Advances in Machine Learning Models for Predicting Enzyme Kinetic Parameters.

Area of Science:

• Biochemistry
• Computational Biology
• Machine Learning

Background:

• Enzyme kinetic parameters (kcat, Km, kcat/Km, Ki) are vital for enzyme engineering, metabolic modeling, and synthetic biology.
• Experimental determination is costly and time-consuming; traditional computational methods are insufficient.
• Machine learning (ML) models offer a promising alternative for in silico prediction of these parameters.

Purpose of the Study:

• To review recent advancements in ML-based prediction of enzyme kinetic parameters.
• To highlight the applications and limitations of current ML models.
• To outline future opportunities for improving ML predictions.

Main Methods:

• Review of global ML models trained on diverse enzyme classes.
• Review of local ML models tailored to specific enzyme families.
• Discussion of ML model applications in mutation effect prediction, enzyme mining, and metabolic modeling.

Main Results:

• ML models, both global and local, have shown success in predicting enzyme kinetic parameters.
• These models aid in various applications, including protein engineering and systems biology.
• Data scarcity is a primary limitation, impacting model performance and scope.

Conclusions:

• ML offers a powerful approach to predict enzyme kinetic parameters, accelerating research in enzyme engineering and synthetic biology.
• Overcoming data scarcity through high-throughput data generation and semisupervised learning is key for future progress.
• Accurate ML-based prediction facilitates better protein sequence annotation for desired functions.