Using genetic associations to map disease substructure from UK biobank phenotype data (360G-Wellcome-215097_Z_18_Z)

Health resources such as the UK biobank (UKBB) offer a breadth of phenotypical data that allow disease heterogeneity and comorbidity to be explored, although these are subject to clinical practice preferences and center-specific biases, and relatively imprecise. Crucially, gold standard diagnoses for many diseases are lacking in the UKBB, making it impossible to confidently assess how these diseases are associated with other healthcare data, and as such, to reveal their substructure. We therefore intend to infer gold standard disease diagnoses from other phenotypic data in the UKBB, using established genetic associations from cleaner data as a guide for the quality of our inference. Once more accurate diagnoses are attained, we intend to explore disease substructure by identifying which clinical features are associated with multiple diseases simultaneously. Working towards an improved understanding of disease substructure will ultimately help target both existing and new treatments to the forms of disease in which they are most likely to be effective.