Prioritizing Stability-enhancing Mutations using the ESM Protein Language Model in conjunction with Physics-based MM/GBSA Predictions.
TL;DR
This work investigates the effectiveness of two different computational methods as triaging tools for prioritizing target positions and identifying specific mutations that are likely to improve protein thermodynamic stability and proposes a hybrid mutation prioritization and selection strategy that achieves better accuracy than either method alone.
Abstract
Directed evolution for protein engineering, as currently practiced in the biotechnology and pharmaceutical industries, is both tedious and expensive. Computationally driven protein design has the potential to expedite the engineering process and generate high-quality variants at a lower cost than traditional approaches. We investigated the effectiveness of two different computational methods as triaging tools for prioritizing target positions and identifying specific mutations that are likely to improve protein thermodynamic stability. Our benchmarking study used a comprehensive dataset consisting of 174,945 mutations across 180 distinct proteins and evaluated the ESM (Evolutionary Scale Modeling) protein language model alongside a physics-based method, MM/GBSA (Molecular Mechanics Generalized Born Surface Area). We found prediction biases in each method but also determined that these biases can be mitigated by applying the two methods in a complementary manner. We propose a hybrid mutation prioritization and selection strategy that achieves better accuracy than either method alone. Through re-ranking, the combined prioritization strategy attained a higher overall average ROC (receiver operating characteristic) AUC (area under curve) of 0.743 across the dataset compared to either MM/GBSA alone (0.685) or ESM Log Odds alone (0.597). The integrated framework can be adapted and applied to newer AI and physics-based models as the field advances.