Protein structure prediction from sequence has dramatically improved in both accuracy and efficiency in recent years, sparking renewed interest in applying artificial intelligence (AI) approaches to structural biology. In this study, we introduce bAIes, a Bayesian approach that combines the strengths of AlphaFold with physics-based molecular dynamics simulations, while addressing their individual limitations. bAIes substantially increases the likelihood of sampling protein conformations that are relevant for small-molecule docking. We demonstrate that our approach outperforms traditional molecular dynamics simulations in effectively sampling ligand-binding pocket conformations, leading to marked improvements in docking accuracy. Furthermore, in large virtual screening campaigns, bAIes shows improved discrimination between binders and non-binders compared to AlphaFold models or models refined with molecular dynamics. By enhancing the usability of AlphaFold2 models without requiring extensive experimental or computational resources, bAIes offers a practical solution to a longstanding challenge in structure-based drug design, with the potential to accelerate the early phases of drug discovery.
Sen et al. (Sun,) studied this question.