Mechanistic association of epigenetic bone age and frailty measures with bone imaging studies using novel artificial intelligence approaches

Vanderbilt University Medical Center, Nashville TN

Investigators

Abstract

Abstract Aging is the biggest risk factor for the development of chronic disease and disability, including development of osteoporosis (OP) and osteopenia. These conditions affect more than 60 million Americans and are associated with great morbidity and mortality in the setting of fragility fractures. The number of people affected by these conditions will only increase as the population ages. Screening for OP and osteopenia is limited, facilitating intervention only in more advanced stages of bone loss. Limited studies have identified the potential of artificial intelligence (AI) and deep learning (DL) tools, as applied to opportunistically-acquired imaging, for the early identification of changes in bone structure that may herald the onset of OP. In addition, the growing field of epigenetics has facilitated assessment of aging and its comorbidities at a molecular and genetic level. Although one study suggested that bone mineral density (BMD)-associated loci discovered by genome-wide association studies could only explain up to 6% of BMD variation (Nat Genet 2012; 44:491-501), recent research has shown that methylation levels at five CpGs that differ significantly between healthy and osteoporotic women could explain 14% of BMD variation (Epigenetics 2017;12:674-687). Despite these promising findings, there is a lack of understanding of the connection between epigenetic modification of genes and imaged bone findings. This research area represents a crucial gap in existing knowledge that this project seeks to fill. This proposal aims to carry out a discovery-focused effort to identify epigenetic factors that cause early OP associated with imaging findings on CTs. The first aim of the study is to refine and to validate, in an American sample, DL approaches to measure BMD from opportunistically-acquired CT scans with a corresponding comprehensive electronic health record. The second and third aims are to identify quantitative markers of OP in CT scans that are associated with DNA methylation of genes implicated in OP, as assessed in blood (Aim 2) and bone (Aim 3) samples. Using these data, frailty and resilience will be quantified with reference to chronological and epigenetic age, through alignment of imaging-based quantitative metrics and molecular data. Knowledge and skills acquired and developed as part of this career award will involve AI/DL technologies, geroscience, epigenetic assessment, and advanced bioinformatics. Ultimately, this study will facilitate identification of early evidence of bony changes in younger patients, predating a disease state, which will allow for earlier intervention and identify targets for pharmacological intervention based on epigenetic patterns. Notably, a 13-year study recently demonstrated that reversal of epigenetic modification at set loci can reverse cellular evidence of aging (Cell 2023;186:305-326.e27). Building on this work, I seek to identify evidence of OP through DL/AI-mediated imaging findings and through epigenetic assessment at a time point when early intervention on factors affecting epigenetic pathways could halt, slow, or reverse progression of OP.