BADGERS Workflow.

BADGERS takes (a) Alzheimer’s disease GWAS, (b) LD reference panel, and (c) Human traits GWAS from the UK biobank as input. The generated result will be the (d) Association between Alzheimer’s disease and human traits. In graph (d), each triangle represents one human trait, and different colors represent different trait categories.

Simulation results.

BADGERS and regression analysis based on individual-level data showed (A) highly consistent effect size estimates for 1,738 PRS in simulation and (B) comparable statistical power (setting 3).

PRS-based BWAS identifies risk factors for AD.

Meta-analysis p-values for 1,738 heritable traits in the UK biobank are shown in the figure. P-values are truncated at 1e-15 for visualization purposes. The horizontal line marks the Bonferroni-corrected significance threshold (i.e. p=0.05/1738). Positive associations point upward, and negative associations point downward.

PRS correlation matrix for the 48 traits identified in marginal association analysis.

Trait categories and association directions with AD are annotated. The dendrogram indicates the results of hierarchical clustering. We used 1000 Genomes samples with European ancestry to calculate PRS and evaluate their correlations. Label “irnt” means that trait values were standardized using rank-based inverse normal transformation in the GWAS analysis.

Influence of the APOE region on trait-AD associations.

The horizontal and vertical axes denote association p-values before and after removal of the APOE region, respectively. Original p-values (i.e. the x-axis) were truncated at 1e-20 for visualization purposes.

Associations between identified AD risk factors and various AD subgroups, CSF biomarkers, and neuropathologic features.

Asterisks denote significant associations based on an FDR cutoff of 0.05. P-values are truncated at 1e-5 for visualization purposes.

Associations between six traits and pre-clinical cognitive phenotypes in WRAP.

Error bars denote the standard error of effect estimates.