BPS2026 – Utilizing AlphaFold2 multi-state modeling for improving structure-based drug discovery

As the name suggests, structure-based virtual screening is contingent on the receptor structure employed. The receptor conformation is generally treated as rigid, despite the high flexibility of proteins. The resolution of this issue is achieved through the implementation of an ensemble screening method, which employs a multitude of structures obtained from the Protein Data Bank (PDB) and/or molecular dynamics simulations. The structures for the ensemble screening should be as diverse as possible to cover the conformational space. Recent advancements in structure prediction methods, such as AlphaFold, were anticipated to address this challenge by providing more precise models. However, it has been observed that the predicted structures exhibit a bias toward the conformations that are present in the PDB. Recently, a methodology referred to as multi-state modeling was presented, which furnishes a template of a specific conformational state to AlphaFold2 to facilitate the attainment of a desired structure. In this talk, the results of the multi-state modeling will be presented and discussed in the context of structure-based drug discovery for kinases, a prominent target family characterized by its diverse structural complexity.