS36 (Saccharomycotina subset of this set's Eukaryota scope)."
- number: 10
conditions:
- condition_type: FUNFAM
value: 2.40.160.20:FF:000002
curie: CATH.FunFam:2.40.160.20:FF:000002
label: "Lipid A palmitoyltransferase PagP"
negated: false
- condition_type: TAXON
value: Bacteria
curie: NCBITaxon:2
label: "Bacteria"
negated: false
notes: "MODIFICATION, NOT BIOSYNTHESIS. PagP transfers palmitate from a phospholipid onto lipid A in the outer membrane. This is lipid A tailoring in response to envelope stress, not core synthesis. Better: a lipid A modification term. 'Bacteria' scope is again too broad."
- number: 11
conditions:
- condition_type: FUNFAM
value: 3.90.550.20:FF:000001
curie: CATH.FunFam:3.90.550.20:FF:000001
label: "MIPC synthase subunit (SurA)"
negated: false
- condition_type: TAXON
value: Fungi
curie: NCBITaxon:4751
label: "Fungi"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. Fungal MIPC synthase subunit; mannosylinositol phosphorylceramide is the major complex sphingolipid of fungal membranes. Better: GO:0051999 mannosyl-inositol phosphorylceramide biosynthetic process, the exact biosynthetic term for this enzyme's product. Do not fall back on the metabolic parent GO:0006673 inositol phosphoceramide metabolic process, which repeats the over-generality this review criticises in the rule."
- number: 12
conditions:
- condition_type: FUNFAM
value: 1.10.510.10:FF:000151
curie: CATH.FunFam:1.10.510.10:FF:000151
label: "Serine/threonine-protein kinase"
negated: false
- condition_type: FUNFAM
value: 2.30.29.30:FF:000056
curie: CATH.FunFam:2.30.29.30:FF:000056
label: "Serine/threonine-protein kinase"
negated: false
- condition_type: TAXON
value: Primates
curie: NCBITaxon:9443
label: "Primates"
negated: false
notes: "MECHANISTICALLY UNRELATED. 1.10.510.10 is the protein kinase catalytic domain and 2.30.29.30 is the PH/PTB domain, so this condition set describes a PH-domain-containing Ser/Thr protein kinase restricted to Primates. There is no route from that architecture to lipid biosynthesis. The taxon restriction cannot rescue a mechanistically unrelated signature."
- number: 13
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000001
curie: CATH.FunFam:3.40.50.2000:FF:000001
label: "UDP-glucuronosyltransferase"
negated: false
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000033
curie: CATH.FunFam:3.40.50.2000:FF:000033
label: "2-hydroxyacylsphingosine 1-beta-galactosyltransferase"
negated: false
- condition_type: TAXON
value: Chordata
curie: NCBITaxon:7711
label: "Chordata"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. 3.40.50.2000:FF:000033 is UGT8/CGT, which makes galactosylceramide. Better: GO:0006682 galactosylceramide biosynthetic process (not GO:0006679, which is the glucosylceramide term and belongs to CS40). The conjoined FF:000001 'UDP-glucuronosyltransferase' is the generic UGT FunFam of the same GT-B superfamily and contributes drug-glucuronidation enzymes, weakening the set."
- number: 14
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000069
curie: CATH.FunFam:3.40.50.2000:FF:000069
label: "Alpha-(1-6)-phosphatidylinositol monomannoside mannosyltransferase"
negated: false
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000115
curie: CATH.FunFam:3.40.50.2000:FF:000115
label: "Alpha-(1-6)-phosphatidylinositol monomannoside mannosyltransferase"
negated: false
- condition_type: TAXON
value: Bacillati
curie: NCBITaxon:1783272
label: "Bacillati"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. Two FunFams of the GT-B superfamily both named for PimA, the alpha-(1-6)-phosphatidylinositol monomannoside mannosyltransferase that starts the PIM/LM/LAM branch of the actinobacterial envelope. Better: GO:0009247 glycolipid biosynthetic process. GO currently has no phosphatidylinositol mannoside biosynthesis child (the chemistry is represented only as CHEBI:59466), so a pathway-specific term would have to be requested rather than cited. Overlaps CS34."
- number: 15
conditions:
- condition_type: FUNFAM
value: 2.60.40.2300:FF:000001
curie: CATH.FunFam:2.60.40.2300:FF:000001
label: "N-acylsphingosine amidohydrolase 2"
negated: false
- condition_type: TAXON
value: Mammalia
curie: NCBITaxon:40674
label: "Mammalia"
negated: false
notes: "CATABOLIC - WRONG DIRECTION. ASAH2 neutral ceramidase; the CATH superfamily is explicitly 'Neutral/alkaline non-lysosomal ceramidase, C-terminal domain'. Hydrolysis of ceramide opposes the biosynthetic assertion. Better: GO:0046514."
- number: 16
conditions:
- condition_type: FUNFAM
value: 3.10.120.10:FF:000011
curie: CATH.FunFam:3.10.120.10:FF:000011
label: "Fatty acid 2-hydroxylase"
negated: false
- condition_type: TAXON
value: Craniata
curie: NCBITaxon:89593
label: "Craniata"
negated: false
notes: "SIGNATURE IS THE WRONG DOMAIN. FA2H genuinely 2-hydroxylates sphingolipid acyl chains, but 3.10.120.10 is the cytochrome b5-like heme/steroid binding superfamily - the fused electron-donor module, not the di-iron catalytic domain. The condition therefore keys on a module shared with desaturases and many unrelated b5-domain proteins."
- number: 17
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000111
curie: CATH.FunFam:3.40.50.2000:FF:000111
label: "Monogalactosyldiacylglycerol synthase 3, chloroplastic"
negated: false
- condition_type: TAXON
value: Viridiplantae
curie: NCBITaxon:33090
label: "Viridiplantae"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. MGD3 makes monogalactosyldiacylglycerol, the dominant thylakoid membrane lipid. This is one of the clearest positives in the rule. Better: GO:0009247 glycolipid biosynthetic process or a galactolipid child."
- number: 18
conditions:
- condition_type: FUNFAM
value: 3.40.640.10:FF:000049
curie: CATH.FunFam:3.40.640.10:FF:000049
label: "serine palmitoyltransferase 1 isoform X1"
negated: false
- condition_type: TAXON
value: Vertebrata
curie: NCBITaxon:7742
label: "Vertebrata"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. SPT1 catalyses the committed first step of de novo sphingolipid synthesis. Better: GO:0030148. Note the superfamily, 3.40.640.10 'Type I PLP-dependent aspartate aminotransferase-like', is the same fold the rule follows into 5-aminolevulinate synthase at CS37."
