COMbining Deep-Learning with Physics-Based AffinIty EstimatiOn 4 (COMPBIO4)”

PI Peter Coveney, Peter Coveney, University College London
Co-PI Shantenu Jha, Pacific Northwest National Laboratory
Philip Fowler, University of Oxford
Ola Engkvist, Astra Zeneca
Eric Stahlberg, MD Anderson Cancer Center
Mariano Vazquez, ELEM Biotech
Dilip Asthagiri, Oak Ridge National Laboratory
Antigoni Georgiadou, Oak Ridge National Laboratory
Francis Joseph Alexander, Argonne National Laboratory
Tom Beck, Oak Ridge National Laboratory
Rick L Stevens, Argonne National Laboratory
Award INCITE
Hours 500,000 Aurora node-hours
Year
Domain Biological Sciences
The IMPECCABLE workflow integrates machine learning (ML) and physics-based (PB) methods to accelerate drug discovery. By incorporating multiple ML and PB approaches, it leverages their strengths while mitigating individual limitations. The workflow is being scaled on exascale machines to efficiently identify potential drug candidates for various target proteins.

Peter Coveney, University College London

Project Summary

This project will use exascale computing, AI, and multiscale simulation to advance human digital twins for personalized medicine, focusing on molecular-scale drug discovery and organ-scale hemodynamics.

Project Description

The advent of exascale computing and advances in artificial intelligence (AI) have opened up immense possibilities to realize a fully personalized digital twin (DT) of the human body. DT technology will not only enable clinicians to make reliable and actionable predictions to support clinical decision-making, but it will also facilitate the adoption of informed lifestyle choices that support healthy ageing and well-being. However, achieving this aim is challenging. 

To address this challenge, our research goal is to advance the modeling and simulation of the human body in health and disease such that we are at the forefront of the development of human DTs for applications in personalized medicine and healthcare. This requires access to large-scale computing resources. We aim to combine multiple scales of human biology through simulation combined with AI to deliver actionable outcomes in the medical context. In this project, we focus our attention on two specific use cases of human DTs at two different scales: (a) drug discovery and personalized medicine (molecular scale) and (b) hemodynamics (organ scale).

Project Type
Learning
Domains
Allocations

More Biological Sciences Projects