Gil McVean, PhD, FRS, FMedSci
Professor of Statistical Genetics
Director of the Big Data Institute
University of Oxford
Using genetics to classify common human disease
Disease classification is fundamental to clinical practice, but current taxonomies do not necessarily reflect the pathophysiological processes that are common or unique to different disorders, such as those determined by genetic risk factors. Genome-wide association studies of risk for common medical conditions have revealed widespread connections between diseases at the molecular level. To date, however, it has not been possible to integrate and interrogate information from the full range of clinical phenotypes, as GWAS have focused on a relatively small number of traits and have often studied patients with only the most clear-cut diagnoses and uniform clinical manifestations. The availability of population-based cohorts, such as the UK Biobank, provides a unique opportunity to take a disease-agnostic perspective to investigate cross-trait genetic associations across a heterogeneous patient population. We have used routine healthcare data from the 500,000 participants in the UK Biobank to map genome-wide associations across 19,628 diagnostic terms, finding several thousand independent genetic risk loci affect multiple clinical phenotypes, which we cluster into >500 distinct disease association profiles. We use multiple approaches to link clusters to different underlying biological pathways and show how these clusters define the genetic architecture of common medical conditions, including hypertension and immune-mediated diseases. I will discuss clusters can be used to re-define disease relationships and to inform therapeutic strategies.
