Researchers from the University of Virginia have pioneered a revolutionary machine-learning approach to discover drugs that potentially reduce harmful scarring after heart injuries such as heart attacks. This innovative approach utilizes both human knowledge and machine learning to understand the impact of drugs on fibroblasts, the cells responsible for heart repair following an injury.
A New Approach to Machine Learning
Fusing human knowledge with machine learning, the University of Virginia scientists have developed a unique approach to better comprehend how drugs affect fibroblasts. Fibroblasts are crucial not only for heart repair but are also the culprits in causing harmful scarring. The researchers combined a computer model with machine learning, creating a tool that has already located a promising candidate to help prevent damaging heart scarring. This work is supported by the National Institutes of Health and has been published in the renowned scientific journal PNAS.
Identifying Promising Drugs
The interdisciplinary team at the University of Virginia used machine learning to identify drugs with the potential to reduce harmful scarring following a heart attack. Their model singled out a promising candidate, pirfenidone, an FDA-approved drug for idiopathic pulmonary fibrosis. Additionally, the model unearthed a new explanation of how pirfenidone suppresses contractile fibres within fibroblasts. An experimental drug, WH4023, also showed potential in suppressing fibroblast contraction.
The Role of Machine Learning in Drug Discovery
Machine learning’s role in drug discovery could transform the pharmaceutical industry. By aiding in the early discovery of drugs, it can expedite the process of identifying candidate drugs, decrease costs, and improve accuracy. The researchers’ novel approach, termed ‘logic-based mechanistic machine learning,’ identified pirfenidone as a promising drug that suppresses contractile fibers in fibroblasts, thereby mitigating the risk of damaging scarring post heart injuries. This breakthrough aligns with developments at the University of Amsterdam, where the autonomous chemical synthesis robot, RoboChem, integrated with an AI-driven machine learning unit, has demonstrated superior performance to human chemists in terms of speed and accuracy.
Preventing Harmful Heart Scarring
University of Virginia researchers utilized the combination of human and computer learning to discover drugs that could potentially minimize harmful scarring after a heart attack. Their machine-learning tool identified a promising candidate to help prevent such scarring, something previous drugs failed to do. The approach involved studying the impact of 13 drugs on human fibroblasts and using the data to train the machine learning model to predict the drugs’ effects on the cells and cellular behavior.
Advancing Heart Fibrosis Treatment
The University of Virginia researchers’ new approach, logic-based mechanistic machine learning, has identified pirfenidone as a promising drug in the fight against harmful scarring post-heart attack. Pirfenidone suppresses contractile fibers within a fibroblast that stiffen the heart. The team believes this approach could be a powerful tool in understanding biological cause-and-effect and in developing targeted treatments for heart fibrosis.
With its potential to advance the development of new treatments for diverse diseases, not just for heart injury, this research holds transformative potential. The use of machine learning in drug discovery could revolutionize the pharmaceutical industry, leading to a new era of accelerated drug discovery and precision medicine.