- number: 19
conditions:
- condition_type: FUNFAM
value: 3.60.21.10:FF:000045
curie: CATH.FunFam:3.60.21.10:FF:000045
label: "Sphingomyelin phosphodiesterase"
negated: false
- condition_type: TAXON
value: Euteleostomi
curie: NCBITaxon:117571
label: "Euteleostomi"
negated: false
notes: "CATABOLIC - WRONG DIRECTION. Sphingomyelin phosphodiesterase FunFam. Better: GO:0006685. Its superfamily 3.60.21.10 also supplies CS38 (LpxH, biosynthetic) and CS48 (unknown), so one metallophosphoesterase fold yields three different answers to whether a membrane lipid is being made."
- number: 20
conditions:
- condition_type: FUNFAM
value: 3.90.1480.20:FF:000002
curie: CATH.FunFam:3.90.1480.20:FF:000002
label: "CMP-N-acetylneuraminate-beta-galactosamide- alpha-2,3-sialyltransferase 2"
negated: false
- condition_type: TAXON
value: Eutheria
curie: NCBITaxon:9347
label: "Eutheria"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. ST3GAL2 does contribute to ganglioside synthesis (GD1a/GM1b), but GT29 members act on glycoproteins too. Retain only with evidence that the orthologue's physiological acceptor is a glycolipid; then use GO:0001574 ganglioside biosynthetic process."
- number: 21
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000037
curie: CATH.FunFam:3.90.550.10:FF:000037
label: "Beta-1,4-galactosyltransferase 6"
negated: false
- condition_type: TAXON
value: Glires
curie: NCBITaxon:314147
label: "Glires"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. B4GALT6 is implicated in lactosylceramide synthesis, but the B4GALT family has many non-lipid acceptors. If retained, GO:0006688 glycosphingolipid biosynthetic process."
- number: 22
conditions:
- condition_type: FUNFAM
value: 3.40.720.10:FF:000015
curie: CATH.FunFam:3.40.720.10:FF:000015
label: "GPI ethanolamine phosphate transferase 1"
negated: false
notes: "WRONG PATHWAY. PIGN transfers ethanolamine phosphate onto the GPI precursor. GPI is a membrane glycolipid, so the annotation is not absurd, but the informative term is GO:0006506 GPI anchor biosynthetic process. Overlaps CS47 (same superfamily 3.40.720.10)."
- number: 23
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000119
curie: CATH.FunFam:3.90.550.10:FF:000119
label: "Dolichol-phosphate mannosyltransferase subunit 1"
negated: false
notes: "WRONG PATHWAY - THIS IS THE CURATOR'S COMPLAINT. DPM1 synthesises dolichyl-phosphate mannose, a lipid-linked sugar DONOR consumed by N-glycosylation, O- and C-mannosylation and GPI assembly; it is not a structural membrane lipid. go-annotation#5835 states 'Dolichol phosphate mannose (DPM) is not directly involved in membrane lipid biosynthesis, but rather in protein glycosylation processes'. UniProt already annotates the flagged protein O14466 to GO:0180047 dolichol phosphate mannose biosynthetic process, which is the correct term. Duplicated by CS50."
- number: 24
conditions:
- condition_type: FUNFAM
value: 3.30.1700.10:FF:000001
curie: CATH.FunFam:3.30.1700.10:FF:000001
label: "UDP-3-O-acyl-N-acetylglucosamine deacetylase"
negated: false
- condition_type: FUNFAM
value: 3.30.230.20:FF:000001
curie: CATH.FunFam:3.30.230.20:FF:000001
label: "UDP-3-O-acyl-N-acetylglucosamine deacetylase"
negated: false
- condition_type: TAXON
value: Pseudomonadati
curie: NCBITaxon:3379134
label: "Pseudomonadati"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. LpxC (both CATH domains required) performs the first committed step of lipid A biosynthesis. Strong, well-evidenced positive. Better: GO:0009245."
- number: 25
conditions:
- condition_type: FUNFAM
value: 3.40.1090.10:FF:000014
curie: CATH.FunFam:3.40.1090.10:FF:000014
label: "Patatin like phospholipase domain containing 1"
negated: false
- condition_type: FUNFAM
value: 3.40.1090.10:FF:000016
curie: CATH.FunFam:3.40.1090.10:FF:000016
label: "Patatin like phospholipase domain containing 1"
negated: false
- condition_type: TAXON
value: Euarchontoglires
curie: NCBITaxon:314146
label: "Euarchontoglires"
negated: false
notes: "MODIFICATION / SPECIALISED PRODUCT. PNPLA1 is a neofunctionalised patatin-family member that makes epidermal omega-O-acylceramides. If retained, use an acylceramide biosynthesis term, never a generic lipid term propagated from the patatin fold - most PNPLA members are lipases. Euarchontoglires scope is an annotation-bias artefact."
- number: 26
conditions:
- condition_type: FUNFAM
value: 3.40.50.1260:FF:000007
curie: CATH.FunFam:3.40.50.1260:FF:000007
label: "Phosphoglycerate kinase"
negated: false
- condition_type: FUNFAM
value: 3.40.50.1260:FF:000014
curie: CATH.FunFam:3.40.50.1260:FF:000014
label: "Phosphoglycerate kinase"
negated: false
- condition_type: TAXON
value: Streptophyta
curie: NCBITaxon:35493
label: "Streptophyta"
negated: false
notes: "MECHANISTICALLY UNRELATED. CATH names both FunFams 'Phosphoglycerate kinase' and the superfamily is 'Phosphoglycerate kinase, N-terminal domain'. PGK is a glycolytic enzyme; there is no basis for a lipid biosynthetic annotation in Streptophyta or anywhere else."
- number: 27
conditions:
- condition_type: FUNFAM
value: 3.40.640.10:FF:000047
curie: CATH.FunFam:3.40.640.10:FF:000047
label: "serine palmitoyltransferase 2 isoform X1"
negated: false
- condition_type: FUNFAM
value: 3.90.1150.10:FF:000004
curie: CATH.FunFam:3.90.1150.10:FF:000004
label: "2-amino-3-ketobutyrate coenzyme A ligase"
negated: false
- condition_type: TAXON
value: Hominidae
curie: NCBITaxon:9604
label: "Hominidae"
negated: false
notes: "BIOSYNTHETIC BUT INCOHERENT LOGIC. 3.40.640.10:FF:000047 is SPT2, a genuine sphingolipid enzyme. Conjoining it with 3.90.1150.10:FF:000004 ('2-amino-3-ketobutyrate coenzyme A ligase') most likely reflects domain segmentation: 3.90.1150.10 is the small domain of the same Type I PLP fold, mislabelled after a different family member. The Hominidae restriction is meaningless for a pan-eukaryotic enzyme."
