Genetic Variation in AMD Progression and Treatment Response - Summary/Abstract This project aims to explore how genetic variation influences disease progression and therapeutic response in age-related macular degeneration (AMD) by integrating advanced optical coherence tomography (OCT) imaging with machine learning (ML) and artificial intelligence (AI). AMD, a leading cause of vision loss in older adults, presents with highly variable progression, ranging from slow decline to rapid vision loss. This variability remains poorly understood, largely due to the genetic and phenotypic diversity of AMD. This project leverages ML for deep phenotyping of OCT data to refine the classification of AMD subtypes, combined with genetic analysis for a deeper understanding of disease progression. The hypothesis is that genetic variations influence distinct AMD subtypes and stages, shaping both disease progression and therapeutic response. To test this hypothesis, the following aims are proposed: Specific Aim 1: Determine how genetic variation influences AMD progression using AI-driven analysis of OCT biomarkers and ML to classify patients as slow or rapid progressors. Three-dimensional ML models, such as SLIViT and retina-specific models like RETFound, will enable detailed phenotype analysis, revealing high-risk features and novel subtypes correlated with progression rates. Specific Aim 2: Investigate how genetic variation affects functional and therapeutic outcomes in AMD by integrating OCT data with patient treatment responses through ML-based models, exploring genetic factors linked to visual acuity outcomes and treatment efficacy. The project will refine polygenic risk scores (PRS) by combining these insights with genome-wide association study (GWAS) data to improve predictions of rapid progression and treatment responsiveness. Datasets from the UK Biobank and UCLA Biobank will be utilized, applying ML-based imaging analysis, transfer learning for 3D data, ML-based deep phenotyping, and traditional and post-GWAS analysis. Techniques include ML-based progression analysis (e.g., pySuStain) and causal ML for treatment response. This research aims to identify novel genetic loci associated with AMD subtypes, offering new insights into disease mechanisms and potentially unrecognized pathways. These findings will enhance PRS models, enabling better stratification of patients by genetic risk and advancing personalized approaches to AMD management. Supported by a team of mentors with expertise in retinal imaging, genetics, and bioinformatics, this project seeks to impact public health by guiding diagnosis, therapeutic, and prognosis to reduce vision loss in patients with AMD.
