Drug-Repurposing for Covid-19 with 3D-Aware Machine Learning

Investigators: Rafael Gomez-Bombarelli (Massachusetts Institute of Technology), Simon Axelrod (Harvard University)

Rafael Gomez-Bombarelli (left), is the Toyota Assistant Professor in Materials Processing at MIT. Simon Axelrod is a graduate student in Chemical Physics at Harvard University

Novel active therapeutics against coronaviruses like the one responsible for Covid-19 (SARS- CoV2) are in urgent need. New discovery of small-molecule drugs, however, is slow and costly. Many promising molecules that strongly interact with their biological target fail as drugs because of poor processing within the body or toxicity that is only discovered late in their development. Furthermore, the synthesis-experimentation loop is typically slow. Repurposing known drugs for the treatment of a new disease, such as Covid-19, effectively bypasses many of these challenges, since the drugs are already FDA approved.

Rafael Gomez-Bombarelli from MIT and Simon Axelrod from Harvard are exploring whether the repurposing of drugs for COVID-19 treatment can be accelerated with a combination of physical simulation and machine learning (ML). They are using affordable electronic structure simulations to calculate molecular conformations and train 3D-based message-passing neural networks from existing molecular screens against the related SARS-CoV1 and SARS-CoV2 data as it becomes available.

Publications

•  GEOM: Energy-annotated molecular conformations for property prediction and molecular generation, arXiv:2006.05531 [physics.comp-ph]
• Molecular machine learning with conformer ensembles, arXiv:2012.08452 [cs.LG] 

Data Release: https://www.dropbox.com/sh/1aptf9fi8kyrzg6/AABQ4F7dpl4tQ_pGCf2izd7Ca?dl=0