# NCBIFAM -> GO mapping (SSSOM) -- curated seed for a missing "ncbifam2go" pipeline # # These are hand-verified NCBIFAM(protein family) -> GO mappings. Unlike RHEA, GO publishes NO # ncbifam2go external2go file (confirmed: https://current.geneontology.org/ontology/external2go/ # ncbifam2go -> HTTP 403; only interpro2go is served). NCBIFAM therefore reaches GOA only where # its signature is integrated into an InterPro entry that itself has an interpro2go row -- but # NCBIFAM already assigns its own GO terms in the PGAP HMM metadata (hmm_PGAP.tsv `go_terms` # column), which GO does not ingest. This seed adopts a sample of those NCBI assignments after # independent verification, so they could seed an ncbifam2go mapping (see ../NCBIFam.md and # NCBIFAM-METHODOLOGY.md, gap classes G4/G5). # # 28 mappings: 27 skos:exactMatch + 1 skos:broadMatch (as in the RHEA project's rhea2go.sssom.yaml). # The object is always the mapping WE propose for ncbifam2go (not a transcription of NCBI's go_terms): # * skos:exactMatch -- the GO term IS the family's function; a ready-to-add ncbifam2go row. # Enzyme rows are EC-bridge supported: ec2go maps the family's EC to exactly # this GO term (verified live). Where NCBI's go_terms was too broad (even the # GO:0003824 near-root) but a precise child EXISTS and is EC-bridged, we # propose the specific child here and record NCBI's broad value in the comment. # Conversely, FtsX (TIGR00439) shows the opposite judgement: we choose the # curator-consensus GO:0051301 (cell division) over the more specific # cytokinesis terms because propagating those would be over-annotation. # * skos:broadMatch -- reserved for the case where NO more-specific GO term exists to adopt: the # best available term is a whole-complex CC term a subunit is part_of (VirB5). # Only 1 such row remains. # # Verification (do NOT trust the NCBI assignment blindly -- the scoping sample contained an OBSOLETE # id, GO:0009448 on a GABA transaminase, AND an outright WRONG one, GO:0003951 "NAD+ kinase activity" # on the diacylglycerol-kinase model NF009874 (EC 2.7.1.107; correct term GO:0004143) -- both # excluded): # * every GO id/label was checked non-obsolete against QuickGO on 2026-06-20; # * every EC->GO bridge was confirmed against the live ec2go file; # * every family accession + product name + family_type + EC + go_terms is from hmm_PGAP.tsv; # * `gain_*` in each comment is the live UniProtKB propagation gain from ncbifam_go_gain.py # (entries carrying the NCBIFAM signature that lack the GO term or any descendant; rev = the # Swiss-Prot/reviewed subset, the curation-relevant number; all = mostly TrEMBL). # # Validate with: just validate-ncbifam-mappings # (linkml-validate against the SSSOM 'mapping set' shape, then GO term/label validation against # the full GO graph -- MF, BP, and CC, since NCBIFAM families span all three aspects). mapping_set_id: https://w3id.org/ai4curation/ai-gene-review/mappings/ncbifam2go mapping_set_title: NCBIFAM to GO mapping (curated seed for a missing ncbifam2go pipeline) mapping_set_description: >- Curated NCBIFAM protein-family -> GO mappings adopting NCBI's own hmm_PGAP.tsv go_terms assignments after independent QuickGO/ec2go verification. GO publishes no ncbifam2go external2go file, so these NCBI-assigned functions reach GOA only via InterPro integration (interpro2go, GO_REF:0000002) and are largely absent from the unreviewed (TrEMBL) entries that carry the family signature. exactMatch rows are ready-to-add ncbifam2go entries (enzyme rows EC-bridge supported); broadMatch rows attach a family to the best available broader GO term and name the narrower term to use/request. Each mapping records the live UniProtKB propagation gain (reviewed and all). Seeds the NCBIFam project's curation of an ncbifam2go mapping and quality-checks GO annotation of the corresponding families. Aspects span MF, BP, and CC (NCBIFAM is a whole-protein family resource, not an enzyme-only one). license: https://creativecommons.org/licenses/by/4.0/ creator_label: - AI Gene Review project mapping_date: "2026-06-20" subject_source: ncbifam object_source: GO curie_map: NCBIFAM: https://www.ncbi.nlm.nih.gov/protfam/ GO: http://purl.obolibrary.org/obo/GO_ skos: http://www.w3.org/2004/02/skos/core# semapv: https://w3id.org/semapv/vocab/ mappings: # ===== exactMatch: ready-to-add ncbifam2go rows (GO term == the family's function) ===== - subject_id: NCBIFAM:NF009803 subject_label: "formamidase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004328 