MU researcher's AI model advances coronavirus drug development

A University of Missouri professor is among those who have successfully identified new early-stage molecules that could be used to combat future coronaviruses.

Dr. Dimitri Kireev, a computational chemist at MU, ranked third in the latest CACHE – or Critical Assessment of Computation Hit-finding Experiments – Challenge from the non-profit biotech company, Conscience.

The challenge asked participants to use their own AI drug-design algorithms to predict new molecules that could be used as the basis for future drug development.

This iteration of the challenge focused on the family of coronaviruses, not COVID-19 specifically, in hopes that the AI-predicted molecules would bind to a specific part of the virus.

Ryan Merkley, the CEO of Conscience, said by focusing on coronaviruses as a whole, future drugs developed from these newly identified early-stage binding molecules may be more resistant to virus mutations.

“If there was a discovery on this target, it would be something that would be resilient to the many mutations that we potentially see,” Merkley said. “Because it's a conserved target, which just means it exists in all the variants.”

These new molecules are one of the earliest stages of the drug development process and it is likely years before patients would see any future antiviral therapeutics, like Paxlovid.

Merkley said the goal of the CACHE challenges is to encourage “radical collaboration” in drug development, a field of science that historically has been largely limited to large pharmaceutical companies with large budgets.

“We're really trying to shorten the path to the next step and eliminate the friction that keeps people from working on important conditions,” Merkley said.

Kireev and his team were among the 22 teams from across the world selected to participate in the challenge. Participants were allowed to submit up to 100 potential molecules, which were then tested experimentally by scientists at Conscience.

Kireev says this experimental step, in particular, can be timely and costly – further preventing wide-scale participation in drug development. So, it was a big deal that Conscience tested the potential new molecules in the lab setting and determined which ones worked – at no cost to these smaller academic groups and companies.

Illustration by Jennifer Harris

/

The University of Texas at Austin

“It’s an equalizer, indeed, between players of different scales in drug discovery, which, obviously, eventually is a great benefit to patients,” Kireev said.

Kireev, alongside postdoctoral scholars Xiaowen Wang and Akhila Mettu, used their AI model, called FRASE-bot (short for "FRAgments in Structural Environments hit-finding robot"), to design 100 possible molecules. Only a handful were then experimentally confirmed as “active,” or able to bind to the identified target on the coronavirus.

The team was then asked to present additional possible analogs of those active molecules, which underwent additional testing to find which were successful and most promising.

According to Conscience, of the 2,576 molecules proposed by the 22 teams, seven new early-stage molecules were identified as “promising.”

All of the data from the CACHE challenge, including the new molecules, are now available for free to any interested party online.

“The fact that we have many more people in the play who would contribute to the drug [development process], this would also make drug discovery more efficient and would bring higher success rate into drug discovery,” Kireev said. “So that patients will get new treatments faster and for lower.”

Kireev is the only researcher who was identified as a top performer in CACHE challenge #2, which focused on coronaviruses, and in CACHE challenge #1 announced in January, which focused on Parkinson’s disease.

“I came into pharmaceutical research, into drug discovery, because I hope that my computational skills – whatever I can do – would eventually benefit to the patients,” Kireev said. “That it's not just mathematical or computational exercise. This is something that will positively affect lives of many people.”

Kireev added the CACHE challenges have already opened doors for new research projects aimed at moving their molecules into the next steps of drug development.