A computational framework integrating a protein language model with alchemical simulation for gain-of-function enzyme design

Abstract

The engineering of enzymes with enhanced activity and stability is a central goal of biotechnology, yet the inherent trade-off between global protein fitness and specific substrate binding affinity poses a significant challenge. Here, we present ESM-FEP, a computational framework that synergistically integrates a fine-tuned protein language model with alchemical free energy perturbation (FEP) to overcome this limitation. Our workflow uses a parameter-efficient fine-tuned ESM-2 model to perform high-throughput saturation mutagenesis and rapidly identify mutations that preserve protein fitness. Top-ranked candidates are then subjected to rigorous FEP simulations to precisely quantify changes in substrate binding affinity. When used to engineer the Zea mays dioxygenase ZmHSL1B to enhance detoxification of the herbicide mesotrione, ESM-FEP efficiently navigated the mutational landscape and identified a quadruple mutant, M5 (Q140H/Y205F/L332R/K336F), with a catalytic efficiency approximately 7-fold higher than that of the wild-type enzyme, as confirmed by in vitro assays and a detailed kinetic analysis. Furthermore, transgenic Arabidopsis thaliana plants expressing the M5 mutant exhibited significantly enhanced herbicide tolerance, validating the functional efficacy of M5 in a biological context. The ESM-FEP framework provides an efficient and generalizable strategy for the rational design of gain-of-function enzymes, with broad applications in biocatalysis, bioremediation, and precision agriculture.