object_label: formamidase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge supported: hmm_PGAP gives EC 3.5.1.49 and ec2go maps EC 3.5.1.49 -> exactly GO:0004328. equivalog family. Propagation gain: gain_rev=0 (27/27 reviewed already have it) but gain_all=54 -- the gap is entirely in unreviewed TrEMBL entries carrying NF009803. - subject_id: NCBIFAM:NF005824 subject_label: "acetolactate synthase large subunit (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0003984 object_label: acetolactate synthase activity mapping_justification: semapv:ManualMappingCuration comment: >- The large (catalytic) subunit carries the activity. equivalog. gain_rev=0 (1/1), gain_all=6. Clean propagation case -- nearly fully covered already, included as an exactMatch exemplar. - subject_id: NCBIFAM:NF045700 subject_label: "AttM family quorum-quenching N-acyl homoserine lactonase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0102007 object_label: acyl-L-homoserine-lactone lactonohydrolase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge supported: EC 3.1.1.81 -> GO:0102007 in ec2go. equivalog, PMID:11930013. Quorum quenching. gain_rev=0 (4/4), gain_all=115 -- large TrEMBL gap. - subject_id: NCBIFAM:TIGR03230 subject_label: "lipoprotein lipase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004465 object_label: lipoprotein lipase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge supported: EC 3.1.1.34 -> GO:0004465 in ec2go. equivalog, PMID:8308035. gain_rev=0 (11/11), gain_all=9. - subject_id: NCBIFAM:NF033545 subject_label: "IS630 family transposase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004803 object_label: transposase activity mapping_justification: semapv:ManualMappingCuration comment: >- equivalog. The single largest propagation gap in the scoping sample: gain_all=18874 of 18881 entries carrying NF033545 lack GO:0004803 (or any descendant), and gain_rev=2. Mobile-element families are precisely where InterPro integration lags, so NCBIFAM's GO is unused. - subject_id: NCBIFAM:NF041162 subject_label: "family 2A encapsulin nanocompartment shell protein (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0140737 object_label: encapsulin nanocompartment mapping_justification: semapv:ManualMappingCuration comment: >- Cellular-component mapping: the shell protein localises to / constitutes the encapsulin nanocompartment. equivalog, PMID:25024436,34362927,35146412. gain_all=897 of 901, gain_rev=0. - subject_id: NCBIFAM:NF042963 subject_label: "anti-phage-associated DUF1156 domain-containing protein (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0051607 object_label: defense response to virus mapping_justification: semapv:ManualMappingCuration comment: >- Biological-process mapping for an anti-phage defense family. equivalog, PMID:32855333. gain_all=151 of 151 (0 carry the term), gain_rev=0 -- a brand-new defense family with no GO propagation at all; a clear gap-fill / proposed-annotation candidate. # ----- exactMatch, EC-bridge supported (ec2go(EC) == this GO term, verified live 2026-06-20) ----- - subject_id: NCBIFAM:NF000320 subject_label: "PEN family class A beta-lactamase, Bcc-type (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0008800 object_label: beta-lactamase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.5.2.6 -> GO:0008800 in ec2go. AMR-relevant. equivalog, PMID:19075063,9371340. gain_all=0 of 45 (already fully covered in this reviewed-heavy family), included as a clean AMR exemplar; cf. NF033105 (a different beta-lactamase family) which maps to the same GO term. - subject_id: NCBIFAM:NF033105 subject_label: "subclass B3 metallo-beta-lactamase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0008800 object_label: beta-lactamase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.5.2.6 -> GO:0008800. AMR-relevant (carbapenem-hydrolysing metallo class). equivalog. gain_all=165 of 1170 -- a substantial TrEMBL gap. Two distinct NCBIFAM families (NF000320 serine class A, NF033105 metallo B3) legitimately share one GO term: family -> GO is many-to-one, the structural analog of RHEA's many-reactions-to-one-activity finding. - subject_id: NCBIFAM:NF002525 subject_label: "D-alanine--D-alanine ligase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0008716 object_label: D-alanine-D-alanine ligase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 6.3.2.4 -> GO:0008716. Peptidoglycan biosynthesis (Ddl, a vancomycin-resistance locus). equivalog. gain_all=3 of 777, gain_rev=0. - subject_id: NCBIFAM:NF003009 subject_label: "5'-deoxynucleotidase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0002953 object_label: 5'-deoxynucleotidase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.1.3.89 -> GO:0002953. equivalog. gain_all=170 of 1714, gain_rev=0. - subject_id: NCBIFAM:NF004018 subject_label: "uridine kinase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004849 object_label: uridine kinase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 2.7.1.48 -> GO:0004849. Pyrimidine salvage. equivalog. gain_all=329 of 15080 -- the largest TrEMBL gap among the enzyme rows in this batch; gain_rev=0. - subject_id: NCBIFAM:NF006707 subject_label: "class I fructose-bisphosphate aldolase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004332 object_label: fructose-bisphosphate aldolase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 4.1.2.13 -> GO:0004332. Glycolysis. equivalog. gain_all=4 of 1476, gain_rev=0. - subject_id: NCBIFAM:NF007054 subject_label: "alpha-amylase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004556 object_label: alpha-amylase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.2.1.1 -> GO:0004556. equivalog. gain_all=1 of 37, gain_rev=0. - subject_id: NCBIFAM:NF011000 subject_label: "acylphosphatase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0003998 object_label: acylphosphatase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.6.1.7 -> GO:0003998. equivalog. gain_all=4 of 1135, gain_rev=0. - subject_id: NCBIFAM:NF040791 subject_label: "glycerate 2-kinase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0043798 object_label: glycerate 2-kinase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 2.7.1.165 -> GO:0043798. equivalog, PMID:16684110. gain_all=6 of 7 -- small but nearly total gap for this family. - subject_id: NCBIFAM:NF045654 subject_label: "acid phosphatase PhoC (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0003993 object_label: acid phosphatase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.1.3.2 -> GO:0003993. equivalog, PMID:10877772,8081499. gain_all=1 of 80. - subject_id: NCBIFAM:TIGR02321 subject_label: "phosphonopyruvate hydrolase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0033978 object_label: phosphonopyruvate hydrolase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 3.11.1.3 -> GO:0033978. Phosphonate catabolism. equivalog, PMID:12697754. gain_all=71 of 137, gain_rev=0. - subject_id: NCBIFAM:TIGR02694 subject_label: "arsenate reductase (azurin) small subunit (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0050611 object_label: arsenate reductase (azurin) activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 1.20.9.1 -> GO:0050611. Arsenic detoxification/respiration. equivalog, PMID:12679550. gain_all=144 of 355, and gain_rev=2 -- reviewed entries are missing it too. - subject_id: NCBIFAM:TIGR03828 subject_label: "1-phosphofructokinase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0008662 object_label: 1-phosphofructokinase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 2.7.1.56 -> GO:0008662. Fructose catabolism. equivalog. gain_all=16 of 5974, gain_rev=0. - subject_id: NCBIFAM:NF001277 subject_label: "adenosylcobinamide-GDP ribazoletransferase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0051073 object_label: adenosylcobinamide-GDP ribazoletransferase activity mapping_justification: semapv:ManualMappingCuration comment: >- EC-bridge: EC 2.7.8.26 -> GO:0051073. Cobalamin (B12) biosynthesis (CobS). equivalog. gain_all=4 of 1293, gain_rev=0. # ===== exactMatch (our proposed mapping): specific term we curate to REPLACE NCBI's broad ===== # NCBI's hmm_PGAP go_terms gives only a broad/near-root term for these families, but a precise GO # term already exists and is EC-bridge confirmed (ec2go(EC) = the specific term). The mapping we # propose for ncbifam2go is therefore the SPECIFIC term as an exactMatch -- not NCBI's broad value # (which is recorded in the comment as the thing being corrected). Reclassifying broadMatch->exactMatch # here also UNMASKS the real propagation gain: the broad parent was near-universal so its gain looked # ~0, whereas the specific term reveals large gaps, including reviewed (Swiss-Prot) ones. - subject_id: NCBIFAM:NF002326 subject_label: "deoxyguanosinetriphosphate triphosphohydrolase / dGTPase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0008832 object_label: dGTPase activity mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term. NCBI's go_terms