ADAM5 (human, Q6NVV9) — catalytic-motif, ortholog-length and GO-census analysis

Generated by analyze_adam5.py; every number below is re-fetched from UniProt and
QuickGO on each run. Reproduce with uv run python analyze_adam5.py.

A. Ortholog length panel

Every reviewed ADAM5 orthologue is 709–777 aa and carries an annotated Peptidase M12B domain. The human entry is 412 aa with no M12B domain at all — 344 residues shorter than the macaque protein in which tMDC II was characterised as a sperm-surface antigen.

All five orthologues, human included, carry UniProt's CAUTION: Not expected to have protease activity; only the human entry adds Could be the product of a pseudogene.

B. Alignment of human ADAM5 to macaque ADAM5 (Q28483)

Global BLOSUM62 alignment, free end gaps. 339/401 identities over aligned columns (84.5%).

The retained blocks are near-identical to macaque, so this is not a diverged paralogue: it is the same gene with material missing. The missing material is:

• macaque 1–43 (N-terminus, no human counterpart)
• macaque 196–441 (246 aa internal deletion, fused after human residue 162)
• macaque 691–756 (C-terminus, 66 aa)

Coverage of each annotated macaque feature by the human sequence:

The claim the review rests on is stated as an ordering, with no threshold:

• lost: Domain:Peptidase M12B 183-380; Signal:Signal 1-16; Transmembrane:Helical 699-719
• retained: Domain:Disintegrin 389-478; Domain:EGF-like 630-664; Propeptide:Propeptide 17-142

The most-covered lost feature (Domain:Peptidase M12B 183-380, 0.066) is less covered than the least-covered retained feature (Domain:Disintegrin 389-478, 0.411), margin 0.345. The run asserts that ordering, asserts that no reference feature is left unclassified, and fails loudly if either breaks.

So the catalytic domain is not merely degenerate in the human gene — at 0.066 coverage the sequence encoding it is essentially absent from the putative translation product. And it is not alone: the signal peptide and the transmembrane helix are lost outright, and the disintegrin domain is itself truncated because the deletion cuts through it. A product lacking both secretory targeting and a membrane anchor could not be displayed on the sperm surface even if it were translated, which bears on the cellular-component aspect as much as the molecular-function one.

Note a consequence for the human entry's own feature table: its annotated Disintegrin domain (111–199) straddles the deletion junction at residue 162/163. Only its C-terminal portion derives from macaque disintegrin sequence; the N-terminal portion derives from the region upstream of the macaque disintegrin domain. The human disintegrin call is a fusion created by the deletion, not a conserved intact domain.

C. Catalytic zinc-site scan (controls in both directions)

Reprolysin zinc-binding motif HExxHxxGxxHD; minimal core HExxH.

The scan discriminates: all four catalytic controls carry an intact motif, and every sperm ADAM that UniProt describes as non-catalytic lacks even the HExxH core. Human ADAM5 contains no HExxH substring anywhere in its 412 residues.

D. GO census of the catalytic terms

QuickGO, goUsage=descendants, is_a/part_of; counts are annotations, not entities.

E. Detector: annotated catalytic activity on a fold with no zinc site

Fires when a protein has an annotated Peptidase M12B domain and lacks the HExxH core and nevertheless carries GO:0004222.

Fires:

• Q28483 ADAM5 macaque — M12B 183–380, no zinc core, 1 × GO:0004222 (IEA)
• Q3TTE0 Adam5 mouse — M12B 185–378, no zinc core, 2 × GO:0004222 (IBA,IEA)
• Q5BK84 Adam5 rat — M12B 141–334, no zinc core, 2 × GO:0004222 (IBA,IEA)
• Q60472 ADAM5 guinea pig — M12B 185–382, no zinc core, 1 × GO:0004222 (IEA)
• Q99965 ADAM2 human — M12B 178–375, no zinc core, 2 × GO:0004222 (IBA,IEA)
• Q9Y3Q7 ADAM18 human — M12B 184–381, no zinc core, 2 × GO:0004222 (IBA,IEA)
• Q9H2U9 ADAM7 human — M12B 199–394, no zinc core, 2 × GO:0004222 (IBA,IEA)

Clean:

• Q6NVV9 ADAM5 human (subject) — no M12B fold and no catalytic term
• O14672 ADAM10 human — zinc motif intact
• P78536 ADAM17 human — zinc motif intact
• Q13443 ADAM9 human — zinc motif intact
• Q9H013 ADAM19 human — zinc motif intact

The boundary is the finding. The pipelines that place GO:0004222 on this clade key on presence of the M12B fold, not on integrity of the zinc site, so every close relative of ADAM5 that retains the fold receives the activity term despite lacking the catalytic residues. Human ADAM5 is the one member that escapes — and it escapes because pseudogenisation deleted the fold the pipeline matches on, not because any pipeline recognised the protein as non-catalytic. Nothing on human ADAM5 needs retracting; the correctable defect sits on its relatives.

F. Complete GO annotation set for human ADAM5

3 annotations, all ND against GO_REF:0000015 — the three ontology roots. There is no molecular-function, cellular-component or biological-process claim on this gene to evaluate.

G. The rule, measured across the whole reviewed PTHR11905 family

Section E infers a rule from a 12-member hand-picked panel. This section tests it over all 331 Swiss-Prot reviewed members of PANTHER PTHR11905. Note the scope: the family contains 29,886 proteins in total, so this is the reviewed subset (1.1%), and every number below is a statement about reviewed entries, not about the family.

The metric here is exact GO:0004222 presence in the entry's GO cross-references, which is not the same measurement as section D's descendant-aware QuickGO count. So the two are reconciled rather than assumed to agree: the family-wide detector is required to reproduce the hand panel's verdict for all 10 panel members present in the reviewed set, and the run fails if any disagrees.

Panel members absent from this family (reported, not silently skipped): O14672, P78536 — ADAM10 and ADAM17 are classified in a different PANTHER family, so the family-wide run cannot corroborate them.

This converts the rule from inferred to measured. If the annotation discriminated on the catalytic site, the second row would be near zero. It is 37/40 (92%) — statistically indistinguishable from the 100% of intact members. Losing the zinc-binding site has almost no effect on whether a reviewed family member is annotated with metalloendopeptidase activity, which is precisely the claim raised for InterPro and GO Central in suggested_questions.

Checks run

• motif expectations asserted in both directions across the panel
• subject asserted to have no M12B domain, no HExxH, and zero catalytic annotations
• panel asserted all-Swiss-Prot with an anchored startswith test
• every QuickGO query asserted numberOfHits == len(results)
• every UniProt fetch asserted to return an entry name (dead-accession guard)
• family fetch paginated to exhaustion and asserted against x-total-results
• family-wide detector asserted to reproduce the hand panel before its wider number is used

Problems reported: 0