創薬ターゲット探索のための富岳へのプラットフォーム構築

抄録

The demand for AI in drug discovery has been increasing, and various AI application prototypes have been developed to optimize the drug development process. However, unfortunately, these AI applications are not yet widely used in pharmaceutical companies' drug discovery efforts. One of the reasons for this is the decentralized nature of individual applications and the lack of an integrated platform. In the current state, researchers cannot easily utilize applications when needed.

To address this issue, we have incorporated approximately 20 AI and analytical technology applications we have developed into an integrated platform in a common computing infrastructure centered around the Supercomputer “Fugaku.” This platform is designed to construct workflows for drug target discovery, estimating target proteins. Specifically, we have incorporated information technologies like Graph Convolutional Networks and Bayesian Networks, which can predict candidate drug target molecules from large-scale biomedical data, including clinical data, omics data, and literature data. Ultimately, this platform allows the input of disease names, patient sample data, and more, enabling the estimation of disease mechanisms and target proteins using an HPC/AI workflow.