Mount Sinai researchers use AI to discover drugs targeting opioid addiction

Leslie Salas Estrada sits in a lab at the Icahn School of Medicine at Mount Sinai with a problem that has stumped pharmaceutical researchers for years: how to find, among millions of molecular candidates, the few compounds that might actually block the kappa-opioid receptor and help people escape addiction.

The science behind the approach is straightforward enough. When someone uses opioids repeatedly, their brain chemistry shifts. The neural pathways that drive craving and dependence become entrenched. Withdrawal—the physical and psychological torment that follows cessation—becomes a wall too high for most people to climb. Animal studies have shown that blocking the kappa-opioid receptor can reduce that wall. It can dampen the brain's rewired demand for the drug during the worst of withdrawal, when relapse is most likely.

But finding a drug that actually does this blocking is another matter entirely. The universe of possible molecular structures is vast. Testing each one through traditional methods—synthesizing it, running it through assays, waiting for results—takes time and money that most research budgets cannot sustain. This is where Estrada's approach diverges from the conventional path. She is using artificial intelligence to compress the search space, to let algorithms screen through thousands of candidates and surface the most promising ones for human researchers to examine more closely.

Estrada explained the human stakes plainly. Addiction is not a failure of willpower. It is a biological reorganization. "After a lot of opioid exposure, your brain gets rewired to need more drugs," she said. When someone tries to quit, withdrawal symptoms arrive with force—physical pain, anxiety, insomnia, cravings that feel like survival instincts. For many people, those symptoms are insurmountable without help. A drug that could ease them, that could reduce the neurological demand for opioids during recovery, could change the trajectory of treatment.

The kappa-opioid receptor has emerged as a target precisely because preclinical work—studies in animal models—has suggested that blocking it might do exactly that. It is not a cure. It is a tool. But in the context of addiction treatment, where the gap between wanting to quit and actually quitting can be measured in the number of people who relapse, a tool that shifts the odds even slightly matters enormously.

Estrada's use of computational methods to accelerate drug discovery is not new in principle, but applying it to this particular problem, with this particular receptor, represents a focused effort to translate what we know works in mice into something that might work in people. The next phase—moving from computational prediction to laboratory synthesis to animal testing to, eventually, human trials—remains years away. But the bottleneck that has always slowed this process, the sheer difficulty of finding the right molecule in the first place, is what artificial intelligence is positioned to solve.