CatRange enables robust prediction of enzyme variant kinetic regimes

Abstract Predicting enzyme kinetics directly from sequence remains a central challenge in computational biology, particularly in resolving the effects of mutations at catalytically essential residues. Existing models frequently overlook the functional consequences of such perturbations, defaulting to wild-type predictions even in cases of substantial activity loss, thereby limiting their reliability for enzyme design and mechanistic inference. Here, we introduce CatRange, a machine learning framework trained on CatLog-27k, a human-in-the-loop, AI-agent trustworthy dataset of 27,176 in vitro enzyme–substrate kinetic records created by a systematic audit and correction of BRENDA and SABIO-RK. All mutant entries are manually reconciled against 2,158 source articles. CatRange reframes kinetic prediction from exact numerical regression into classification over log10-spaced bins for catalytic turnover (kcat) and substrate affinity (KM), matching the order-of-magnitude scale at which experimental enzyme kinetic measurements are commonly interpreted. This biologically grounded formulation mitigates assay-level variability while preserving distinctions among functional catalytic and binding states. Using joint enzyme–substrate representations and gradient-boosted classifiers, CatRange predicts kinetic ranges for wild-type and mutant enzymes across standard held-out, out-of-distribution, and few-shot mutation settings. The model shows robust order-of-magnitude recovery with class-balanced discrimination and captures mutation-induced movement across kinetic regimes, including losses associated with perturbation of annotated catalytic residues. CatRange detects non-enzyme sequence inputs and emphasizes rigorous data curation, transparent training data dissemination (CatLog), biochemically informed task formulation, and balanced evaluation metrics. These position CatRange as an interpretable, mutation-sensitive framework with utility in enzyme engineering and kinetic metabolic modeling.