AI Cuts Drug Discovery From Months to Days

Creating new medicines just got dramatically faster and cheaper, thanks to a tool that does in days what used to take months.

Researchers at the University of Utah and UCLA developed a machine learning system that predicts how atoms snap together to form potential new drugs. Instead of running 50 to 60 expensive lab experiments, chemists can now get accurate predictions from just 5 to 10 tests.

"Instead of running 50-60 reactions, we are now able to run 5-10, potentially saving weeks or months," said Erin Bucci, a doctoral student at UCLA who tested the tool. Each reaction component costs money to purchase or create from scratch, making this advancement a budget saver too.

The tool tackles one of drug development's trickiest puzzles. Molecules can exist as mirror images, like left and right hands, and one version might heal while the other could harm. Finding the right combination of ingredients to build the correct version typically requires months of trial and error.

What makes this system special is how little data it needs. Most AI requires enormous datasets to work well, but this tool was trained on just four academic papers. It then accurately predicted outcomes for completely new chemical combinations it had never seen before.

The researchers focused on asymmetric cross-coupling reactions, a powerful technique that stitches together carbon-based molecular fragments using metal catalysts. These reactions can produce 95% of the desired molecular form and just 5% of the unwanted mirror image, but only if chemists pick exactly the right ingredients.

"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, the study's co-lead author.

The Ripple Effect

Pharmaceutical companies stand to benefit immediately from this breakthrough. When a company needs to deliver large quantities of a compound, this tool can guide them to the most efficient path forward without burning through research budgets.

Beyond saving time and money, the system helps scientists understand chemistry itself. "We can learn something about the chemistry from the predictions, even if they're off," said Abigail Doyle, a chemist at UCLA and study coauthor. The workflow isn't a black box. Researchers can apply their expertise to extract insights they wouldn't have discovered without the tool.

The workflow can expand beyond nickel-based reactions to other fields of chemistry. What started as a solution for one specific type of reaction could transform how scientists approach molecular design across multiple disciplines.

The pharmaceutical industry spends billions on drug development, with many promising compounds failing during testing. A tool that accelerates the early discovery phase while cutting costs could help more medicines reach patients faster.