Summary: Researchers are using artificial intelligence to design novel compounds that can block kappa opioid receptors, aiming to develop new treatments to reduce opioid dependence and withdrawal-driven relapse.
Source: Biophysical Society
Approximately three million Americans suffer from opioid use disorder, and more than 80,000 die each year from overdoses.
Opioids such as heroin, fentanyl, oxycodone and morphine exert their effects by binding to opioid receptors in the brain. Activation of mu-opioid receptors produces pain relief and euphoria but can also cause physical dependence and dangerous respiratory depression that leads to fatal overdoses. By contrast, kappa-opioid receptors modulate different aspects of reward and stress-related behaviors, and blocking these receptors has emerged in preclinical studies as a promising strategy to reduce drug-seeking and withdrawal-driven relapse.
Leslie Salas Estrada, a postdoctoral researcher in Marta Filizola’s lab at the Icahn School of Medicine at Mount Sinai, is applying artificial intelligence to accelerate the discovery of molecules that inhibit the kappa-opioid receptor. Her work, presented at the 67th Annual Biophysical Society Meeting in San Diego, focuses on designing novel chemical compounds that could blunt the neurobiological drivers of opioid dependence.
“If you are addicted and trying to quit, withdrawal symptoms can be extremely difficult to overcome,” Salas Estrada explained. “After prolonged opioid exposure, brain circuits become altered in ways that increase craving and drug seeking. In animal models, blocking kappa-opioid receptors can reduce that heightened drive during withdrawal. Our goal is to find compounds that selectively target that receptor and could one day help people through recovery.”
Discovering drugs that block a specific protein such as the kappa-opioid receptor has traditionally required lengthy and costly screening of large compound libraries. Computational screening can speed this process, but brute-force approaches that evaluate billions of molecules still take considerable time and resources. To overcome these limitations, Salas Estrada and colleagues turned to machine learning techniques that can learn chemical patterns and propose new candidate molecules more efficiently.
Their approach trains a generative computer model on available structural data for the kappa-opioid receptor and on known bioactive molecules. The system uses a reinforcement learning framework: the model proposes new chemical structures and receives feedback based on properties that predict favorable drug behavior. Those properties can include predicted binding to the target receptor, molecular features associated with stability and permeability, and other attributes that influence whether a compound is likely to succeed as a therapeutic.
“Artificial intelligence excels at extracting patterns from vast datasets,” Salas Estrada said. “By training models on chemical databases and receptor information, we can design novel molecules from scratch rather than relying solely on existing libraries. This strategy has the potential to shorten discovery timelines and reduce costs.”
Using this pipeline, the team has already generated several candidate compounds with promising predicted profiles. The next steps involve synthesizing these molecules with collaborating chemists and testing them in cell-based assays to confirm their ability to block the kappa-opioid receptor. Compounds that demonstrate target engagement and acceptable safety profiles in vitro will advance to preclinical animal studies to evaluate efficacy and tolerability.
Salas Estrada emphasized the iterative nature of the work: as experimental data are collected, those results feed back into the computational models to refine future designs. “The cycle of design, synthesis, testing and model updating is central to making progress,” she said. “Our ultimate hope is that AI-designed compounds will lead to new medications that help people struggling with addiction.”
About this AI and opioid addiction research news
Author: Leann Fox
Source: Biophysical Society
Contact: Leann Fox – Biophysical Society
Image: The image is in the public domain
Original Research: The findings were presented at the 67th Annual Biophysical Society Meeting in San Diego, California.