Prioritizing Stability-enhancing Mutations using the ESM Protein Language Model in conjunction with Physics-based...

Area of Science:

• Biotechnology
• Computational Biology
• Protein Engineering

Background:

• Directed evolution is costly and time-consuming for protein engineering.
• Computational protein design offers a faster, cheaper alternative for generating protein variants.

Purpose of the Study:

• To evaluate computational methods for prioritizing mutations that enhance protein thermodynamic stability.
• To compare the effectiveness of AI-based (ESM) and physics-based (MM/GBSA) models.

Main Methods:

• Benchmarking a dataset of 174,945 mutations across 180 proteins.
• Assessing Evolutionary Scale Modeling (ESM) and Molecular Mechanics Generalized Born Surface Area (MM/GBSA).
• Developing and testing a hybrid mutation prioritization strategy.

Main Results:

• Both ESM and MM/GBSA showed prediction biases.
• Combining ESM and MM/GBSA mitigated individual biases and improved accuracy.
• The hybrid strategy achieved a higher average ROC AUC (0.743) than individual methods (MM/GBSA: 0.685, ESM: 0.597).

Conclusions:

• A hybrid approach using AI and physics-based models enhances protein engineering.
• This integrated framework offers a more accurate and efficient method for identifying stabilizing mutations.
• The strategy is adaptable to future advancements in computational protein design models.