UVA Researchers Discover Scar-Reducing Drug for Post-Injury Treatment

Scientists at the University of Virginia (UVA) have developed a new approach to machine learning, a form of artificial intelligence, to identify drugs that can help reduce the severity of post-traumatic stress disorder, heart or other injuries.

 A new machine learning tool has identified drug candidates that may prevent heart damage in a different way than traditional drugs. According to UVA researchers, their latest computer model can predict and explain the effects of drugs on other diseases.

"Many common diseases, such as heart disease, metabolic syndrome and cancer, are complex and difficult to treat," said researcher Anders R. Nelson, a computational biologist and alumnus of UVA's Jeffrey J. Lab. Dr. Saucerman. "Machine learning will help reduce this complexity, identify the most important factors that affect disease, and better understand how drugs affect diseased cells."

"Machine learning can help identify cellular properties created by drugs," said Saucerman of the Department of Biomedical Engineering, a joint program of UVA's School of Medicine and Engineering. "The combination of machine learning and human learning helped us predict drugs for fibrosis and explain how the drugs work. This knowledge is needed to design clinical trials and identify side effects."

Saucerman and his team combined machine learning and computer models based on decades of human knowledge to better understand how drugs affect fibroblast cells. These cells help repair the heart after injury by producing collagen and reducing scarring. However, an ugly scar called fibrosis also occurs as part of the repair process. Saucerman and his team wanted to see if choosing an effective drug could increase doctors' ability to prevent scarring and improve patient outcomes.

Previous attempts to identify drugs that target fibroblasts have focused only on selected aspects of fibroblast behavior, and the mode of action of these drugs is unclear. This knowledge gap presents a significant challenge in developing treatments for cardiac fibrosis. So Saucerman and his colleagues developed a new approach called "logic-based machine learning," which predicts drugs and how they affect fibroblasts' behavior.