Two and a half years later, much of the world has settled into an uneasy routine with Covid-19 thanks to a host of highly effective vaccines and a handful of effective therapeutic drugs. Variants loom over this tenuous peace, though, and researchers are eager to streamline the discovery process for new drugs and vaccines for the next variant — or the next pandemic. Now, researchers at Argonne National Laboratory are using a combination of supercomputing and machine learning (ML) to accelerate these drug discovery processes.
If the combination of AI and supercomputing for Covid drug discovery sounds familiar, you’re not wrong — but the sheer scale of molecular candidates means that even after years, the process of sifting through potential drugs remains onerous. “When you have billions of possibilities, you have to narrow them down to a few thousand before you can proceed with these procedures,” said Agastya Bhati, a researcher from University College London, in an interview with Argonne’s Emily Stevens. “It is a huge task, and it is tedious.”
“The biggest ambition relates to the conflation or joining of machine learning methods to accelerate discovery through first looking at huge numbers of potential compounds and then ranking them in a way that’s prioritized based on increasingly accurate calculations,” explained Peter Coveney, a professor at University College London and lead of the project.
So the researchers developed an IMPECCABLE method — with “IMPECCABLE” here serving as an ambitious backronym for “Integrated Modeling Pipeline for Covid Cure by Assessing Better Leads.” A modular workflow, IMPECCABLE cycles through a variety of more intensive physics-based simulations and broader, computationally cheaper machine learning analyses meant to brush away the low-hanging fruit. (One of the more intensive physics-based methods, developed by Bhati, is the TIES method, which stands for “Thermodynamic Integration with Enhanced Sampling.”)
“The overall workflow is comprised of about four separate workflows which are already large scale,” Coveney said. After the set of four workflows — which Coveney said could take as little as 24 hours — you could have your best ranking of the molecules and move to lab analysis.
IMPECCABLE was made possible by a grant from the Department of Energy’s INCITE program that allowed the team access to the Theta supercomputer at Argonne. Theta, an HPE system, delivers 6.92 Linpack petaflops and ranked 70th on the most recent Top500 list. Bhati praised Argonne’s hardware for this use case: “[They] have specific hardware which is optimized for running machine learning algorithms in a nice way and also have high-performance computing infrastructure where we can run a large number of simulations in parallel.”
Next, the researchers intend to ramp up use of IMPECCABLE for drug discovery related to Covid and other viruses.
To learn more about this research, read the reporting from Argonne’s Emily Stevens here.