Logistic mixed effect models (logME) can be used to test for associations between binary traits and genetic variants while accounting for population structure and the relationships among individuals using a random effect defined by an empirical genetic relationship matrix (GRM). However, the optimal choice of variants to include when generating the GRM is not well understood. Here, we investigate the performance of logME models under a range of GRM choices using meta and joint analysis when the study sample includes samples from two distinct ancestries.
Genotypes from 4177 adults of European ancestry (Framingham Heart Study, FHS), and 3417 adults of African American ancestry (Jackson Heart study, JHS) were measured from the Trans-Omics for Precision Medicine program freeze4 whole genome deep sequencing data. We computed GRMs in FHS and JHS and in the combined FHS+JHS sample using 1) MAF≥0.1% and 2) MAF≥5% variants after linkage-disequilibrium pruning. A binary trait with 50% heritability and a difference in prevalence of 10% versus 5% in the two samples was simulated based on a mix of variants with large and small differences in frequency between the FHS and JHS samples. On average, power was higher for combined than for meta-analysis. The ≥5% MAF and ≥0.1% MAF GRMs produced equivalent type I error and power with meta-analysis, but for combined analysis the power was lower for the ≥0.1% MAF GRM than for the ≥5% MAF GRM. We recommend combined analysis and GRMs based on common variants for logME association analyses with binary phenotypes.

Background: Genome-wide association studies (GWAS) have identified 14 loci associated with atrial fibrillation (AF). Whole
genome sequencing (WGS) may enable the identification of causal variation underlying AF through greater variant ascertainment
than with genotyping arrays. We report interim results for a subset of the NHLBI Trans-Omics for Precision Medicine WGS
Program, in which ~3000 individuals with early-onset AF (onset ≤65 years of age) and 4000 referent individuals will be
sequenced. Methods: In the current freeze, 1423 individuals with early-onset AF from 9 studies and 1431 referent individuals from
the Framingham Heart Study underwent WGS at the Broad Institute (Cambridge, MA) and passed quality control filters. Variants
were jointly called at the University of Michigan. European ancestry was verified using principal components. Common variants
(≥5%) were tested for association with AF after age- and sex-adjustment. Low-frequency and rare variants were grouped in
regions of interest and tested for association with AF. Results: AF cases were younger than referents (53±14 vs. 59±15, p<0.001)
and a greater proportion were male (68% vs 43%, p<0.001). The mean sequencing read depth ranged from 34.5x-38.9x across
the different studies. In total, 55,106,124 variants were included in association analyses, a substantially greater number than in
studies predating WGS (e.g., ~2.2 million in a prior HapMap GWAS). Most variants were rare (85% with an allele frequency <1%).
We observed expected associations between variants at established AF loci from prior GWAS. For example, variants at the top
AF susceptibility locus on chromosome 4q25 were highly associated with AF (rs2220427_T, OR 1.68, 95%CI 1.46-1.95, p=2x10-12).
Ongoing group-based tests will enable the assessment of variation in individual genes and noncoding regions for an association
with AF. Conclusions: Our results demonstrate the feasibility of performing large-scale WGS for common diseases. WGS enables
greater interrogation of genetic variation as compared to GWAS. Future analyses will include larger study samples, assessment of
structural genomic variation, and integration with high-throughput assays to discover functional elements underlying AF
pathogenesis.