Researchers announced on Mar. 9 that a new machine learning system could make drug discovery faster and less expensive by predicting chemical reactions more efficiently. The study, published as an accelerated preview in the journal Nature on Feb. 11, describes how the tool can help chemists build better molecules for medicines by streamlining the process of testing and optimizing chemical reactions.
The development is significant because creating new drugs typically requires extensive trial and error, which consumes large amounts of time and resources. By using artificial intelligence to predict outcomes, scientists hope to reduce both costs and labor involved in developing pharmaceuticals.
“Sometimes we use sophisticated, physics-based computational chemistry tools to understand novel reactions. However, these tools are too expensive to make predictions on thousands of potential new molecules,” said Simone Gallarati, co-lead author and joint postdoctoral researcher at the University of Utah and the University of California, Los Angeles. “We wanted to train statistical models that were ‘smart’ enough to make accurate predictions on untested reactions, but also as cheap as possible.”
The research focused on asymmetric cross-coupling reactions—a key method for constructing complex molecules with specific three-dimensional arrangements known as handedness. These arrangements are crucial because different versions can have very different effects in the body; one may be therapeutic while its mirror image could be harmful.
Matthew Sigman, chemist at the University of Utah and coauthor of the study, said: “Most AI requires enormous amounts of data to train models on. That’s a problem in chemistry by which obtaining high-quality, large datasets from experimental work is very expensive and extremely time consuming. The coolest thing about this tool is that it allows someone to collect smaller bits of data, build reasonably good models and make accurate predictions for known reactions, and also transfer predictions to reactions that the models haven’t seen yet.”
The team trained their model using results from four academic papers involving nickel-based catalysts with various ligands. They then challenged the system with hypothetical components not included in its training data. Lab tests led by Erin Bucci at UCLA confirmed that the model could reliably forecast reaction outcomes with much less experimental effort than traditional methods.
“As a lab-based chemist, this tool is extremely valuable for saving time spent running experiments,” Bucci said. “For example, instead of running 50-60 reactions, we are now able to run 5-10, potentially saving weeks or months. Each reaction component we test in the lab needs to either be purchased or made from scratch—this tool greatly cuts the amount of money I would typically spend on materials.”
Abigail Doyle, chemist at UCLA and coauthor of the study, added: “One of the nice things about the workflow is—it’s not a black box. We can learn something about the chemistry from the predictions, even if they’re off. We apply our chemistry expertise to help learn something we wouldn’t have learned without the tool.”
Sigman said pharmaceutical companies could benefit immediately from such technology when scaling up compounds for clinical trials: “This is where this tool could be highly applicable… Optimizing a reaction and the time-cost is the value proposition when you build a drug. This streamlined process could make the difference when they need to take a molecule from phase one to phase two.”