- number: 28
conditions:
- condition_type: FUNFAM
value: 1.20.144.10:FF:000034
curie: CATH.FunFam:1.20.144.10:FF:000034
label: "Dihydrosphingosine-1-phosphate phosphatase"
negated: false
- condition_type: TAXON
value: Dikarya
curie: NCBITaxon:451864
label: "Dikarya"
negated: false
notes: "INTERCONVERSION, NOT BIOSYNTHESIS. Fungal dihydrosphingosine-1-phosphate phosphatase (LCB3/YSR2) dephosphorylates sphingoid base phosphates. Better: GO:0006665 sphingolipid metabolic process. Shares superfamily 1.20.144.10 with CS29 despite acting in a different kingdom and pathway."
- number: 29
conditions:
- condition_type: FUNFAM
value: 1.20.144.10:FF:000044
curie: CATH.FunFam:1.20.144.10:FF:000044
label: "Lipid A 1-phosphatase"
negated: false
- condition_type: TAXON
value: Campylobacterota
curie: NCBITaxon:29547
label: "Campylobacterota"
negated: false
notes: "MODIFICATION, NOT BIOSYNTHESIS. LpxE-type lipid A 1-phosphatase removes a phosphate from assembled lipid A. This is remodelling, typically for cationic-peptide resistance, not synthesis."
- number: 30
conditions:
- condition_type: FUNFAM
value: 3.10.120.10:FF:000021
curie: CATH.FunFam:3.10.120.10:FF:000021
label: "Delta(8)-fatty-acid desaturase 2"
negated: false
- condition_type: TAXON
value: Embryophyta
curie: NCBITaxon:3193
label: "Embryophyta"
negated: false
notes: "PLAUSIBLE, TERM WRONG. Plant Delta(8) desaturase acts on sphingolipid long-chain bases. As with CS16 the signature is the cytochrome b5-like module (3.10.120.10), not the catalytic domain, so substrate specificity is not established by the condition."
- number: 31
conditions:
- condition_type: FUNFAM
value: 3.40.50.10330:FF:000005
curie: CATH.FunFam:3.40.50.10330:FF:000005
label: "Sphingosine kinase 2"
negated: false
- condition_type: TAXON
value: Homo
curie: NCBITaxon:9605
label: "Homo"
negated: false
notes: "INTERCONVERSION / SIGNALLING. SPHK2 phosphorylates sphingosine to S1P, a signalling metabolite that is exported or degraded, not incorporated into membranes. The superfamily is 'Probable inorganic polyphosphate/ATP-NAD kinase; domain 1'. The 'Homo' taxon restriction on a pan-eukaryotic enzyme is an annotation-bias artefact rather than a biological claim."
- number: 32
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000072
curie: CATH.FunFam:3.40.50.2000:FF:000072
label: "Glycosyl transferase"
negated: false
- condition_type: TAXON
value: Mycobacteriales
curie: NCBITaxon:85007
label: "Mycobacteriales"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. A Mycobacteriales FunFam of the GT-B superfamily named only 'Glycosyl transferase'. It may well be a PIM/LM/LAM enzyme, but the name does not establish it. Retain only with validated orthology."
- number: 33
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000145
curie: CATH.FunFam:3.40.50.2000:FF:000145
label: "Probable glycosyl transferase"
negated: false
- condition_type: TAXON
value: Actinomycetota
curie: NCBITaxon:201174
label: "Actinomycetota"
negated: false
notes: "NON-DIAGNOSTIC. 'Probable glycosyl transferase' across all Actinomycetota. A FunFam whose own name hedges cannot support a specific process assertion."
- number: 34
conditions:
- condition_type: FUNFAM
value: 3.40.50.2000:FF:000207
curie: CATH.FunFam:3.40.50.2000:FF:000207
label: "Phosphatidyl-myo-inositol mannosyltransferase"
negated: false
- condition_type: TAXON
value: Actinomycetes
curie: NCBITaxon:1760
label: "Actinomycetes"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. PimB-type phosphatidyl-myo-inositol mannosyltransferase in Actinomycetes; a real envelope glycolipid enzyme. Better: GO:0009247 glycolipid biosynthetic process. As noted for CS14, GO currently has no phosphatidylinositol mannoside biosynthesis child, so a pathway-specific term would have to be requested rather than cited. Overlaps CS14 and CS32 in the same superfamily."
- number: 35
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000317
curie: CATH.FunFam:3.40.50.720:FF:000317
label: "Very-long-chain 3-oxoacyl-CoA reductase"
negated: false
- condition_type: TAXON
value: Ascomycota
curie: NCBITaxon:4890
label: "Ascomycota"
negated: false
notes: "PRECURSOR SUPPLY, NOT MEMBRANE LIPID SYNTHESIS. Fungal very-long-chain 3-oxoacyl-CoA reductase (KAR/Ifa38) acts in the VLCFA elongation cycle. VLCFAs feed sphingolipids among other fates. Better: GO:0042761 very long-chain fatty acid biosynthetic process."
- number: 36
conditions:
- condition_type: FUNFAM
value: 3.40.50.720:FF:000578
curie: CATH.FunFam:3.40.50.720:FF:000578
label: "3-ketodihydrosphingosine reductase"
negated: false
- condition_type: TAXON
value: Saccharomycotina
curie: NCBITaxon:147537
label: "Saccharomycotina"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG, AND REDUNDANT. Saccharomycotina KDSR (TSC10). Correct enzyme, but its taxon scope is a subset of CS9's Eukaryota scope and the enzyme is already reached by CS4. Third route to one enzyme."
