Abstract

In this talk, I will describe some recent attempts to carry out molecular science studies during COVID times. Specifically, I will describe two projects: (1) Crowd-sourced attempts to search the space of ML strategies and develop algorithms for predicting atomic-pairwise nuclear magnetic resonance (NMR) properties in molecules. Using an open-source dataset, we worked with Kaggle to design and host a 3-month competition which received 47,800 ML model predictions from 2,700 teams in 84 countries. Within 3 weeks, the Kaggle community produced models with comparable accuracy to our best previously published ‘in-house’ efforts. A meta-ensemble model constructed as a linear combination of the top predictions has a prediction accuracy which exceeds that of any individual model, 7-19x better than our previous state-of-the-art. [1] (2) Efforts to develop Narupa, [2] a flexible, open-source, cloud-mounted, multi-person VR software framework which enables groups of researchers distributed across the world to simultaneously cohabit real-time simulation environments and interactively build, inspect, visualize, and manipulate the dynamics of complex molecular structures with atomic-level precision. [3,4] I will outline a range of application domains where we are using Narupa to obtain microscopic insight into 3D dynamical concepts and enable effective research and communication, including protein-ligand binding, [5] and machine learning potential energy surfaces. [6]

1. L. A. Bratholm et al., “A community powered search of machine learning strategy space to find NMR property prediction models,” arxiv. 2008.05994 2. M. O’Connor et al., An open-source multi-person virtual reality framework for interactive molecular dynamics: from quantum chemistry to drug binding, J. Chem Phys 150, 224703, 2019. 3. M. O’Connor et al., Sampling molecular conformations and dynamics in a multiuser virtual reality framework. Science Advances, 2018, 4 (6). 4. https://vimeo.com/420036282 5. H. M. Deeks et al., “Sampling protein-ligand binding pathways to recover crystallographic binding poses using interactive molecular dynamics in virtual reality”, arXiv:1908.07395, 2019 6. S. Amabilino et al., Training neural nets to learn reactive potential energy surfaces using interactive quantum chemistry in virtual reality. J Phys Chem A, 2019, 123, 20, 4486, 2019