Workshop: Machine Learning Directed Multiscale Simulations To Explore RAS Biology
Abstract: Computational modeling has the potential to provide significant new insights into the detailed molecular mechanisms underlying many diseases such as cancer. However, existing approaches have difficulty bridging the experimentally observable effects at the macro scale with the molecular level of detail needed to provide new insight. Here we present a first step towards a new multi-resolution framework able to address these challenges. Using emerging deep learning approaches, we couple a continuum macro-scale model at experimentally relevant time- and length-scales with a large ensemble of molecular dynamic simulations of the micro scale. Given sufficient resources, the system will converge to macro-scale results at micro-scale resolution and provide fundamentally new insights. We discuss the general ideas and initial implementation and report on our experience with the first large-scale simulation campaign that utilized all of Sierra, the second fastest supercomputer in the world.