Integrating deep learning, magnetic resonance imaging and genomics to study myocardial fibrosis

/Abstract Myocardial fibrosis is commonplace in the aging heart. Accelerated extracellular matrix remodeling and collagen deposition, triggered by chronic myocardial tissue injury, are associated with pathologic myocardial fibrosis, and contribute to the rising burden of cardiovascular disease. Although the initiating triggers for myocardial tissue injury are disease-specific, they activate a highly consistent set of core pathways of fibrosis that lead to accumulation of myofibroblasts and fibrotic tissue remodeling. Unraveling the mechanisms that contribute to myocardial fibrosis will enable prevention of myocardial fibrosis progression and design of therapeutic interventions that could have a wide-ranging impact on multiple cardiovascular diseases. Cardiac magnetic resonance imaging (MRI) with T1 mapping is a non-invasive histologically validated quantitative method for assessment of myocardial fibrosis. Genome-wide association analysis is a powerful analytical approach that can identify biological pathways linked to cardiovascular traits. However, adequately powered genome-wide association studies of myocardial fibrosis in humans require large sample sizes with combined genomic, cardiac MRI and clinical outcome data, which to date have been elusive. The UK Biobank (UKB) is a unique longitudinal prospective cohort comprising ~500,000 participants with comprehensive phenotyping, imaging, and multiple genomic data types. Approximately 100,000 participants have contributed to the UKB imaging study, offering paired cardiac MRI and whole genome sequence data, with 60,000 participants undergoing repeat cardiac MRI during follow-up. The current proposal aims to: (1) harness machine learning to measure, at an unprecedented scale in 100,000 individuals in the UKB, myocardial segment-specific fibrosis burden and tissue heterogeneity and examine the clinical relevance of these measured traits by examining their association with cardiovascular, metabolic and inflammatory diseases; (2) leverage whole genome sequence data to identify coding and non-coding genomic loci associated with myocardial fibrosis for potential therapeutic targeting; and (3) examine clinical and genetic factors that contribute to myocardial fibrosis progression in the subset with repeat cardiac MRI. This work will benefit from the robust scientific and computational infrastructure at the Cardiovascular Medicine Division at the Brigham and Women's Hospital and the Cardiovascular Disease Initiative (CVDI) at the Broad Institute of Massachusetts Institute of Technology (MIT) and Harvard. Dr. Nauffal will be mentored by Dr. Patrick T. Ellinor, the Director of the CVDI and an expert in defining the molecular basis of cardiac fibrosis using genetics. By completing the research and training aims outlined in this proposal, Dr. Nauffal will acquire proficiency in deep learning methods, human genetic discovery, fibrosis imaging and longitudinal data analysis. These skills are critical to achieve his goal of becoming an independent R01-funded computational cardiovascular geneticist. Project Number: 1K08HL179484-01 | Fiscal Year: 2025 | NIH Institute/Center: National Heart Lung and Blood Institute (NHLBI) | Principal Investigator: Victor Nauffal | Institution: BRIGHAM AND WOMEN'S HOSPITAL, BOSTON, MA | Award Amount: $166,240 | Activity Code: K08 | Study Section: NHLBI Mentored Clinical and Basic Science Study Section[MCBS (JA)] View on NIH RePORTER: https://reporter.nih.gov/project-details/1K08HL17948401