CG18507 (M9PBB3): evidence and ProtNLM claim review

CG18507 is a 510-residue DUF4481-domain protein with two predicted transmembrane helices. It is likely a membrane-associated protein, but its molecular activity, organelle distribution, and biological pathway remain unresolved.

Exact input: M9PBB3, 510 residues. The accession was fetched explicitly with the gene-directory alias; no canonical-sequence substitution is made.

Raw emitted predictions: CG18507-predictions-source.json. Source features: CG18507-uniprot.txt, with an exact extraction in CG18507-sequence-evidence.json.

Sequence and domain evidence

These are sequence/domain observations or explicitly named feature predictions, not measurements of biological function. ARBA assertions and ProtNLM-derived UniProt names are not counted as validation.

DR   InterPro; IPR028054; DUF4481.
FT   TRANSMEM        292..310
FT                   /note="Helical"
FT                   /evidence="ECO:0000256|SAM:Phobius"
FT   TRANSMEM        316..339
FT                   /note="Helical"
FT                   /evidence="ECO:0000256|SAM:Phobius"

ProtNLM claims

The snapshot emits names and location/keyword statements, with no GO or EC prediction for this target. Each actual statement is assessed below; no GO term has been substituted for it. Categories follow the function-prediction rubric, with nonspecific “uncharacterized” names marked UNC because they contain no testable function. CNN records an independently supported existing annotation; it does not assert a particular training-set composition.

Literature evidence

No target-specific primary finding is used to establish a molecular activity here. The current assessment is bounded by exact-record architecture and the explicitly identified curated inferences.

Annotation decisions

The accession-specific GOA fetch contains no existing annotation rows. The molecular activity remains unresolved; no broad GO root or invented function is added to create apparent coverage.

Research provenance

Genuine external literature research is requested through the repository Falcon wrapper, with perplexity-lite configured as fallback. Provider output is retained separately as CG18507-deep-research-<provider>.md; its source leads are checked against the underlying publications and exact sequence record.

2026-10-21 — Full re-review, independent of the ProtNLM2 predictions

This section was written without consulting any ProtNLM/ProtNLM2 output or prediction-evidence files.

Identity of the benchmark accession

• M9PBB3 (510 aa, TrEMBL, PE 4) is one of five UniProt products of FBgn0028527 (CG18507): M9PBB3 510, M9PCX0 510,
  M9PFU6 499, Q9VJW7 513, Q7KTB0 514. None is a reviewed Swiss-Prot entry, so there is no canonical reviewed product.
• All five share an identical N-terminus through residue 438-492 (including both TM helices and the TMEM268-aligned
  core) and differ only in their C-terminal tails, i.e. alternative 3' exons
  [file:DROME/CG18507/CG18507-bioinformatics/RESULTS.md "M9PBB3 is a full-length alternative product of the gene, not a fragment."].
• M9PBB3 is therefore a full-length alternative isoform, not a fragment or defective gene model; no domain or
  predicted TM segment is missing relative to the other products.
• QuickGO (queried 2026-10-21) returns zero GO annotations for any of the five accessions.

Evolutionary placement

• PANTHER PTHR31193 ("TRANSMEMBRANE PROTEIN C9ORF91") has a single subfamily, SF1 "TRANSMEMBRANE PROTEIN 268",
  containing human TMEM268 (Q5VZI3), mouse Tmem268 (Q8R239), bovine TMEM268 and fly CG18507. The fly gene is thus the
  subfamily-level ortholog of TMEM268. The family review (interpro/panther/PTHR31193/PTHR31193-review.yaml) places
  M9PBB3 in SF1, which agrees with the UniProt cross-reference; no disagreement.
• Local alignment (BLOSUM62): fly 227-399 vs human 39-215, 30.5% identity / 52.4% similarity; mouse alignment extends
  to fly 486 / mouse 331. Fly residues 1-226 (MobiDB-lite disordered, low complexity) are a fly-specific extension.
• No PAINT/IBA annotations exist for this family's members in GOA, so there is no curator node placement to follow.

Literature

• Only gene-specific primary source: Chung, Chavez & Andrew 2011 PMID:21963537. The falcon report places CG18507 in Table 1
  (cDNA RE25285) as Trh-dependent; Table 1 is not in the cached text, so the gene-specific row is unverified here.
  This is an expression/regulation fact about the transcript, not a function of the protein; no GO term is warranted
  (an expression pattern under a transcription factor is not participation in tracheal development).
• Human TMEM268: ER and plasma membrane localization PMID:30361615; binds and stabilizes ITGB4
  (PMID:30361615) and CD11b/ITGAM PMID:38730209. Both partners are vertebrate integrins without fly orthologs, so
  the partner-specific stabilization function is not transferred.

Decisions

• No existing GOA rows to review.
• NEW: GO:0016020 membrane (Phobius TM helices in the conserved core + SF1 membership + human ortholog membrane
  localization). Kept generic: no fly localization data, and the human data name two different membranes.
• No molecular-function or biological-process term proposed; core_functions left empty because no activity is
  established. Gaps recorded as BIOLOGY knowledge_gaps with suggested experiments.
• Description rewritten to be family- and evidence-grounded (previous text was a one-line placeholder).

2026-10-21 — Evidence code on the NEW membrane row

Following the review of PR #5305, the NEW GO:0016020 membrane row now uses evidence_type: ISS
instead of IEA. The inference rests on PANTHER subfamily membership with human TMEM268 (whose
membrane location is experimental) plus Phobius transmembrane predictions, so ISS describes it
better than an electronic code. The term and reasoning are unchanged.