Enhancing Machine-Learning Prediction of Enzyme Catalytic Temperature Optima through Amino Acid Conservation...

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Enzyme Engineering

Background:

• Enzymes are critical industrial catalysts, but determining their optimal catalytic temperature (T_opt) is challenging.
• Existing experimental data for T_opt is limited, and computational prediction methods lack sufficient accuracy.
• Accurate T_opt prediction is essential for optimizing enzyme-driven industrial processes and pharmaceutical development.

Purpose of the Study:

• To develop an accurate computational approach for predicting enzyme T_opt.
• To investigate the impact of removing conserved amino acids on T_opt prediction accuracy.
• To provide a foundation for rapid enzyme selection for industrial applications.

Main Methods:

• A machine learning model was developed using amino acid frequency and protein molecular weight as features.
• The K-nearest neighbors regression algorithm was employed for T_opt prediction.
• The model was applied to phosphatase sequences, both complete and with conserved amino acids removed, for comparative analysis.

Main Results:

• The machine learning model achieved a mean coefficient of determination (R²) of 0.599 for complete phosphatase sequences.
• Removing conserved amino acids significantly improved the model's predictive performance, with R² increasing to 0.755.
• Experimental validation confirmed that T_opt predictions for phosphatases lacking conserved amino acids were closer to experimentally determined values.

Conclusions:

• Excluding conserved amino acids enhances the accuracy of machine learning-based T_opt prediction for enzymes.
• This approach facilitates the selection of enzymes with optimal thermostability for industrial applications.
• The study provides a valuable tool for accelerating enzyme engineering and process optimization.