Accelerating COVID-19 Drug Discovery Through Leadership-scale Graph Neural Net Machine Learning

Project Description

In urgent response to the COVID-19 pandemic, Entos is partnering with NVIDIA to fundamentally disrupt the drug-discovery process via a physics-based machine learning model for biomolecular simulation and property prediction of unprecedented scale, accuracy, and efficiency. The project will yield a leadership-scale graph-neural-net model (over 8 billion parameters), informed by billions of underlying first-principles quantum mechanical calculations. This proposal seeks the needed GPU resources to demonstrate the training efficiency of the GNN model at scale and to validate the representation generality for learning tasks that include energies, gradients, and other molecule properties such as ligand-protein binding for drug binding-affinity predictions.