# AI tool streamlines drug synthesis

_Friday, June 26, 2026 at 8:04 PM EDT · science · Latest · Tier 2 — Notable_

![AI tool streamlines drug synthesis — Primary](https://d26toa8f6ahusa.cloudfront.net/wp-content/uploads/2026/03/06123947/Sigman_Header.jpg)

A machine learning workflow predicts how components in asymmetric cross-coupling reactions will combine to favor one mirror image form of a molecule. The system screens tens of thousands of chemical structures to forecast reaction outcomes using data converted into numerical inputs for computer analysis.

Researchers trained the model on results from only four academic papers that used nickel-based catalysts with different ligands. The workflow then made predictions for hypothetical components not included in the training data, including materials increasingly dissimilar to the original set.

Co-lead author Simone Gallarati said the statistical models make accurate predictions on untested reactions while remaining far less expensive than physics-based computational chemistry tools. Co-lead author Erin Bucci said the tool cuts lab experiments from 50 to 60 reactions down to 5 to 10, reducing time and material costs.

The study was published as an accelerated preview in the journal Nature on Feb. 11, 2026. Coauthor Matthew Sigman said the approach allows reasonably good models from smaller datasets that transfer predictions to reactions the models have not seen. Coauthor Abigail Doyle said the workflow is not a black box and can reveal new insights about the chemistry even when predictions are off.

## Sources

- [@theU](https://attheu.utah.edu/facultystaff/ai-tool-streamlines-drug-synthesis/)

---
Canonical: https://techandbusiness.org/newswire/dwShKCC5FBZlnWiQ1Ra6pm
Retrieved: 2026-06-27T04:56:28.189Z
Publisher: Tech & Business (techandbusiness.org)