- number: 37
conditions:
- condition_type: FUNFAM
value: 3.40.640.10:FF:000006
curie: CATH.FunFam:3.40.640.10:FF:000006
label: "5-aminolevulinate synthase, mitochondrial"
negated: false
- condition_type: TAXON
value: Bacteroidota
curie: NCBITaxon:976
label: "Bacteroidota"
negated: false
notes: "MECHANISTICALLY UNRELATED - FOLD-LEVEL FALSE POSITIVE. 5-aminolevulinate synthase is a heme/tetrapyrrole biosynthesis enzyme. It shares the Type I PLP fold (3.40.640.10) with SPT1 (CS18) and SPT2 (CS27), which is exactly how the rule reached it. The FunFam is named 'mitochondrial' while the condition restricts to Bacteroidota - the taxon constraint conceals the error rather than catching it."
- number: 38
conditions:
- condition_type: FUNFAM
value: 3.60.21.10:FF:000012
curie: CATH.FunFam:3.60.21.10:FF:000012
label: "UDP-2,3-diacylglucosamine hydrolase"
negated: false
- condition_type: TAXON
value: Pseudomonadota
curie: NCBITaxon:1224
label: "Pseudomonadota"
negated: false
notes: "BIOSYNTHETIC DESPITE HYDROLYTIC CHEMISTRY. LpxH hydrolyses UDP-2,3-diacylglucosamine to lipid X, an obligate step inside the anabolic Raetz pathway. This is the counter-example that shows reaction chemistry does not settle direction - pathway position does. Better: GO:0009245."
- number: 39
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000164
curie: CATH.FunFam:3.90.550.10:FF:000164
label: "Beta-(1-3)-glucosyl transferase"
negated: false
- condition_type: TAXON
value: Cyanobacteriota
curie: NCBITaxon:1117
label: "Cyanobacteriota"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. A cyanobacterial beta-(1-3)-glucosyltransferase FunFam; likely glycolipid synthesis but the acceptor is not established by the name."
- number: 40
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000207
curie: CATH.FunFam:3.90.550.10:FF:000207
label: "Ceramide glucosyltransferase 1"
negated: false
- condition_type: TAXON
value: Ecdysozoa
curie: NCBITaxon:1206794
label: "Ecdysozoa"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. UGCG/ceramide glucosyltransferase makes glucosylceramide, the root of most glycosphingolipids. Better: GO:0006679 glucosylceramide biosynthetic process, which is the exact term for this activity, rather than the parent GO:0006688."
- number: 41
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000323
curie: CATH.FunFam:3.90.550.10:FF:000323
label: "Processive diacylglycerol beta-glycosyltransferase"
negated: false
- condition_type: TAXON
value: Mycoplasmatota
curie: NCBITaxon:544448
label: "Mycoplasmatota"
negated: false
notes: "BIOSYNTHETIC, TERM WRONG. Processive diacylglycerol beta-glycosyltransferase makes glycosyldiacylglycerols, the principal membrane glycolipids of Mycoplasmatota. Better: GO:0009247 glycolipid biosynthetic process."
- number: 42
conditions:
- condition_type: FUNFAM
value: 3.90.550.20:FF:000013
curie: CATH.FunFam:3.90.550.20:FF:000013
label: "Alpha1,4-N-acetylgalactosaminyltransferase"
negated: false
- condition_type: TAXON
value: Arthropoda
curie: NCBITaxon:6656
label: "Arthropoda"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. An arthropod alpha1,4-GalNAc transferase; may extend glycolipid glycans, but the GT-A family is not lipid-directed as a whole."
- number: 43
conditions:
- condition_type: FUNFAM
value: 3.90.76.10:FF:000006
curie: CATH.FunFam:3.90.76.10:FF:000006
label: "Monoacyl phosphatidylinositol tetramannoside-binding protein"
negated: false
- condition_type: TAXON
value: Mycobacteriaceae
curie: NCBITaxon:1762
label: "Mycobacteriaceae"
negated: false
notes: "NOT A BIOSYNTHETIC ENZYME. The condition's own label says 'binding protein', and the superfamily is 'Dipeptide-binding Protein; Domain 1' - a periplasmic-binding-protein-like fold. LprG-type mycobacterial lipoproteins carry and traffic PIMs; ligand binding is not catalysis and cannot support a biosynthetic process term."
- number: 44
conditions:
- condition_type: FUNFAM
value: 1.10.150.50:FF:000037
curie: CATH.FunFam:1.10.150.50:FF:000037
label: "sphingomyelin synthase-related protein 1 isoform X1"
negated: false
notes: "SIGNATURE IS SUSPECT. SMSr/SAMD8 makes ceramide phosphoethanolamine, so the biology could support a sphingolipid term - but CATH places 1.10.150.50 in the 'Transcription Factor, Ets-1' winged-helix superfamily, whereas the SMS catalytic domain is a lipid-phosphate-phosphatase-like fold. This looks like a partial or spurious domain assignment and should not be used as a condition without checking."
- number: 45
conditions:
- condition_type: FUNFAM
value: 3.20.20.80:FF:000030
curie: CATH.FunFam:3.20.20.80:FF:000030
label: "Lysosomal acid glucosylceramidase"
negated: false
notes: "CATABOLIC - WRONG DIRECTION. GBA lysosomal acid glucosylceramidase (a TIM-barrel glycosidase) removes glucose from glucosylceramide. Better: GO:0006680 glucosylceramide catabolic process."
- number: 46
conditions:
- condition_type: FUNFAM
value: 3.40.50.10320:FF:000002
curie: CATH.FunFam:3.40.50.10320:FF:000002
label: "Probable N-acetylglucosaminyl-phosphatidylinositol de-N-acetylase"
negated: false
notes: "WRONG PATHWAY. PIGL de-N-acetylates GlcNAc-PI, an early committed step of GPI assembly. Better: GO:0006506 GPI anchor biosynthetic process."
- number: 47
conditions:
- condition_type: FUNFAM
value: 3.40.720.10:FF:000041
curie: CATH.FunFam:3.40.720.10:FF:000041
label: "GPI ethanolamine phosphate transferase 3"
negated: false
notes: "WRONG PATHWAY. PIGO/GPI7-type GPI ethanolamine phosphate transferase. Better: GO:0006506. Overlaps CS22 in superfamily 3.40.720.10."
- number: 48
conditions:
- condition_type: FUNFAM
value: 3.60.21.10:FF:000022
curie: CATH.FunFam:3.60.21.10:FF:000022
label: "Putative metallophosphoesterase 1"
negated: false
notes: "NON-DIAGNOSTIC. 'Putative metallophosphoesterase 1' in the calcineurin-like superfamily 3.60.21.10, which also contains the catabolic SMPD (CS19) and the biosynthetic LpxH (CS38). If this is MPPE1/PGAP5 it is a GPI-remodelling enzyme, but the FunFam name does not establish that, and the superfamily spans many non-lipid functions."
