We fitted ancestry-aware mixed models for 12 cohorts with ancestrally diverse and admixed participants. These were conducted using GENESIS Bioconductor package in R, which was developed for large-scale genetic analyses in samples with complex structure including relatedness, population structure and ancestry admixture.60 Genotyped variants for each study were first pruned, and the KING-robust method was used to estimate relatedness in the first instance. Subsequently, PC-Air was employed to calculate PCs using the kinship matrix derived from KING-robust method and the pruned variants. PC-Relate was used to re-estimate relatedness utilizing PCs from PC-Air. To enhance precision, a second iteration of PC-Air and PC-Relate was performed. Afterwards, we fitted a null model for MD case-control status, using sex, age, all 32 PCs from PC-Air, and the kinship matrix from PC-Relate. Finally, score tests were conducted using the null model and all imputed variants as predictors. Due to computational limitations, the Million Veterans Program was partitioned into 19 batches, which were then combined using an inverse variance weighted meta-analysis, implemented in METAL. To derive an estimate of the odds ratio (OR) and its
