MISATO - Machine learning dataset for structure-based drug discovery

Developments in Artificial Intelligence (AI) have had an enormous impact on scientific research in recent years. Yet, relatively few robust methods have been reported in the field of structure-based drug discovery. To train AI models to abstract from structural data, highly curated and precise biomolecule-ligand interaction datasets are urgently needed. We present MISATO, a curated dataset of almost 20000 experimental structures of protein-ligand complexes, associated molecular dynamics traces, and electronic properties. Semi-empirical quantum mechanics was used to systematically refine protonation states of proteins and small molecule ligands. Molecular dynamics traces for protein-ligand complexes were obtained in explicit water. The dataset is made readily available to the scientific community via simple python data-loaders. AI baseline models are provided for dynamical and electronic properties. This highly curated dataset is expected to enable the next-generation of AI models for structure-based drug discovery. Our vision is to make MISATO the first step of a vibrant community project for the development of powerful AI-based drug discovery tools.近年来，人工智能（Artificial Intelligence）的发展对科学研究产生了巨大影响。然而，在基于结构的药物发现领域，已报道的稳健方法仍相对匮乏。为训练AI模型从结构数据中提炼抽象信息，亟需获取经高度精心整理且精准的生物分子-配体相互作用数据集。本研究提出MISATO数据集：该数据集包含近20000个蛋白质-配体复合物的实验结构、配套的分子动力学轨迹以及电子性质数据。研究采用半经验量子力学方法，系统优化了蛋白质与小分子配体的质子化状态。蛋白质-配体复合物的分子动力学轨迹是在显式水模型环境下获取的。该数据集可通过简洁的Python数据加载工具向科研社区便捷开放。针对动力学与电子性质任务，本数据集配套提供了AI基准模型。这一经过高度精心整理的数据集有望推动面向基于结构药物发现的新一代AI模型的发展。我们的愿景是将MISATO打造为充满活力的社区项目的第一步，助力开发基于AI的高性能药物发现工具。