ProtNLM's GO:0004252 serine-type endopeptidase prediction is contradicted for X2JEK1. The selected 176-residue product lacks the C-terminal part of the S1 domain carrying the catalytic serine. This is distinct from the longer CG34171 protein's serine-equivalent F222 substitution. Both are incompatible with the canonical S1 catalytic mechanism; they are not the same residue-level defect.
FlyBase FBgn0085200 maps CG34171-PC / FBpp0310560 / NP_001286041 / AHN54555 to X2JEK1 (176 residues), and CG34171-PB / FBpp0293603 / NP_001097175 / ABV53692 to A8DZ15 (292 residues). The record explicitly lists two unique polypeptides and notes “Alternative translation stop created by use of multiphasic reading frames within coding region.” The current short isoform is therefore a documented native gene model, not a sequence truncation introduced by this review. This does not prove the predicted polypeptide is stably expressed, nor identify the sequence actually supplied to the historical ProtNLM run.
The original cohort record and prediction response are preserved as CG34171-uniprot-source.json and CG34171-predictions-source.json, extracted without alteration from the frozen fly-benchmark JSONL files. Raw FlyBase HTML and a BeautifulSoup text extraction are retained. The exact target UniProt record and all reference records were retrieved on 2026-09-08.
The reproducible sequence analysis uses MAFFT v7.526 with localpair and globalpair iterative modes, plus Biopython 1.85 for residue mapping. Reference active-site positions are parsed from UniProt, not recalled from memory. Independent active bovine trypsin P00760 and the first S1 domain of toad OVCH2 Q66TN7 agree on the H/D/S alignment. Q66TN7's second S1 domain serves as a contrasting sequence control: its histidine-equivalent position is asparagine.
Both CG34171 isoforms preserve H72/D124. The long isoform has F222 in the aligned serine position, while X2JEK1 has no aligned sequence there. Full-sequence comparison establishes an identical 167-residue prefix; the short form then has a distinct 9-residue tail. Including this entire tail in the alignment does not supply the missing catalytic region. This is stronger evidence than the SPH244 name, a generic PROSITE caution, or a family-level peptidase annotation.
The intrinsic-activity prediction is NPI, confidence score 0. The optional error_type is omitted: the closest schema enum, PSEUDOENZYME_OVERANNOTATION, describes fold-retaining pseudoenzymes, while retention of a complete fold has not been demonstrated for this short product. The locus is a useful pseudoenzyme-related case, but its selected isoform specifically illustrates failure to check domain completeness. WRONG_INPUT_SEQUENCE is not justified because the predictor's actual input is not known.
The target GOA already contains a NOT GO:0004252 IKR annotation to Cao and Jiang, 2018, PMID:30367934, DOI 10.1016/j.ibmb.2018.10.006. The cached full text explains SP/SPH classification “based on the presence or absence of the His-Asp-Ser catalytic triad.” This is a sequence-based inference, not a negative enzyme assay. The main review ACCEPTs the negation, preserving negated: true and its original evidence code/reference.
Table S1 is described as containing the exact SPH sequences and properties, but its XLSX download returned a browser challenge and the manuscript's supplementary files were unavailable through the attempted EuropePMC route. Its SPH244 row and historical sequence were not inspected. The gene-specific IKR assertion is visible in the current GOA and FlyBase records; the particular contemporary molecular defect is independently established by the local alignment. The paper's general discussion of SPH cofactors does not establish that CG34171-PC has such a role.
The other seeded annotation, IEA GO:0006508 proteolysis from InterPro IPR001254, is UNDECIDED. Loss of intrinsic activity invalidates a direct enzyme inference but does not rule out noncatalytic participation in proteolysis. The short product's target-specific participation is not established. QuickGO definitions for both terms were retrieved and frozen as JSON; the molecular-function term expressly requires a serine nucleophile, whereas the biological-process term describes peptide-bond hydrolysis.
FlyBase also displays a gene-level IBA innate-immune-response assertion. This is a phylogenetic assertion, not mere donor-count evidence, but it was not seeded on the exact target accession. Its ancestral placement and relevance to the shortened PC protein were not independently examined. No new immune-process or protease-inhibitor annotation is proposed, and core_functions is left empty because the short product's positive molecular function is unresolved.
The original GO prediction records model score 0.99, phmmer donor Q66TN7 and phmmer score 65.4. Q66TN7 is a 980-residue toad ovochymase/oviductin with two S1 domains, one bearing the complete catalytic-site annotation. No alignment coordinates accompany the model metadata, so the particular matched domain cannot be inferred. The reviewed donor's function also includes curator-assessed homology transfer; its database status is not direct experimental proof of the target's activity.
The response's “Peptidase S1 domain-containing protein” name is a defensible domain description rather than an enzymatic claim. Its “Secreted” location is consistent with the target's N-terminal hydrophobic signal and SignalP call at residues 1–16; this remains a sequence-based localization inference, not target-specific extracellular detection. These ancillary observations are separate from the GO activity assessment.
The Falcon/Edison report completed in 348.65 seconds on 2026-09-08, with a retained markdown artifact. It usefully warns against confusing SPH244 with the independently studied cSPH242 and does not invent a target-specific enzyme assay. However, it relies heavily on the supplied ARBA-derived CLIP-subfamily description and does not resolve the exact isoforms or the existing Cao/Jiang IKR evidence. That CLIP claim is not adopted: the target has no annotated clip domain, and its sequence before the S1-domain start lacks the cysteine pattern expected of an N-terminal clip domain. S1-family membership is supported; a conventional CLIP architecture is not established.
The report leaves catalysis uncertain because it did not inspect the active-site residues. The independent sequence analysis supplies the missing discriminator and supports NPI for the intrinsic activity. Its general immune-cofactor discussion cannot supply the missing positive molecular function of X2JEK1. No report prose is used as biological validation; the preserved report records the provider's synthesis and its limitations.
All two seeded GOA rows are reviewed, including the explicit NOT flag. Main validation passes with two advisories: no defined positive core function and no annotation-support citation to the provider report. Both are deliberate: the positive function is unresolved, and the provider contributes no independent decisive evidence. Schema status is DRAFT under the repository's warning convention. The ProtNLM sidecar validates with zero errors and warnings, one title and two supporting quotes verified. No GO term or source label was rewritten. No new biological-process or molecular-function claim is introduced merely to fill a field.