Qiu, J.-H.; Lin, Z.; Chen, K.-W.; Sun, T.-Y.; Zhang, X.; Yuan, L.; Tian, Y.; Wu, Y.-D.

The quantitative determination of enzyme kinetic parameters traditionally relies on experimental methods that are both time-intensive and costly. Machine learning models have demonstrated significant potential for predicting enzyme kinetic parameters in recent years. Despite this promise, these methods face challenges, including limited training data, inadequate sensitivity to subtle mutations, and poor alignment with practical enzyme engineering contexts. Here, we introduce SAKPE (Site Attention Kinetic Parameters Prediction Method for Enzyme Engineering), a novel machine-learning framework designed to predict enzyme kinetic parameters with enhanced accuracy in practical application scenarios. By incorporating protein representation, substrate representation, and protein representation with weights for important sites, SAKPE significantly outperforms existing methods in predicting enzymatic kinetic parameters, including turnover number (kcat), Michaelis constant (Km), and inhibition constant (Ki). Incorporating protein representation with weights for important sites enables SAKPE to effectively capture the impact of mutations, especially mutations of important sites and their surrounding amino acids of interest in enzyme engineering, on enzyme kinetics parameters. SAKPE offers a robust and practical tool for predicting enzyme kinetic parameters, providing a superior tool for enzyme engineering scenarios such as enzyme design, directed evolution, and industrial applications.