Abstract
The rapid development of artificial intelligence (AI) has brought us many unique features in recent years. Many of these new advances, represented by AlphaFold, demonstrate that AI can also play an essential role in molecular science. Our group focuses on developing AI-based algorithms, models and software for molecular modelling and simulation, which involve supervised, unsupervised, and reinforcement learning. We developed a deep molecular model MolCT based on the graph neural networks (GNN), which enables efficient and transferable representation learning for molecular systems. We also developed TALOS, a deep reinforcement learning paradigm that borrows ideas from generative adversarial networks (GAN) and can be used for enhanced sampling, force field optimization, and transition state search. In developing these methods, we were aware that the most currently popular molecular dynamics (MD) simulation software has difficulty being well-compatible with these new algorithms based on AI frameworks. Therefore, our group collaborated with Huawei Technologies Co., Ltd. to develop MindSPONGE, a new MD simulation software based on Huawei’s AI framework MindSpore. MindSPONGE is an AI-native MD simulation software, which is not only fully compatible with AI-based algorithms but also makes the best use of the features of the AI framework to extend the capabilities of MD simulation. Our aim was for MindSPONGE to lead the way in developing next-generation MD simulation software.
Biography
Dr. Yi Isaac Yang received his B.Sc. degree from Shandong University in 2010 and his PhD degree from Peking University in 2015. He was a postdoctoral researcher in the group of Professor Michele Parrinello at ETH Zürich and USI Lugano from 2016 to 2019. Dr. Yang joined the faculty of Professor Yi Qin Gao’s group at Shenzhen Bay Laboratory in 2019, where he is now an associate investigator. Dr. Yang is a Senior Technical Evangelist of MindSpore and a MindSpore Technical Experts Committee member. His current research interests focus on developing theories, algorithms and software for molecular dynamics simulation and artificial intelligence.
Seminar Series by the NYU-ECNU Center for Computational Chemistry at NYU Shanghai