Paired horse–human reviews: cases 21–30
All ten pairs have an initial annotation review and human research. Source-specific uncertainty is retained in the YAML; no PENDING rows remain. ProtNLM narrative and GO statements are assessed separately.
| Gene | Human rows | Horse rows | ProtNLM GO | Narrative | Human research |
|---|---|---|---|---|---|
| CDK7 | 146 | 23 | CNN 1, COR 1 | — | Falcon |
| MAP2K2 | 174 | 15 | COR 1 | — | Falcon |
| ZDHHC23 | 13 | 9 | CNN 1 | — | Falcon |
| CH25H | 25 | 14 | LSP 2, CNN 1 | — | Falcon |
| BCAT2 | 42 | 5 | supported | Falcon | |
| SIRT5 | 68 | 1 | UNC 1 | UNC | Falcon |
| USP8 | 83 | 9 | CNN 1 | supported | Falcon |
| OMA1 | 54 | 16 | CNN 1 | — | Falcon |
| EFR3A | 13 | 1 | UNC 1 | — | Falcon |
| CACNB3 | 46 | 10 | COR 1 | — | Falcon |
Findings needing attention
- SIRT5 F6S899: the chosen sequence loses the normal C-terminal conservation pattern, substitutes Trp at two human zinc-ligand cysteines, and changes or loses several NAD-contact segments. Metal binding and the multi-activity narrative remain UNC for this exact protein model. This is not yet a demonstrated horse pseudoenzyme.
- EFR3A A0A9L0S4L8: the N-terminal cysteine cluster is missing, although the downstream sequence is 98.24% identical. Primary human mutagenesis supports the significance of losing this membrane anchor; membrane recruitment remains UNC.
- BCAT2: catalysis is supported, but the extra 60-residue N-terminal segment means the exact-model mitochondrial targeting remains unresolved.
- CDK7: the seven-subunit TFIIH core excludes the CAK module; protein kinase activity should not become ATP-driven DNA modification. The human experimental DNA-activity assignment remains UNDECIDED until its whole-complex assay is resolved.
- CH25H: C4 methylsterol oxidation/zymosterol synthesis should not be transferred from other sterol enzymes to a cholesterol C25 hydroxylase. Lipid biosynthesis itself remains a defensible description of oxysterol formation.
- CACNB3: the beta subunit is a channel regulator rather than a pore-forming subunit.
- USP8: deubiquitination can participate in cargo degradation; the ubiquitin-dependent catabolic-process prediction is not refuted merely by the deubiquitinase label.
Evidence and validation
The comparison scripts, alignments, residue maps and sequence hashes are in each horse gene’s bioinformatics directory. Primary source caches were fetched with the repository pipeline; rate-limited requests were retried with a delay. Human annotations use primary findings and experimentally supported UniProt passages; research reports serve as literature syntheses and source leads. Target annotation overlap is distinct from unknown training-set membership.
Individual gene validation and prediction-evidence validation were run. Remaining advisories concern evidence-specific UNDECIDED rows, omission of unsupported horse core functions, structured propagation metadata, and research-report citation suggestions. The review is an initial evidence pass, not expert sign-off.