Finding: Unresolved for this exact horse sequence despite a well-supported human SIRT5 mechanism.
Target: F6S899, Equus caballus. The exact source is the frozen ProtNLM API response dated 2026-09-08, retained in projects/PROTNLM_EVALUATION/mammal-benchmark/predictions.jsonl.gz.
NAD-dependent lysine demalonylase, desuccinylase and deglutarylase that specifically removes malonyl, succinyl and glutaryl groups on target proteins. Has weak NAD-dependent protein deacetylase activity; however this activity may not be physiologically relevant in vivo
Uncertain. Human SIRT5 has this activity, but the selected horse sequence diverges sharply after roughly residue 200 and lacks multiple human NAD-contact segments. Two conserved zinc-coordinating cysteines map to tryptophans.
Uncertain. The retained substrate-selectivity residues Tyr102 and Arg105 and catalytic His158 support SIRT5-family identity. They do not compensate for the unresolved zinc-binding and C-terminal-fold defects.
Uncertain. Human biochemical evidence establishes deglutarylation, but functional transfer requires an intact catalytic fold and cofactor site; this exact horse sequence fails a straightforward conservation check.
Uncertain. The statement is appropriate for characterized human SIRT5, but activity measurements cannot be assigned to this divergent horse protein model.
Not established for horse. This is a qualification of human enzymology rather than direct evidence about equine physiology. No horse experiment available here resolves the catalytic activity or its physiological significance.
The comparison points to a possible protein-model problem, not a demonstrated equine pseudoenzyme. The current accession sequence may not be identical to the sequence originally supplied to ProtNLM. Establishing a corrected transcript/protein model and testing its zinc/NAD-binding fold are necessary before deciding whether this is a prediction failure. Horse-specific research is justified by equine SIRT5-expression literature (PMID:36361948), but expression alone would not validate enzyme activity.
The reproducible paired sequence comparison records residue-level findings and sequence hashes. The human UniProt source distinguishes experimental supporting papers from inferred statements. ARBA assertions and generated review prose are not used as validating evidence.