About 20% of Americans suffer from chronic pain, and existing treatment methods often fall short. Dr. Feixiong Cheng, Director of the Genome Center at Cleveland Clinic, along with IBM, is leveraging artificial intelligence (AI) to discover new drugs aimed at better managing pain. Their deep-learning approach has uncovered various metabolites derived from the gut microbiome and identified FDA-approved medications that can be repurposed as non-addictive, non-opioid treatments for chronic pain.
The results have been published in Cell Reports Methods.
Dr. Yunguang Qiu, a co-first author and postdoctoral fellow in Dr. Cheng’s lab dedicated to nervous system therapeutics, highlights the ongoing challenge of using opioids for pain management due to the risk of serious side effects and addiction. Recent studies indicate that targeting a particular group of pain receptors within a class of proteins known as G protein-coupled receptors (GPCRs) could lead to non-addictive pain relief. The real challenge lies in effectively targeting these receptors, explains Dr. Qiu.
Rather than creating new molecules from the ground up, the researchers considered applying existing methods to search for previously approved FDA drugs that might also address pain. This research involves mapping gut metabolites to identify viable drug targets.
The team, led by Dr. Yuxin Yang, a computational scientist and one of the prominent authors, worked to refine an earlier AI-based drug discovery algorithm developed by the Cheng Lab. The collaboration with IBM also contributed to writing and refining the research publication.
Dr. Yang expresses gratitude for the insights and advanced computational techniques shared by IBM collaborators: “It’s been a great opportunity to collaborate and learn from industry peers.”
To assess whether a molecule can function as a drug, the researchers aim to predict its interaction with and impact on proteins, specifically pain receptors. This necessitates a comprehensive 3D understanding of both the molecules involved, grounded in extensive 2D data regarding their chemical, structural, and physical characteristics.
Dr. Cheng notes that synthesizing the vast amounts of data necessary for predictive analysis using current computational methods is both complex and labor-intensive. “AI accelerates this process by effectively utilizing both compound and protein data obtained from imaging, evolutionary studies, and chemical experiments to forecast which compounds are likely to interact favorably with our pain receptors,” he explains.
The researchers developed a tool called LISA-CPI (Ligand Image- and Receptor’s Three-Dimensional (3D) Structures-Aware framework to predict Compound-Protein Interactions). This tool employs a type of artificial intelligence known as deep learning to forecast:
- Whether a molecule can bind to a particular pain receptor
- The specific location where a molecule will attach to the receptor
- The strength of the molecule’s attachment to the receptor
- Whether the binding of a molecule will activate or deactivate signaling pathways
Using LISA-CPI, the team predicted interactions between 369 gut microbial metabolites and 2,308 FDA-approved drugs with 13 pain-related receptors, leading to the identification of potential compounds for repurposing in pain management. Laboratory studies to validate these compounds are currently in progress.
Dr. Yang emphasizes that the predictions made by this algorithm can significantly reduce the amount of experimental work needed to compile lists of potential drugs for further investigation. “This tool enables us to evaluate a broader range of drugs, metabolites, GPCRs, and other receptors to find treatments for ailments beyond pain, such as Alzheimer’s disease.”
Dr. Cheng noted that this initiative represents just one aspect of their collaboration with IBM to create foundational models for drug development, including drug repurposing in this instance and a continuing novel drug discovery endeavor.
“We are confident that these foundational models will empower advanced AI technologies for the rapid development of therapies targeting various pressing health challenges,” he states.
