The Minnesota Center for Twin and Family Research genome-wide association study

Abstract

As part of the Genes, Environment and Development Initiative, the Minnesota Center for Twin and Family Research (MCTFR) undertook a genome-wide association study, which we describe here. A total of 8,405 research participants, clustered in four-member families, have been successfully genotyped on 527,829 single nucleotide polymorphism (SNP) markers using lllumina's Human660W-Ouad array. Quality control screening of samples and markers as well as SNP imputation procedures are described. We also describe methods for ancestry control and how the familial clustering of the MCTFR sample can be accounted for in the analysis using a Rapid Feasible Generalized Least Squares algorithm. The rich longitudinal MCTFR assessments provide numerous opportunities for collaboration.

Figure 1

Two histograms of Mahalanobis distances (D²s) to the centroid of the European-American (“White”) group. Both histograms display the same data, but with different axis ranges and bin widths. Histogram A uses bins of width 42 so that the first two bars on the left include all 6,615 White subjects and no others. The subjects with D² near 1000 are mostly mixed-race, Hispanic and Native American. The peak between 3000 and 4000 was caused by 378 Asian subjects, mostly Korean adoptees, and the subjects with D² exceeding 4000 are Black. Histogram B used bin widths of 1 so that all visible bars (values less than 84 on the abscissa) represent counts of White subjects. All subjects with D² less than 84 were used in the core European American group for all initial GWAS analyses.

Figure 2

Scatter plot of the first two principal components from an EIGENSTRAT analysis of 10,000 markers with color-coding to show reported ethnicity. Some of the gray dots of “Unknown” ancestry had been reported as White, but they were sufficiently far from the White centroid that they have been recoded as Unknown. Note that “Mixed” was a vague category that seems to include individuals with some European ancestry mixed with either African, Asian or Native American ancestry. Several subjects have been recoded as White because of Mahalanobis D² less than 84 indicating close proximity to the centroid of the White group in princpipal component space (see Figure 1). These included 2 of 73 reportedly Black subjects, 5 of 60 Hispanic, 17 of 77 Mixed, 9 of 44 Native American and 59 of 66 of Unknown ancestry. An additional 41 of 6564 subjects reported to be White were recoded as Unknown based on the same analysis. The graph reflects the final ethnic classifications that followed the Mahalanobis distance analysis described in the text and in Figure 1.