- number: 49
conditions:
- condition_type: FUNFAM
value: 3.90.1480.20:FF:000009
curie: CATH.FunFam:3.90.1480.20:FF:000009
label: "alpha-N-acetylgalactosaminide alpha-2,6-sialyltransferase 6 isoform X2"
negated: false
notes: "PLAUSIBLE BUT UNVERIFIED. ST6GALNAC6-like GT29 sialyltransferase. Same caveat as CS3 and CS20: sialyltransferase family membership does not establish a glycolipid acceptor."
- number: 50
conditions:
- condition_type: FUNFAM
value: 3.90.550.10:FF:000036
curie: CATH.FunFam:3.90.550.10:FF:000036
label: "Dolichol-phosphate mannosyltransferase subunit 1"
negated: false
notes: "WRONG PATHWAY AND REDUNDANT WITH CS23. A second DPM1 FunFam of the same superfamily, carrying the identical CATH label. Both should collapse into one condition pointing at GO:0180047 dolichol phosphate mannose biosynthetic process."
- number: 51
conditions:
- condition_type: FUNFAM
value: 3.90.550.50:FF:000019
curie: CATH.FunFam:3.90.550.50:FF:000019
label: "Hexosyltransferase"
negated: false
notes: "NON-DIAGNOSTIC. An unqualified 'Hexosyltransferase' FunFam with no taxon constraint. A generic sugar-transfer family cannot establish that the acceptor is a lipid."
go_annotations:
- go_id: GO:0046467
go_label: obsolete membrane lipid biosynthetic process
aspect: BP
entries: []
reviewed_protein_count: 0
unreviewed_protein_count: 0
created_date: '2021-10-20'
modified_date: '2025-12-15'
review_summary: >-
ARBA00028538 asserts a single consequent, GO:0046467, which the Gene Ontology obsoleted
on 2025-12-09 (tracker geneontology/go-ontology#26698) as part of retiring the whole
membrane-lipid grouping branch, on the ground that it grouped lipids by localization
rather than by metabolic function. QuickGO now returns zero annotations to GO:0046467
from any source, so the rule is inert. It is not, however, retired: UniProt still serves
it and its modifiedDate of 2025-12-15 falls six days after the obsoletion, which means
routine maintenance touched the rule without noticing that its only output term had
ceased to exist. That matters because the obvious repair - substituting the replaced_by
term GO:0008610 lipid biosynthetic process - would silently reactivate the rule and
restore the errors that a GO curator raised against it in go-annotation#5835.
Auditing all 51 OR-ed condition sets against what their InterPro entries and CATH
FunFams actually identify, at most 26 describe an enzyme that builds a membrane lipid,
and even those would be better served by pathway-specific children (GO:0030148,
GO:0046513, GO:0006688, GO:0009245). Five sets are directionally wrong: alkaline
ceramidase, acid and neutral sphingomyelinase, neutral ceramidase and lysosomal
glucosylceramidase all hydrolyse membrane sphingolipids. Six describe modification or
interconversion rather than synthesis (PagP, lipid A 1-phosphatase, sphingoid-base
phosphatase, sphingosine kinase 2, PNPLA1, VLCFA reductase). Five sets are
non-diagnostic families whose acceptor could equally be a protein or a wall polymer,
and five concern GPI-anchor or dolichyl-phosphate-mannose assembly, which is the
specific complaint raised in go-annotation#5835 against S. pombe Dpm1 (O14466). Four
sets are mechanistically unrelated and reveal how the rule was built: CS37
(5-aminolevulinate synthase) shares the Type I PLP fold with the serine
palmitoyltransferases of CS18 and CS27; CS26 is phosphoglycerate kinase; CS12 is a
PH-domain Ser/Thr protein kinase; CS43 is a lipid-BINDING lipoprotein, not a synthase.
The apparent specificity of 51 named FunFams collapses onto only 26 CATH superfamilies,
ten of which are reused across multiple condition sets, and in three cases
same-superfamily neighbours disagree about whether a membrane lipid is being made at
all. Two further conditions rest on the wrong domain entirely: CS16 and CS30 key on the
cytochrome b5-like electron-donor module rather than a catalytic domain, and CS44
("sphingomyelin synthase-related protein 1") is assigned to an Ets-1 winged-helix
superfamily. Redundancy is substantial: KDSR is reached three ways (CS4, CS9, CS36) and
DPM1 twice by two FunFams of one superfamily carrying identical labels (CS23, CS50).
The 39 taxon conditions name 34 distinct taxa and 12 sets carry no taxon at all; scopes
such as Homo for sphingosine kinase 2, Hominidae for SPT2 and Glires for B4GALT6 are
annotation-bias fossils, while Bacteria and Eukaryota are wide open. The falcon deep
research reached the same overall verdict independently, from the literature rather
than from fold membership, and supplies the directional anchors for the hydrolases and
for LpxC/LpxH; it was, however, told the pre-obsoletion label by this repository's own
stale label cache and therefore reasoned throughout about a live term.
action: DEPRECATE
action_rationale: >-
The rule's only consequent no longer exists in GO, so there is nothing to accept and no
in-place edit that leaves the rule correct. The natural mechanical fix - remap
GO:0046467 to its replaced_by term GO:0008610 - must be rejected: nine of the 51
condition sets (five catabolic, four mechanistically unrelated) would then produce
annotations that are false rather than merely imprecise, and the remaining sound sets do
not need a term as general as "lipid biosynthetic process" when pathway-specific
children exist. Deprecation is also the honest response to go-annotation#5835. The
obsoletion removed this rule's output without examining the reasoning that produced it,
and the reporter's own follow-up ("oh no, some definitely are there now. Will reopen")
is borne out: the same S. pombe Dpm1 protein still carries GO:0046474 from ARBA00028351
and GO:1901137 from ARBA00026302, making the same category error. The defensible core -
sphingolipid, glycosphingolipid, lipid A, GPI-anchor and dolichyl-phosphate-mannose
biosynthesis - should be re-derived as separate pathway-scoped rules with specific
terms, not preserved by re-pointing this one.
suggested_modifications:
- "Retire ARBA00028538 rather than remapping GO:0046467 to GO:0008610; the remap would reactivate the rule and convert nine condition sets from vague to false."
