PQG Working Group

Caitlin CareyPost-Doctoral FellowAnalytical and Translational Genetics Unit, Mass General Hospital and Broad InstituteGenetic architecture of phenome-wide latent factors in the UK BiobankLarge-scale biobanks and electronic health records can offer unprecedented insight into the genetic architecture of a wide range of traits in the general population. However, the high dimensionality of phenotypic data, with thousands of traits measured per individual, can make it difficult to interpret genetic results. Here we derive and genetically characterize latent phenotypic factors representative of the spectrum of traits measured in UK Biobank (UKB). Beginning with 4203 traits, we performed exploratory factor analysis in a core dataset of 33,860 individuals and 730 items with low overall missingness. The final model includes 36 stable, interpretable factors accounting for ~30% of total phenotypic variance and spanning a wide range of physical (e.g., body size, general pain), behavioral (e.g., neuroticism, smoking), and lifestyle (e.g., education, urbanicity) dimensions. To investigate the genetic architecture of these latent factors, we performed a genome-wide association study (GWAS) of each factor in the full European ancestry subset (N=361,144). Using LD score regression, we observe a roughly twofold increase in median common variant heritability for the derived factors compared to that of the individual items. We contextualize these factors by estimating genetic correlation to psychiatric and somatic disorders studied outside of UKB. In preliminary results, genetic correlations confirm expected associations (e.g., rg=0.58 for coronary artery disease and the heart attack factor, rg=-0.61 for ADHD and the education factor) and also reveal a surprisingly broad, diffuse pattern of genetic correlation between common psychiatric disorders (i.e., ADHD and major depressive disorder) and phenotypic factors.Overall, these results suggest that phenotypic factor analysis enhances interpretability, boosts power for heritability analyses, and can yield meaningful reduction in dimensionality to drive the next generation of genetic studies. Event Url https://ems.sph.harvard.edu/MasterCalendar/EventDetails.aspx?data=hHr80o3M7J4zS74Y6VHRjb7AkaaT8B%2b9X3hWFuirZrKtLmS4MtEywjVwXEdisn9t