Machine Learning-Driven Enzyme Mining: Opportunities, Challenges, and Future Perspectives
Felix Moorhoff, Yanzi Zhang, Sizhe Qiu, Wenjuan Dong, David Medina-Ortiz, Jing Zhao, Mehdi D. Davari
Abstract
High Resolution Image Download MS PowerPoint Slide Enzyme mining is rapidly evolving as a data-driven strategy to identify biocatalysts with tailored functions from a vast landscape of uncharacterized proteins. The integration of machine learning (ML) into these workflows enables high-throughput prediction of enzyme functions─including Enzyme Commission numbers, Gene Ontology terms, and substrate specificity─as well as key catalytic properties, such as kinetic parameters, optimal temperature, pH, solubility, and thermophilicity. This review provides a systematic overview of state-of-the-art ML models and highlights representative case studies that demonstrate their effectiveness in accelerating enzyme discovery. Despite notable progress, current approaches remain limited by data scarcity, model generalizability, and interpretability. We discuss emerging strategies to overcome these challenges, including multitask learning, integration of multimodal data, and explainable AI. We outline how the convergence of machine learning, autonomous experimentation, and agentic AI systems could accelerate progress toward self-driving enzyme discovery. Together, these developments position ML-guided enzyme mining as a scalable, interpretable, and increasingly autonomous framework for uncovering biocatalysts in biotechnology, biocatalysis, and synthetic biology.