- "Re-derive sphingolipid/ceramide biosynthesis (GO:0030148, GO:0046513) as a separate rule from CS1, CS4, CS8, CS9, CS18, CS27, CS36, after collapsing the three redundant KDSR routes into one."
- "Re-derive the glycosphingolipid sets with the specific child each one warrants rather than the shared parent GO:0006688: CS13 (UGT8/CGT) asserts GO:0006682 galactosylceramide biosynthetic process and CS40 (UGCG) asserts GO:0006679 glucosylceramide biosynthetic process; only CS21 (B4GALT6), whose acceptor is unverified, falls back to GO:0006688, and only with evidence that the orthologue's physiological acceptor is a glycolipid."
- "Re-derive plant and bacterial glycolipid biosynthesis (GO:0009247) from CS17 and CS41, the two clearest non-sphingolipid positives: CS17 is MGD3 making the dominant thylakoid galactolipid, for which GO:0019375 galactolipid biosynthetic process is the exact child; CS41 is the processive diacylglycerol beta-glycosyltransferase that makes the principal Mycoplasmatota membrane glycolipids."
- "Re-derive fungal complex-sphingolipid biosynthesis from CS11 (MIPC synthase subunit) using GO:0051999 mannosyl-inositol phosphorylceramide biosynthetic process, the exact term for the enzyme's product, rather than a generic lipid term or the metabolic parent GO:0006673."
- "Consolidate the actinobacterial PIM/LM/LAM sets CS14, CS32 and CS34, which all sit in the same GT-B superfamily and overlap, into one condition. GO has no phosphatidylinositol mannoside biosynthesis term, so either request one or assert GO:0009247 glycolipid biosynthetic process; CS32's FunFam is named only 'Glycosyl transferase' and should be included only with validated orthology."
- "Drop the lipid A modification sets CS10 (PagP palmitoyltransferase) and CS29 (LpxE-type 1-phosphatase). Both remodel assembled lipid A for envelope stress or cationic-peptide resistance and must not be folded into the GO:0009245 lipid A biosynthesis rule; if annotation is wanted, use a lipid A modification term."
- "Drop the sphingoid-base interconversion sets CS28 (fungal LCB3/YSR2 dihydrosphingosine-1-phosphate phosphatase) and CS31 (SPHK2). Neither synthesises a membrane lipid; CS28 at most supports GO:0006665 sphingolipid metabolic process, and CS31's product S1P is an exported signalling metabolite. CS31's 'Homo' scope is an annotation-bias artefact, not a biological claim."
- "Drop CS35 (fungal very-long-chain 3-oxoacyl-CoA reductase) from a membrane-lipid rule; it is precursor supply and belongs under GO:0042761 very long-chain fatty acid biosynthetic process."
- "Retain CS25 (PNPLA1) only under an epidermal omega-O-acylceramide term, never under a generic lipid term propagated from the patatin fold, since most PNPLA members are lipases; its Euarchontoglires scope is an annotation-bias artefact."
- "Leave the unverified-acceptor sets CS20, CS39, CS42 and CS49 out of any re-derived rule until the physiological acceptor is established. CS20 (ST3GAL2) would take GO:0001574 ganglioside biosynthetic process if confirmed; CS39, CS42 and CS49 are glycosyltransferase or sialyltransferase FunFams whose family membership does not establish a lipid acceptor - the same caveat as CS3."
- "Re-derive lipid A biosynthesis (GO:0009245) from CS7, CS24, CS38, with taxon scopes narrower than 'Bacteria' since lipid A enzyme architecture varies across phyla."
- "Re-derive GPI anchor biosynthesis (GO:0006506) from CS22, CS46, CS47."
- "Merge CS23 and CS50 into a single DPM1 condition asserting GO:0180047 dolichol phosphate mannose biosynthetic process - the term UniProt already uses for the protein named in go-annotation#5835."
- "Drop, do not re-point, the catabolic sets (CS2, CS5, CS15, CS19, CS45); if annotation is wanted, use the corresponding catabolic terms."
- "Drop the mechanistically unrelated sets CS12 (PH-domain protein kinase), CS26 (phosphoglycerate kinase), CS37 (5-aminolevulinate synthase) and CS43 (lipid-binding lipoprotein)."
- "Drop the non-diagnostic family sets CS3, CS6, CS33, CS48 and CS51, whose signatures do not establish a lipid acceptor."
- "Re-examine CS16, CS30 and CS44, whose CATH superfamilies (cytochrome b5-like; Ets-1 winged helix) indicate the condition keys on a non-catalytic or spurious domain assignment."
- "Add an obsoletion check to rule enrichment and sync, and refresh rules/_labels.json for obsoleted GO terms: this rule's stale cached label made an obsolete consequent look live to both curators and downstream tooling."
parsimony:
assessment: OVERLY_COMPLEX
notes: >-
51 OR-ed condition sets for one consequent, against a working guideline of 12. The
repository's own analysis tooling refuses to process it: 'just analyze-rule
ARBA00028538' aborts with 'Rule ARBA00028538 has 51 condition sets, which exceeds the
maximum of 12', so no pairwise overlap or heatmap data could be generated and the
condition-level audit had to be done by hand. Complexity here is not the price of
biological breadth; it is what allows five catabolic and four unrelated condition sets
to sit unnoticed alongside genuine sphingolipid and lipid A enzymes, because in an OR
the weakest condition sets the rule's real specificity.
supported_by:
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "Because any single condition triggers GO:0046467, the weakest condition determines the rule’s practical specificity."
literature_support:
assessment: WEAK
notes: >-
Support is bimodal and the rule as a whole is governed by its weakest sets. The
literature is strong and specific for a core: serine palmitoyltransferase, KDSR and
the ceramide synthases as the de novo sphingolipid pathway; LpxC and LpxH as committed
steps of lipid A biosynthesis; the GPI enzymes and Dol-P-Man donor chemistry. For five
condition sets the literature is not merely thin but directly contradictory - acid
sphingomyelinase, alkaline and neutral ceramidase and lysosomal glucosylceramidase are
characterised as catabolic enzymes. For CS12, CS26, CS37 and CS43 no literature
connects the family to membrane lipid synthesis at all, and their presence is
explained by shared folds rather than shared function. The overall grade is therefore
WEAK rather than STRONG or CONTRADICTED: a substantial well-evidenced core exists, but
it cannot be inferred from the rule, only from auditing it set by set.
supported_by:
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "Acid sphingomyelinase is explicitly characterized as a lysosomal/secretory phospholipase C central to phospholipid **catabolism**, with more than 20 phospholipids cleaved in vitro; this directly contradicts conditions 5 and 19 as predictors of biosynthesis"
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "Conditions 12, 26, and 37—protein kinases, phosphoglycerate kinase, and 5-aminolevulinate synthase—have no direct basis for GO:0046467. Condition 43 is a lipid-binding protein rather than a biosynthetic catalyst."
