CG4793 is a noncatalytic serine protease homolog with experimentally supported roles in cellular immunity. ProtNLM's serine-type endopeptidase activity prediction is contradicted by the absence of the catalytic serine in the selected sequence. This is a pseudoenzyme error, not merely disagreement with a preferred gene name or a generic caution in a database record.
The reproducible sequence analysis maps the catalytic serine of two active trypsin controls to Gly290 in Q8IP30. The full 910-residue sequence is identical to RefSeq NP_723941.2; UniProt maps it to FlyBase FBgn0028514 and CG4793. The N-terminal S1 domain covers residues98–340, followed by a long repetitive tail. Its length and architecture differ from small single-domain SPH proteins. Alignment covers about97% of each mature control, and the separate active-control run recovers an intact triad. The more divergent histidine loop makes the precise Glu145 site mapping less secure than Gly290. Neither historical prediction-time input nor isoform-specific experimental action is established by present-day accession identity.
Kr et al. 2021, PMID:33644706, DOI 10.17912/micropub.biology.000370, provides full-text gene-specific RNAi evidence. Figure1 and its legend state: “Conversely, knockdown of CG4793 in lamellocytes resulted in a significant increase in the encapsulation rate.” Two independent constructs show this effect; plasmatocyte knockdown does not. This supports NEW GO:0140540, negative regulation melanotic encapsulation of foreign target, in the tested Leptopilina clavipes infection model. The paper's mechanistic interpretation remains explicitly provisional: “However, a detailed mechanistic understanding of how these S1A family genes regulate cellular immunity and how their activity may be linked to JAK-STAT pathway signaling remain to be established.” The experiments establish regulation, not direct cleavage or a particular binding partner. The published reagent table appears internally inconsistent in its TRiP genotype/stock labeling; the distinct NIG4793R RNAi line and the reported convergent gene-specific results provide a second experimental handle, but construct identity would merit checking in follow-up replication.
Bina et al. 2010, PMID:20168330, DOI 10.1038/embor.2010.1, is abstract-only in the standard cache. The full article was accessed and read separately: PDF, unmodified extracted text, and download provenance. Figure3F and the page206 results place CG4793 among the genes whose knockdown suppresses melanotic tumour formation in the constitutively active hopTumL JAK model. The text states: “CG4793, CG1572 and CG13559—were identified whose knockdown significantly reduces the development of tumours in the adult fly”. Direct circulating-haemocyte counts in Figure3G/H concern baz, not CG4793. Retain FlyBase's positive-regulation-of-haemocyte-proliferation IGI annotation as KEEP_AS_NON_CORE, interpreted in this activated-JAK tumour context. This is distinct from the physiological encapsulation response and does not establish a universal positive regulator of cell division.
The four GOA NOT activity annotations (three serine-type endopeptidase, one peptidase) are accepted. They agree with target-specific catalytic-site loss and the published SPH framework. Cao and Jiang2018, PMID:30367934 explicitly classify SPs versus SPHs by presence/absence of the His–Asp–Ser triad. Its family-level discussion is consistent with the local analysis; it is not a substitute for checking Q8IP30. Ross et al.2003, PMID:12568721 is available as an abstract and records the original genome-wide framework; its detailed target alignment was not accessible here.
The extracellular-region IBA is accepted as an inference consistent with the exact sequence's predicted signal peptide, residues1–18. ARBA family text and keywords are not counted as independent support. Both positive proteolysis annotations remain UNDECIDED. An inactive domain cannot catalyze peptide-bond hydrolysis, but could regulate another enzyme. Neither gene-specific study establishes or excludes that proteolytic participation. The IBA's ancestral node was not inspected in PAINT, so no unsupported propagation-root-cause block is asserted; the corresponding validation advisory is retained.
The frozen source JSON is Q8IP30-predictions.json, copied exactly in content from the project's 2026-09-08 release snapshot. It records ProtNLM2, model score0.99, phmmer accessionQ7Z410, score226.0 for GO:0004252. These values describe prediction provenance. They cannot compensate for loss of the catalytic nucleophile. A separate secreted-location prediction is present in raw metadata but is outside this functional GO claim review.
Falcon deep research returned successfully through the repository workflow, concurrently with publication caching. The report correctly separates family annotation from intrinsic catalytic activity, but did not retrieve the gene-specific primary experiments in enough detail: its statement that the accessible evidence does not identify whether CG4793 controls encapsulation is resolved by the full 2021 article and Figure1. Its catalytic-status uncertainty is resolved at the sequence-inference level by the independently reproduced replacement of the catalytic serine by Gly290; no purified-protein assay is claimed. The report also repeats a generic CLIP architecture from supplied family metadata, whereas the target has a long repetitive C-terminal extension and an N-terminal protease-like domain. The primary articles, exact sequence, and local computations support the specific conclusions above. The current review's unresolved molecular activity and process assignments are intentional; a specific peptidase-regulator molecular function would require a demonstrated target interaction.