Protein-protein interactions (PPIs) are essential targets in drug discovery because of their association with various diseases. PPI-targeting modulators have very different physicochemical properties from conventional small molecule oral drugs, such as the "Rule-of-Five" (RO5). Therefore, it has been difficult to efficiently generate and design PPI inhibitors using conventional methods, including molecular generation models. In this study, we propose a molecular generation model based on deep reinforcement learning, specialized for generating PPI inhibitor candidates. By improving the scoring function of the existing molecular generation model for small molecules, we have made it possible to generate compounds that are likely to inhibit PPIs. For future use in a biochemical assay, we also try to build a virtual library consisting of generated compounds by the proposed method. The compounds in this library are considered more suitable for recent PPI inhibitor design than those in the existing PPI-oriented library.
各種検索
サポート
ヘルプ