Scientists have used AI to discover promising drug-like compounds.

The universe of molecules that could be turned into potentially life-saving drugs is mind-boggling in size: researchers estimate the number at around 10⁶⁰. That’s more than all the atoms in the solar system, offering virtually unlimited chemical possibilities—if only chemists could find the worthwhile ones.

Now machine-learning tools can explore large databases of existing molecules and their properties, using the information to generate new possibilities. This could make it faster and cheaper to discover new drug candidates.

In September, a team of researchers at Hong Kong–based Insilico Medicine and the University of Toronto took a convincing step toward showing that the strategy works by synthesizing several drug candidates found by AI algorithms.

Using techniques like deep learning and generative models similar to the ones that allowed a computer to beat the world champion at the ancient game of Go, the researchers identified some 30,000 novel molecules with desirable properties. They selected six to synthesize and test. One was particularly active and proved promising in animal tests.

Chemists in drug discovery often dream up new molecules—an art honed by years of experience and, among the best drug hunters, by a keen intuition. Now these scientists have a new tool to expand their imaginations.