The Future of AI in Drug Discovery

As we wrap up 2024 and look ahead to 2025 and beyond, it is clear that artificial intelligence has become an indispensable part of our everyday lives. And the end of the year seems like an opportune time to make predictions about the future. So, what is the future of AI in drug discovery?

AI is already ubiquitous in drug discovery today. Almost every company uses some form of AI for some aspect of their work: from predicting how a drug will bind to its target, to making sense of large datasets, to suggesting and even designing molecules or experiments. Early in the AI revolution, many people predicted that AI will come to replace humans in drug discovery, removing our inherent bias from the process. But I disagree with that prediction.

In my view, AI for drug discovery is like a car. There are different makes and models of AI tools that specialize in different functions. You need to choose the right kind of AI for the job at hand, just as you don’t want an F1 racer for getting groceries or a semi truck for driving your kids to soccer practice. And we are seeing a few leading companies who are making the most popular models that most people use, with a variety of smaller providers with highly specialized offerings. But importantly, a car can’t prepare your dinner or set a broken bone. It is just one (admittedly indispensable, for many people) tool in the toolbox of modern life.

Building on this analogy, we’ve passed the AI equivalent of the early 1900’s where everyone had to build the tool that they wanted to use. Right now we’re in the equivalent of the mid-20th century, where there are off-the-shelf options and just about everyone has access to at least one of them. But at the same time, every owner is still expected to know how to change their own oil or fix their own radiator. In the next few years I think we’ll get to the AI equivalent of early 21st century cars, where most companies will use off-the-shelf solutions without looking “under the hood” and there will be specialized people whose jobs are to tinker with the internal workings as needed, either to fix problems or to customize the functionality. And in 5-10 years we may indeed see “self-driving” AI that can handle discreet parts of the drug discovery process. But will AI be able to do the drug discovery equivalent of deciding on your next vacation spot, making all the necessary reservations, packing your bags, driving you to the airport and onto the plane, flying the plane, and going see the museums for you, seamlessly and without intervention or guidance from you? I don’t think so.

I’m sure there are many people who would disagree with my prediction. And the fun thing about predictions is that they’re all likely to be wrong in some way. What do you think will be the future of AI in drug discovery?