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "LpxC deacetylates UDP-3-O-acyl-GlcNAc and drives the first committed step of lipid-A biosynthesis; LpxH hydrolyzes UDP-2,3-diacylglucosamine to lipid X as an obligatory biosynthetic reaction. Thus, hydrolytic chemistry does not imply catabolism when the reaction lies inside an anabolic pathway"
condition_overlap:
assessment: SIGNIFICANT
notes: >-
The 51 named FunFams occupy only 26 distinct CATH superfamilies, and ten superfamilies
are reused across two or more condition sets: 3.40.50.2000 across six (CS13, 14, 17,
32, 33, 34), 3.90.550.10 across six (CS21, 23, 39, 40, 41, 50), and 3.40.50.720,
3.40.640.10 and 3.60.21.10 across three each. Outright redundancy: KDSR is matched
three times, by InterPro (CS4) and by two FunFams of one Rossmann superfamily (CS9
Eukaryota, CS36 Saccharomycotina - a taxon subset of CS9); DPM1 is matched twice by
CS23 and CS50, two FunFams of 3.90.550.10 carrying the identical CATH label; GPI
ethanolamine phosphate transferases appear as CS22 and CS47 in 3.40.720.10. Worse than
the duplication is the disagreement: 3.60.21.10 supplies the catabolic sphingomyelin
phosphodiesterase (CS19), the biosynthetic LpxH (CS38) and an unidentified
metallophosphoesterase (CS48), and 3.40.640.10 supplies both serine
palmitoyltransferases and 5-aminolevulinate synthase. Quantitative pairwise overlap
could not be computed because the analysis tool declines rules with more than 12
condition sets; the superfamily census in scripts/census_arba00028538.py --cath is
the substitute.
supported_by:
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-analysis.md
supporting_text: "Ten superfamilies are reused across two or more condition sets, and the\nsame-superfamily neighbours frequently disagree about the biology"
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "Conditions 23 and 50 appear redundant and should be merged after checking whether they represent distinct historical FunFam models."
go_specificity:
assessment: MISMATCHED
notes: >-
The consequent is obsolete. GO:0046467 was obsoleted on 2025-12-09 and renamed
"obsolete membrane lipid biosynthetic process", with replaced_by GO:0008610 lipid
biosynthetic process; the stated reason is that the term "groups lipids by their
localization rather than by their metabolic function, which is problematic since most
lipids can function in multiple locations". QuickGO returns 0 annotations to it from
any source. Beyond the obsoletion the term choice was already wrong in two distinct
ways: directionally, for the five hydrolase condition sets, where the correct terms lie
in the catabolic subtree; and by pathway, for the six GPI/dolichol sets, where GPI
anchor biosynthesis (GO:0006506) and dolichol phosphate mannose biosynthesis
(GO:0180047) are the right terms - the latter being exactly the point made in
go-annotation#5835 about Dpm1, and the term UniProt already assigns to the flagged
protein O14466. For the ~26 defensible sets the term was merely far too broad;
GO:0030148, GO:0046513, GO:0006688 and GO:0009245 are available. Because the errors
are of three different kinds, no single replacement term fixes the rule.
supported_by:
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-analysis.md
supporting_text: "*** the rule asserts an OBSOLETE GO term ***"
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-analysis.md
supporting_text: "GO:0046467 exact, any source : 0"
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "For hydrolases, the term is directionally wrong rather than merely broad."
taxonomic_scope:
assessment: TOO_NARROW
notes: >-
The 39 taxon conditions name 34 distinct taxa, and 12 of the 51 sets carry no taxon
constraint at all - a distribution that reflects where training annotations happened
to sit rather than any phylogenetic judgment. Most scopes are far narrower than the
biology: Homo for sphingosine kinase 2 (CS31) and Hominidae for serine
palmitoyltransferase 2 (CS27), both pan-eukaryotic enzymes; Glires for B4GALT6 (CS21);
Euarchontoglires for ST3GAL (CS3) and PNPLA1 (CS25); Primates for the protein kinase
(CS12). A handful run the other way and are too broad: Bacteria for the lipid A
acyltransferases (CS7) and PagP (CS10), despite well-documented plasticity of lipid A
architecture across bacterial phyla, and Eukaryota for the generic glycosyltransferase
set (CS6). In CS37 the taxon constraint is actively misleading rather than merely
arbitrary: a FunFam named "5-aminolevulinate synthase, mitochondrial" is restricted to
Bacteroidota, and the restriction hides the fold-level false positive instead of
catching it. TOO_NARROW is recorded because arbitrary narrow clades are the dominant
pattern, but the underlying problem is that taxon conditions here are doing filtering
work, not encoding conservation.
supported_by:
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
supporting_text: "Taxonomic restrictions such as Primates, Streptophyta, or Bacteroidota cannot rescue a mechanistically unrelated signature."
- reference_id: file:rules/arba/ARBA00028538/ARBA00028538-analysis.md
supporting_text: "The 39 taxon conditions name **34 distinct taxa**, and 12 of the 51\nsets carry no taxon constraint at all."
confidence: 0.95
references:
- id: file:rules/arba/ARBA00028538/ARBA00028538-analysis.md
title: "ARBA00028538 analysis - GO:0046467, a consequent that GO obsoleted"
findings:
- statement: "GO:0046467 was obsoleted on 2025-12-09 with replaced_by GO:0008610; QuickGO returns 0 annotations to it from any source, so the rule is inert."
- statement: "The rule's modifiedDate of 2025-12-15 is six days after the obsoletion, so it is inert but not retired."
- statement: "The 51 FunFams occupy only 26 CATH superfamilies; 3.40.640.10 supplies both serine palmitoyltransferases and 5-aminolevulinate synthase, and 3.60.21.10 supplies both the catabolic SMPD and the biosynthetic LpxH."
- statement: "At most 26 of 51 condition sets describe an enzyme that builds a membrane lipid; 9 would become false rather than merely vague if the consequent were remapped to GO:0008610."