gave only the broad parent GO:0016793 (triphosphoric monoester hydrolase activity); the specific GO:0008832 (dGTPase activity) is correct (product name = dGTPase, EC 3.1.5.1; ec2go EC 3.1.5.1 -> GO:0008832). equivalog. Against the specific term gain_all=456 of 3719 and gain_rev=13 -- a substantial reviewed gap the broad parent (gain~0) hid entirely. - subject_id: NCBIFAM:NF005804 subject_label: "enoyl-CoA hydratase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004300 object_label: enoyl-CoA hydratase activity mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term. NCBI assigned the ontology near-root GO:0003824 (catalytic activity) to this EC 4.2.1.17 enzyme; ec2go maps EC 4.2.1.17 -> GO:0004300 directly. equivalog. Against the specific term gain_all=184 of 485, gain_rev=1 (the GO:0003824 parent gain was ~0 -- altitude masking). - subject_id: NCBIFAM:NF006559 subject_label: "dihydroorotase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004151 object_label: dihydroorotase activity mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term. NCBI assigned the broad parent GO:0016810; the specific GO:0004151 is the ec2go target of EC 3.5.2.3. Pyrimidine biosynthesis (PyrC). equivalog. Against the specific term gain_all=491 of 1354 (vs 4 for the broad parent), gain_rev=0. - subject_id: NCBIFAM:TIGR00417 subject_label: "spermidine synthase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0004766 object_label: spermidine synthase activity mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term. NCBI assigned the ontology near-root GO:0003824 (catalytic activity) to this EC 2.5.1.16 enzyme; ec2go maps EC 2.5.1.16 -> GO:0004766 directly. Polyamine biosynthesis. equivalog. Against the specific term gain_all=575 of 9260 and gain_rev=1 (the GO:0003824 parent gain was ~0). - subject_id: NCBIFAM:TIGR03542 subject_label: "LL-diaminopimelate aminotransferase (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0010285 object_label: "L,L-diaminopimelate:2-oxoglutarate transaminase activity" mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term. NCBI assigned the broad class GO:0008483 (transaminase activity); the specific GO:0010285 is the ec2go target of EC 2.6.1.83. Lysine/peptidoglycan biosynthesis. equivalog, PMID:17093042,17583737. Against the specific term gain_all=1185 of 2649 (vs 77 for the broad class), gain_rev=2. # ===== broadMatch: only a broader/whole-complex term exists; no specific child to adopt ===== # One row remains broadMatch: VirB5 is a subunit part_of a complex (the term is the whole complex, # no subunit-specific CC term exists). - subject_id: NCBIFAM:TIGR02791 subject_label: "P-type DNA transfer protein VirB5 (equivalog)" predicate_id: skos:broadMatch predicate_label: broad match object_id: GO:0043684 object_label: type IV secretion system complex mapping_justification: semapv:ManualMappingCuration comment: >- VirB5 is a minor pilus/component subunit part_of the T4SS; the GO term is the whole complex, so the relation is part_of (broadMatch), not exact, and there is no VirB5-specific CC term to propose. equivalog, PMID:12855161,14673074,15901731. gain_all=298 of 298 (0 carry it), gain_rev=4. # ===== exactMatch (our proposed mapping): curator-consensus term, declining the over-annotation ===== - subject_id: NCBIFAM:TIGR00439 subject_label: "permease-like cell division protein FtsX (equivalog)" predicate_id: skos:exactMatch predicate_label: exact match object_id: GO:0051301 object_label: cell division mapping_justification: semapv:ManualMappingCuration comment: >- Our proposed term, chosen against NCBI's GO:0000910 (cytokinesis) on empirical + ontology grounds. Ontology: GO:0000910 cytokinesis is part_of GO:0051301 cell division, so NCBI's term is actually NARROWER, not broader. Empirically all 7 reviewed FtsX proteins carry GO:0051301 (cell division) but only 2/7 carry cytokinesis or GO:0043093 (FtsZ-dependent cytokinesis). FtsX/FtsEX regulates septal peptidoglycan hydrolysis and divisome assembly, so the conservative participation term (cell division) is the curator consensus. gain_all=22 of 3306 (already near-universal -- a confirmatory, low-gap mapping). NOTE: mapping to the specific GO:0043093 instead would show a huge apparent gain (3304) but curators apply it to only 2/7 entries, so blanket propagation would be OVER-ANNOTATION -- we deliberately keep the safe consensus term. The opposite call from the five altitude rows above, where the specific term WAS the correct, universally-applicable one.