Computational screening finds elusive drug leads among hundreds of millions of compounds

Finding a promising drug from within the vastness of chemical space is a task that’s often described as finding a needle in a haystack—though arguably, it’s much more difficult. Now, two teams of scientists have done just that, finding drug leads that show promise as treatments for depression and pain by virtually screening a library of 75 million molecules in the first case and 301 million molecules in the second.

Virtual libraries of compounds have been around for a long time, but drugmakers didn’t have the confidence that they could make the molecules in them without a lot of effort, says Brian Shoichet, an expert in molecular docking at the University of California, San Francisco, who was a leader on both teams. That’s all changed in recent years, Shoichet says, as academic labs and specialty chemical companies like Enamine have developed large libraries of building blocks with interesting chemical motifs. “It’s not just methyl, ethyl, propyl, butyl, futile around a small number of scaffolds,” he says. Rather, the new libraries are full of diverse molecules, “and it’s that chemical difference that plays out into the new biology they reveal,” Shoichet says.

In the new reports, both teams were looking to disentangle the beneficial properties of certain drugs from those drugs’ liabilities. One team looked at a bespoke library of tetrahydropyridines to find molecules that could bind to the serotonin 5-HT_2A receptor (5-HT_2AR)—a target for antidepressants—without also initiating the signaling that is thought to cause hallucinations in drugs like LSD and psilocybin (Nature 2022, DOI: 10.1038/s41586-022-05258-z).

The other team used the ZINC15 library of commercially available small molecules to search for compounds that would bind to the α_2A-adrenergic receptor (α_2AAR)—a pain relief target—without initiating the signaling process that causes sedation associated with opioids (Science 2022, DOI: 10.1126/science.abn7065).

In both studies, screening the vast libraries led the researchers to more than a dozen promising compounds, which they synthesized and tested for their ability to bind the targets. The scientists optimized the best of these agonists and tested them in mice. Two molecules that bind to 5-HT_2AR, (R)-69 and (R)-70 (shown), showed antidepressant activity: mice that were given the compounds swam for longer or struggled for longer when suspended by their tails compared with mice that had not taken any drugs. Neither compound induced hallucinations in the mice, which is measured by head twitching in the rodents. The most promising agonists in the α_2AAR study, ‘9087 and PS75 (shown), relieved pain in mice without causing sedation.

David E. Gloriam, an expert in computational drug design at the University of Copenhagen, says the drug leads found from these studies are potentially game-changing. Given that 86% of agents in clinical trials fail, “these studies open critical paths towards safer drugs for depression and pain mitigating the opioid crisis,” he says in an email.

Shoichet says that scientists are looking to further develop the most promising compounds from these studies.