Approaching Optimal pH Enzyme Prediction with Large Language Models.

Area of Science:

• Biotechnology
• Computational Biology
• Enzymology

Background:

• Enzymes are crucial biocatalysts in various industries, but their activity is highly sensitive to pH.
• Identifying enzymes with optimal activity at specific pH levels is a significant challenge in enzyme engineering.
• Current methods for enzyme optimization often require extensive experimental screening.

Purpose of the Study:

• To develop an accurate computational method for predicting the optimal pH range of enzymes directly from their amino acid sequences.
• To enable rapid in silico screening of enzyme candidates for desired pH profiles.
• To support the design of novel enzymes for specific biotechnological applications.

Main Methods:

• A language-model-based approach was developed to predict the optimal pH range from enzyme sequences.
• Machine learning models were trained and validated using diverse datasets.
• Robustness was assessed using various data splitting strategies, including sequence similarity, protein family, and enzyme classification.

Main Results:

• The developed models demonstrated high prediction accuracy for optimal pH across different protein families and sequence similarities.
• The approach proved robust even when applied to proteins with low sequence similarity to the training set.
• The method is computationally efficient, facilitating high-throughput virtual screening.

Conclusions:

• The language-model-based approach provides a fast and accurate in silico tool for predicting enzyme optimal pH.
• This method can significantly accelerate the discovery and engineering of enzymes for targeted biotechnological applications.
• The approach enhances the efficiency of enzyme design by reducing the need for extensive experimental validation.