Skip to content


Using deep learning for structure-based rational drug design, without the structure

The genes linked to human disease number in the thousands — but in all but a few hundred cases, researchers don’t understand the structure or activity of the proteins these genes encode, and therefore don’t have the data they need for traditional drug design. Atomwise uses convo­lu­tional neural networks, similar to those used in speech recognition or computer vision, to screen billions of synthe­siz­able small molecules for those that might interact with target proteins, even without full structure data.

Explore Companies