A Rutgers researcher has teamed up with a local high school student to develop a method using machine learning to screen current and existing drugs to determine which could potentially become effective treatments for addiction-related disorders.
Morgan James, assistant professor of psychiatry at Robert Wood Johnson Medical School, core member of the Rutgers Brain Health Institute, and member of the executive board at the Rutgers Addiction Research Center, has been working on developing better treatments for psychiatric illnesses such as eating disorders, substance use disorders, and depression.
Instead of attempting to create an entirely new drug, James took a novel approach; he decided to focus on existing medications to determine if their off-target effects include blocking the orexin 1 receptor, which regulates various physiological phenomena such as wakefulness, feeding and the brain's reaction to rewards.
He received help from Vanessa Zhang, a student from West Windsor-Plainsboro High School South and champion figure skater with an interest in computer programming who has witnessed the toll eating disorders has taken among her peers. Together they developed method using artificial intelligence to screen all existing drugs to determine if they may have so-called off-target effects at the orexin 1 receptor.
"Vanessa made contact with me through William Welsh, is a professor in the department of pharmacology and a long-time collaborator of mine," said James. "Vanessa is a very accomplished figure skater so working on this project was a way to combining two of her interests, machine learning and combatting eating disorders. She took on the project of creating the algorithm, with guidance from myself and Professor Welsh, and has done an amazing job."
Zhang explained what inspired and motivated her to take part in the project, including the use of computational methods.
"Eating disorders are a prevalent issue in figure skating," said Zhang. "As a competitive figure skater, I've witnessed the detrimental impact these mental struggles have on skaters' health and careers. I wanted to make a change in the skating community, and this passion inspired me to conduct scientific research with the hopes of finding new eating disorder treatments."
She added, "Professor James' project on the orexin 1 receptor especially appealed to me because this receptor represents a new approach to combating eating disorders. Furthermore, I was very excited to use computational methods while researching, and Professor James was very open to this approach."
Although there are no clinically-available drugs that act on orexin 1 directly, there are experimental compounds that do. James and his team collected those compounds into a database to better understand why and how they impact the orexin 1 receptor. That information was then used to create an algorithm that predicts the likelihood of another drug being able to do the same or similar.
"We developed a machine-learning program to determine what features predicted the likelihood of a drug acting at the orexin 1 receptor, and the algorithm did a really good job of doing that," said James. "We then looked at 1,600 compounds that are FDA approved, so theoretically available to give to humans, and that produced a short list of candidate drugs. We carried those out in vitro screening, essentially testing each drug in a dish to see if it acts at the orexin 1 receptor. We found several of our candidate drugs had some of the features we are looking for."
"My favorite part of this project was the process of developing the machine learning model," said Zhang. "Although it was a learning experience and full of trials and errors, each failure motivated me to learn more methods and refine my approach. Ultimately, I learned how to use a diverse array of computational tools to create a model that could contribute to the project's success."
Satisfied with the knowledge that the process worked, James and his team turned their patented AI method to a database of almost 20 million chemical entities, and then focused on the top 100 results to test further in vitro.
The AI platform as well as certain candidate drugs that were discovered during the process have been patented through the Office for Research's technology transfer department, and James and his team will continue their medicinal chemistry study to find drugs that have high promise on their own and could potentially be modified to become viable medications focusing on the orexin system.
"The Office for Research's technology transfer team have been great - they always are," said James. "We went to them pretty early on, once we'd found data, and worked with them to discuss if there was anything patentable there, and they helped us delineate two patentable things: the platform, the AI approach to screening drugs, as well as some of the compounds that popped out in our initial screening that could potentially be repurposed to exploit their actions at the orexin 1 receptor. They helped us draft and file the provisional application and have been supportive all the way."
"Working with Dr. James goes hand-in-hand with the Office for Research's mission to support the research, scholarship, and creative endeavors of all Rutgers faculty," said Deborah Perez Fernandez, executive director of Technology Transfer. "His research into addiction disorders could become the key in addressing an issue that impacts so many people and families around the world."
