Deciphering AMD by deep phenotyping and machine learning (360G-Wellcome-210572_Z_18_Z)

We will identify the structural changes leading to and associated with cell degeneration in the retina in patients with early age-related-macular-degeneration (AMD). This will pinpoint what makes AMD progress towards visual loss (late AMD). This will be achieved via machine learning, genotyping and high resolution phenotyping of early AMD patients using retrospective and prospective data. Patients will undergo state-of-the-art imaging of the major tissues; neurosensory retina, retinal pigment epithelium and choriocapillaris to identify the sequence of cell degeneration. We will control for genetic risk by genotyping to infer ancestry and to identify individuals with extreme polygenic risk scores. Thus accounting for confounder effects of cryptic genetic diversity. Machine learning will be utilized to generate a generalized model of AMD progression on a population basis, enabling us to assess individual deviation from normal ageing. Structural imaging biomarkers will be developed in already collected extensive imaging databases, and validated in a prospective cohort study. Biomarkers will: 1) detect conversion to late AMD earlier; 2) discriminate slow/fast progressors and 3) identify therapeutic targets. Results should improve clinical trial design by better characterizing study populations and result in novel therapies by identifying the underlying mechanisms of one of the largest unmet medical needs.