A recent study published in the Proceedings of the National Academy of Sciences suggests that a novel drug developed using artificial intelligence can significantly reduce fentanyl consumption in animal models. The experimental medication targets specific serotonin receptors in the brain to restore neural pathways altered by addiction. These findings provide evidence that this new compound could eventually offer a non-addictive treatment option for people experiencing opioid use disorder.
Opioid use disorder currently affects millions of people, with synthetic opioids like fentanyl driving a severe public health crisis. Seeking alternative treatments, scientists focused on creating a therapy that addresses the neurological changes caused by addiction without relying on opioid-based medications.
“New therapeutics for opioid use disorder are desperately needed,” said study author Christie D. Fowler, a chancellor’s fellow and professor in the Department of Neurobiology and Behavior at the University of California, Irvine. “Just about everyone has been impacted by the opioid epidemic. These are people’s mothers, fathers, sisters, brothers, sons, and daughters.”
Fowler, who also serves as the co-director of the UC Irvine Center for Addiction Neuroscience, noted that opioid use disorder does not discriminate based on age or socioeconomic boundaries. “Thus, if we can help people who still have [so] much benefit to give our society, and their families, then I believe that it is our ethical responsibility to do so,” she explained.
Historically, the pharmaceutical industry has hesitated to invest heavily in this area. “For too long, drug companies have stayed out of the addiction therapeutic space due to the misperception that the people that would buy their drugs wouldn’t be able to afford them – based on the stigma of an ‘addict’,” Fowler said. “This is just not the case. We cannot give up hope on people who have an opioid use disorder, just like we wouldn’t give up hope on someone who has another disease, like cancer or diabetes.”
Current treatment options, such as methadone and buprenorphine, tend to be limited by safety concerns, inconsistent long-term effectiveness, and difficulties with patient adherence. These medications are also opioid-based, which presents additional challenges for long-term recovery.
To find a different path, the researchers analyzed how fentanyl interacts with the serotonin system, a network of chemical messengers involved in mood, reward, and learning. Prolonged opioid use alters these serotonin pathways and changes the physical structure of brain cell connections. The authors aimed to discover a drug that could target these specific serotonin-related changes rather than opioid receptors.
“The development of this new drug will hopefully lead to more therapeutic options for those attempting to quit or reduce use of an opioid,” Fowler said. “Moreover, the targeting approach does not include the receptors that opioids directly act on in the brain, so this new drug may lead to more beneficial and persistent changes in the brain to overcome the harms induced by prior opioid exposure.”
The researchers used an artificial intelligence platform to analyze large sets of genetic and chemical data from the postmortem brain tissue of human patients who had been dependent on opioids. This system identified two specific types of serotonin receptors as highly probable targets for a new therapy.
“In these studies, we used an AI platform developed by GATC Health, so these studies also provide evidence to support the use of AI technology to reduce the time needed for drug development efforts,” Fowler noted. Based on these predictions, the artificial intelligence system generated dozens of potential chemical compounds, leading the researchers to select two top candidates named GATC-021 and GATC-1021.
The researchers first conducted laboratory tests on cells to confirm that the two new drugs accurately targeted the intended serotonin receptors. They found that GATC-021 activated the desired receptors but also caused unwanted side effects. In tests involving 14 rats exploring an open enclosure, the highest dose of GATC-021 significantly reduced their normal movement and exploration.
In contrast, GATC-1021 demonstrated high precision. Laboratory cell tests revealed that it selectively activated the target serotonin receptors without binding to unrelated pathways. When tested in a group of eight rats, GATC-1021 produced no negative effects on general movement or behavior across various doses.
Based on these initial safety profiles, the scientists evaluated how the drugs influenced fentanyl consumption. They surgically implanted intravenous catheters into adult male and female Wistar rats. The animals were placed in specialized chambers where they could press a lever to receive a small dose of fentanyl, a procedure known as intravenous self-administration.
After the rats learned the behavior and established a stable pattern of fentanyl intake over ten days, the researchers injected them with either the experimental drugs or a placebo. These tests involved groups of 12 to 13 rats per specific dose and drug combination. Initially, both drugs reduced fentanyl consumption, but GATC-021 lost its effectiveness over a five-day testing period.
GATC-1021, on the other hand, maintained its effectiveness and showed no signs of tolerance. Across doses ranging from 25 to 70 milligrams per kilogram of body weight, GATC-1021 consistently reduced the number of times the rats pressed the lever for fentanyl. When the scientists analyzed the data to account for individual variations, they found that GATC-1021 reduced fentanyl intake by more than 60 percent.
Because GATC-1021 targets the same serotonin receptors that are typically activated by psychedelic drugs, the researchers tested whether it might cause hallucinations. In rodents, hallucinogenic effects are measured by observing rapid, involuntary head twitches. While a known hallucinogenic drug caused significant head twitches, GATC-1021 did not trigger any such responses.
The scientists then examined how the drug affected the physical structure of brain cells, which communicate through tiny, branch-like protrusions called dendritic spines. The shape and density of these spines change as the brain learns and adapts, a process known as neuroplasticity. The animals that self-administered fentanyl and received GATC-1021 showed a higher percentage of adaptable, thin dendritic spines compared to rats that only consumed fentanyl.
To understand the genetic mechanisms behind these physical changes, the authors analyzed gene expression in three brain areas linked to reward and addiction. They found that fentanyl consumption alone caused widespread changes in gene activity, while treatment with GATC-1021 modified these patterns in a beneficial way. In the prefrontal cortex, a brain region responsible for decision-making, the drug significantly increased the activity of specific genes associated with neuroplasticity and brain cell survival.
While the integration of artificial intelligence accelerated the discovery process, the physical experiments revealed nuances that the computer models missed. For example, the artificial intelligence system predicted that pairing the experimental drugs with sulbutiamine, a synthetic form of vitamin B1, would enhance brain absorption. However, the animal tests showed that GATC-1021 was actually more effective at entering the brain without the addition of sulbutiamine.
“While the AI platform was essential for the initial development of the drug, preclinical testing in an animal model was equally essential to determine how the drug would act in a complex biological system,” Fowler said. “In these studies, we found that while the AI predictions were overall correct, some aspects of the prediction were not validated once we tested them in the rodent model. Thus, preclinical testing remains important so that we can prevent off-target effects from occurring in humans.”
Any predictions about how the drug will perform in humans remain speculative until clinical trials are conducted. “We are still in the early stages of developing this drug, and thus, it will likely be several years before all requirements are met with regard to the FDA regulatory pipeline,” Fowler explained.
Future research will need to explore how the drug is absorbed and processed across different areas of the brain under varying dosing schedules. Scientists will also need to investigate the long-term effects of the medication.
“We would like to see if this drug holds benefit for other disorders that are commonly found in [people] experiencing opioid use disorder, such as anxiety and depression,” Fowler added.
The study, “AI-derived therapeutic development of a serotonin receptor–targeting drug for the treatment of opioid use disorder,” was authored by Valeria Lallai, Samuel Kho, A. C. Martin, James P. Fowler, Madison L. Roach, Kevin Wang, Kendyl N. Laumann, Tyler G. Morrison, Mina Palaniappan, Malia Bautista, Allison S. Mogul, Jinjutha E. Cheepluesak, Bijay Shrestha, Dhanaji M. Lade, Julia E. Lagomarsino, Vaishnavi Narayan, Jayson Uffens, Waldemar Lernhardt, Saman Mirzaei, Ian Jenkins, Arturo R. Zavala, Jonathan R. T. Lakey, Robert Tinder, and Christie D. Fowler.
