Pioneering the Future of Non-Addictive Pain Management with Deep Learning

A new AI model offers a faster and more effective approach to drug discovery and repurposing, facilitating the screening of potential drug candidates for non-addictive pain management.  

The findings, published in Cell Press, represent one of the many ways Cleveland Clinic (OH, U.S.) and IBM (NY, U.S.) are partnering together in this field.

The physical and emotional toll of chronic pain is immense. It can often lead to increased rates of anxiety, depression, suicidal thoughts and a reduced quality of life. Pain affects approximately 50 million adults in the U.S. and 20% of the population worldwide. Opioids are commonly used as first line treatment for pain, however, their addictive nature remains problematic.

Now, Cleveland Clinic’s Genome Centre and IBM, are using AI for drug discovery and repurposing in advanced non-addictive pain management.

LISA-CPI: How Does it Work?

Ligand Image-and Structure Aware Compound-Protein-Interaction (LISA-CPI) is an advanced deep learning framework designed to predict interactions between drug compounds and their protein targets. The Cell Press study reported that LISA-CPI successfully analyzed and predicted how 369 gut microbial metabolites and 2,308 FDA-approved drugs would interact with 13 pain receptors.

LISA-CPI combines two advanced computational techniques: ImageMOI, which uses deep learning to extract structural and chemical information from molecular images of ligands; and the Evoformer algorithm, which generates accurate 3D protein structures. Together, they model compound-protein interactions with remarkable precision.

In terms of developing treatments for pain management, LISA-CPI can predict whether a molecule can bind to specific pain receptors. It can then determine the strength of this binding, which allows researchers to deduce whether the binding activates or deactivates pain signals.

To determine whether a molecule could function as a drug or be repurposed, a 3D model of its structure is necessary, built from comprehensive 2D data on its physical, structural, and chemical properties. The study highlights that LISA-CPI identified several drugs that can be repurposed, such as methylergometrine, however, further studies are needed to confirm this.

Dr Yuxin Yang, the first author of the Cell Press paper explained:

“This algorithm’s predictions can lessen the experimental burden researchers must overcome to even come up with a list of candidate drugs for further testing. We can use this tool to test even more drugs, metabolites, GPCRs and other receptors to find therapeutics that treat diseases beyond pain, like Alzheimer’s disease.”

AI can Revolutionize Drug Repurposing

The process of developing new drugs for chronic pain is expensive and time-consuming. The traditional drug discovery procedures can take over a decade and often lead to high failure rates with huge losses due to the high cost associated with it. LISA-CPI offers researchers a tool to test these FDA-approved drugs to be repurposed in a time-efficient manner.

Recent evidence showed that certain receptors within a group known as G-protein coupled receptors (GPCRs) are involved in pain perception, and that these can be drugged in a non-addictive manner. LISA-CPI can help in this endeavour by identifying molecules that bind to GPCRs and elicit pain relief responses. Dr Feixiong Cheng, Director of Cleveland Clinic’s Genome Center and IBM commented:

"AI can rapidly make full use of both compound and protein data gained from imaging, evolutionary and chemical experiments to predict which compound has the best chance of influencing our pain receptors in the right way."

The Cell Press study highlights that LISA-CPI demonstrated 20% greater accuracy than previous state-of-the-art models in predicting how molecules interact with pain-related GPCRs. This could revolutionize the entire drug discovery timeline as it could speed up the process, decrease the cost of finding new drug candidates, and allow for more drug candidates to be approved and available to patients at a more affordable price.

Challenges and Future Directions

The current LISA-CPI framework has some limitations. Firstly, it encodes only 2D molecular images, omitting any 3D spatial information about atomic positions. Additionally, the model relies on single protein representations from AlphaFold2's Evoformer, rather than 3D structures of GPCR targets or ligand-receptor complexes. A potential solution is to integrate 3D structural data for both ligands and receptors using graph-based or 3D mesh representations of molecules and proteins.

The recent release of AlphaFold 3 in biomolecular interaction prediction holds promise for improving GPCR model accuracy. The researchers stated that they believe integrating AlphaFold 3 could potentially elevate the performance of LISA-CPI.

Dr Cheng, head of the lab behind the Cell Press paper, concluded by saying:

"We believe that these foundation models will offer powerful AI technologies to rapidly develop therapeutics for multiple challenging human health issues.”

With deep learning frameworks like LISA-CPI, researchers can now predict drug-receptor interactions with unprecedented speed and accuracy. For those suffering from chronic pain, this development could be life-changing. It also represents a broader transformation for the use of AI in drug discovery and repurposing, possibly allowing for effective and non-addictive pain management that is accessible worldwide.