Open-source competition finds potential coronavirus drugs

The CACHE (Critical Assessment of Computational Hit-Finding Experiments) Challenge has spurred more new research—this time identifying seven promising molecules that could become pan-coronavirus drugs.

Twenty-two teams were challenged to find a molecule that could bind to the RNA binding site on the NSP13 helicase, a SARS-CoV-2 replication protein that is preserved across multiple coronavirus types.

These molecules, or “hits,” were identified in silico by teams using different drug discovery platforms. These hits were subsequently tested experimentally at the Structural Genomics Consortium at the University of Toronto to see if they could actually bind to the identified target. The molecules and the data behind them are freely available for anyone to develop.

According to Ryan Merkley, CEO of Conscience, which organized the competition, these open science competitions offer drug discovery groups a way to see how their technology measures up against others.

“Pharma, very traditionally, works in its silos,” he says. “There aren’t a lot of places where there are apples-to-apples comparisons on performance.”

The winning team, determined by the number and strength of molecules identified, was led by Karina dos Santos Machado, who hails from the Universidade Federal do Rio Grande and Universidade Federal de Pelotas—both universities in Brazil. The team used both open-source and internally developed AI methods to identify their molecules.

Another high-scoring team from Vanderbilt University, led by Rocco Moretti, used Drugit, a design mode of the citizen-science game Foldit. Players were given a series of puzzles to design a molecule, and over 150 players submitted 7598 candidate compounds. After winnowing the number down to 111 compounds to order and test, CACHE chose one molecule as a hit.

“I think there’s a lot of room for human intuition in discovery of these new molecules,” Moretti says. “There’s a lot we don’t necessarily understand about protein–small molecule interactions. And so having that human insight, as far as what might fit into the pocket, how the hydrogen bonding pattern should be, I think there’s a contribution that can be made by citizen scientists.”