How AI is accelerating drug research – thanks to the professor ‘trying to change chemistry’

·4-min read

Never has there been a greater reminder that time is of the essence when it comes to cutting-edge research than the global race towards a Covid-19 vaccine. The development of these vaccines in such a short space of time was possible thanks to decades of previous research. So, it’s natural to wonder; how can science get the right answers faster, to help us face an uncertain future?

The Acceleration Consortium could help solve this puzzle. A collaboration between academia, industry and government, the University of Toronto-based consortium seeks to speed up the tedious and repetitive parts of scientific discovery by using artificial intelligence (AI) and robotics.

When it comes to research, typically you don’t go in knowing what the right answer is – especially if you are trying to devise a new material or improve a drug. Researchers use trial and error, building on what is already known, to discover the unknown. Many paths are dead ends and some are successful. But whether the result is a gamechanging discovery, or a dud, it all takes time.

Wouldn’t it be useful to have a roadmap that decides which research avenues should be explored and which are scientific cul-de-sacs? That’s where AI comes in. The developers train the algorithm to solve a problem and provide insights that would either take one scientist a long time or require many scientists working simultaneously. Imagine it as an extremely well-prepared assistant that can make estimates on which approaches should be prioritised.

“AI is driving what we like to call ‘self-driving labs’, a lab where every experiment is analysed by the AI on the fly and you pick what is the next best experiment to make so you don’t waste time in the discovery process,” says Prof Alán Aspuru-Guzik, director of the Acceleration Consortium. “I am a person who is trying to change chemistry.”

It’s a bold ambition, and one that took Aspuru-Guzik from Harvard to the University of Toronto in 2018 to work on material acceleration. He is a professor of chemistry and computer science and is also the Canada 150 research chair in theoretical chemistry and a Canada CIFAR AI chair at the Vector Institute, an independent, not-for-profit corporation dedicated to research in the field of artificial intelligence.

“We moved from Boston to Toronto and I decided to start doing material acceleration. I wanted to see if we could build the University of Toronto as the global centre of this movement. So, if you hear materials acceleration platforms or self-driving labs, you think of Toronto, Canada.”

Aspuru-Guzik’s Acceleration Consortium aligns with the University of Toronto’s interdisciplinary culture. The self-driving lab concept has already benefited a diverse range of research there, from material sciences and chemistry, through to biomedical research – including research into combating Covid-19 led by Dr Masoud Vedadi, associate professor in the department of pharmacology and toxicology.

Related: ‘The obesity and diabetes epidemic is unsustainable’: how a team of academics aims to halt the trend

His research looks at disrupting two enzymes that are crucial for the replication of SARS-CoV-2, the virus responsible for Covid-19. The self-driving lab approach is helping predict which compound would work better in disrupting these enzymes, based on current knowledge, shaving years off the process, and hopefully delivering an effective drug soon.

“We are working on drug discovery, and drug discovery is a long process. On average it takes 10 years to start from the initial experiments to get to the point that, for example, the FDA approves the drug. So anything which can help to shorten this process is fantastic to use,” says Vedadi. “Every minute that a drug is delayed might equal many people dying. The recent pandemic actually showed us how time is crucial.”

While the self-driving lab is already working in many aspects, it is still considered the lab of the future. The issue right now is scaling up; every particular problem that could benefit needs its own bespoke self-driving lab.

“What we need to do is advance the learning curve of material acceleration platforms,” Aspuru-Guzik says. “We need a few tens of millions of dollars to start making more [acceleration] platforms … we need to find the commonalities between [different acceleration platforms], simplify them, generalise them, make it more accessible. That’s the mission of the consortium.”

Dr Alfredo Carpineti is an astrophysicist and science writer

Our goal is to create a safe and engaging place for users to connect over interests and passions. In order to improve our community experience, we are temporarily suspending article commenting