Fig. 4: Analysis of the genetic architecture of phenotypes imputed with AutoComplete. | Nature Genetics

Fig. 4: Analysis of the genetic architecture of phenotypes imputed with AutoComplete.

From: Deep learning-based phenotype imputation on population-scale biobank data increases genetic discoveries

Fig. 4

a, Effect sizes of significantly associated loci based on imputed phenotypes were examined in the association studies of the observed phenotypes in UKBB (Obs or observed) and comparable nonUKBB studies (Ext or external). Genome-wide analysis was performed across 5,776,313 SNPs. For imputed phenotypes, circles indicate the mean effect based on multiple imputation (black bars indicate the 95% CI). Mismatches in effect directions are highlighted in orange. Effects that were not significantly different from zero in Obs or Ext (at P < 0.05, two-sided t-test) are denoted using empty markers. Loci are visualized that were present across compared studies for each phenotype. b, Genetic correlation (\({r}_{G}\)) for bilirubin, LifetimeMDD and cannabis ever taken between UKBB observed and imputed (Obs and Imp, in blue) phenotypes and nonUKBB cohorts (Ext, in orange and purple). Bar heights for genetic correlations that involve imputed phenotypes indicate mean \({r}_{G}\) based on multiple imputation. Black bars indicate the 95% CI.

Back to article page