- id: file:rules/arba/ARBA00028538/ARBA00028538-deep-research-falcon.md
title: "Deep research analysis via Falcon (Edison Scientific), 21 citations"
findings:
- statement: "Independent per-condition audit of all 51 sets concludes the rule is not valid in its present OR-ed form and should be split by pathway and reaction direction."
- statement: "Flags CS12, CS26, CS37 and CS43 as mechanistically unrelated, and CS2, CS5, CS15, CS19, CS45 as catabolic - matching the fold-based analysis derived independently from CATH superfamily membership."
- statement: "Did not know GO:0046467 is obsolete: the prompt supplied the pre-obsoletion label from this repository's stale rules/_labels.json cache, so its recommendation to rely on GO ancestor propagation to GO:0046467 is no longer available."
- statement: "Its executive summary enumerates 14 removal candidates while claiming 12, and cites Grabner et al. 2022 (adipocyte lipolysis inhibitors) for PNPLA1's role in epidermal acylceramide synthesis; the per-set verdicts are sound but the roll-up counts and that one citation should not be quoted."
- id: file:rules/arba/ARBA00028538/scripts/census_arba00028538.py
title: "Reproducible census of the rule, its consequent term, and the protein flagged in go-annotation#5835"
findings:
- statement: "Fetches the rule from rest.uniprot.org/arba, the term from QuickGO, and the CATH superfamily names live; all numbers in the analysis are reproducible from this script."
- statement: "S. pombe Dpm1 (O14466) no longer carries ARBA00028538, but still carries GO:0046474 from ARBA00028351 and GO:1901137 from ARBA00026302, which make the same category error."
- id: PMID:34445706
title: "Acid Sphingomyelinase, a Lysosomal and Secretory Phospholipase C, Is Key for Cellular Phospholipid Catabolism."
findings:
- statement: >-
Frames acid sphingomyelinase as a catabolic enzyme: "sphingolipids of mammalian cells
are degraded in a stepwise manner without an effective bypass in human lysosomal
catabolism", and "The recessively inherited ASMD mainly leads to impaired cellular SM
catabolism causing progressive SM accumulation in the cellular membranes." This is the
direct basis for calling CS5 (SMPD1 architecture) and CS19 (sphingomyelin
phosphodiesterase FunFam) directionally wrong.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.3390/ijms22169001 via NCBI esearch and cached with full text
on 2026-08-29. The quoted sentences were checked verbatim against the cached text. The
title alone carries the argument: this enzyme is classified as a phospholipase C for
catabolism, not a biosynthetic enzyme.
- id: PMID:38458396
title: "Common and varied molecular responses of Escherichia coli to five different inhibitors of the lipopolysaccharide biosynthetic enzyme LpxC."
findings:
- statement: >-
"The essential enzyme LpxC catalyzes the first committed step of LPS synthesis and thus
represents a promising antibiotic target", and "Instead of the first enzyme LpxA, which
catalyzes a reversible reaction, the second enzyme, LpxC, is the main driver in this
process as it catalyzes the first committed step". This supports CS24 (LpxC) as a
genuine biosynthetic condition and, with it, the general point that hydrolytic chemistry
inside an anabolic pathway (CS38, LpxH) is still biosynthesis.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1016/j.jbc.2024.107143 and cached with full text on
2026-08-29; both quotes checked verbatim against the cached text.
- id: PMID:26563290
title: "Biosynthesis of GPI-anchored proteins: special emphasis on GPI lipid remodeling."
findings:
- statement: >-
Describes the GPI backbone as "ethanolamine phosphate (EtNP), three mannoses (Mans), one
non-N-acetylated glucosamine, and inositol phospholipid" whose "Fatty chains of inositol
phospholipids are inserted into the outer leaflet of the plasma membrane". GPI is
therefore a membrane glycolipid, which is why CS22/CS46/CS47 are not absurd - but the
informative term is GO:0006506, not a generic lipid-biosynthesis term.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1194/jlr.R063313; the cached record is abstract-only
(full_text_available: false), so only the abstract was quoted. It establishes GPI
structure and membrane insertion, not the Dol-P-Man donor claim, for which UniProt's own
GO:0180047 annotation on O14466 is the checkable anchor.
- id: PMID:26139634
title: "Functional classification of CATH superfamilies: a domain-based approach for protein function annotation."
findings:
- statement: >-
The FunFam method paper states that "the functional annotations provided by FunFams are
found to be more precise compared with those generated by other domain-based resources".
FunFam membership is nonetheless a prediction of functional coherence within a
superfamily, not a measurement of a specific reaction - which is what CS8 (a ceramide
synthase FunFam inside a Homeodomain-like superfamily), CS16/CS30 (cytochrome b5-like
superfamily) and CS44 (Ets-1 winged-helix superfamily) illustrate.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1093/bioinformatics/btv398 and cached with full text on
2026-08-29; quote checked verbatim. Cited for the method's own scope, not against it.
- id: PMID:34357010
title: "Ceramide Metabolism Enzymes-Therapeutic Targets against Cancer."
findings:
- statement: >-
States plainly that "ceramidases catalyze the degradation of ceramide to sphingosine",
naming "alkaline ceramidases (encoded by ACER1, ACER2 and ACER3), acidic ceramidase
(encoded by ASAH1) and neutral ceramidase (encoded by ASAH2)". This covers CS2 (ACER
InterPro entry) and CS15 (ASAH2 FunFam), both of which the rule assigns to a
biosynthetic term.
reference_review:
relevance: HIGH
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.3390/medicina57070729 and cached with full text on 2026-08-29;
quotes checked verbatim. This is the review the Falcon report leans on for the de novo
sphingolipid pathway as well as for the ceramidases.
- id: PMID:28942130
title: "Biosynthesis and structure-activity relationships of the lipid A family of glycolipids."
findings:
- statement: >-
"Lipid A, the conserved, lipid component of LPS, anchors LPS to the outer membrane", and
while "E. coli acyl transferases are chain length specific, orthologs from other
Gram-negative bacteria are more tolerant, yielding a heterogeneous Lipid A composition of
the outer membrane." The first clause supports CS7/CS24/CS38 as membrane-lipid
biosynthesis; the second is why the "Bacteria" taxon scope on CS7 and CS10 is too broad.
reference_review:
relevance: MEDIUM
correctness: VERIFIED
review_notes: >-
PMID resolved from DOI 10.1016/j.cbpa.2017.07.008 and cached with full text on
2026-08-29; both quotes checked verbatim against the cached text.