Haplotypic analysis of Wellcome Trust Case Control Consortium data

Abstract

We applied a recently developed multilocus association testing method (localized haplotype clustering) to Wellcome Trust Case Control Consortium data (14,000 cases of seven common diseases and 3,000 shared controls genotyped on the Affymetrix 500 K array). After rigorous data quality filtering, we identified three disease-associated loci with strong statistical support from localized haplotype cluster tests but with only marginal significance in single marker tests. These loci are chromosomes 10p15.1 with type 1 diabetes (p = 5.1 x 10(-9)), 12q15 with type 2 diabetes (p = 1.9 x 10(-7)) and 15q26.2 with hypertension (p = 2.8 x 10(-8)). We also detected the association of chromosome 9p21.3 with type 2 diabetes (p = 2.8 x 10(-8)), although this locus did not pass our stringent genotype quality filters. The association of 10p15.1 with type 1 diabetes and 9p21.3 with type 2 diabetes have both been replicated in other studies using independent data sets. Overall, localized haplotype cluster analysis had better success detecting disease associated variants than a previous single-marker analysis of imputed HapMap SNPs. We found that stringent application of quality score thresholds to genotype data substantially reduced false-positive results arising from genotype error. In addition, we demonstrate that it is possible to simultaneously phase 16,000 individuals genotyped on genome-wide data (450 K markers) using the Beagle software package.

Figure 1

Flowchart summarizing the genotype quality filters applied to the data to screen out results that may be false positive due to poor genotype quality. A detailed description of each filter is given in the Materials and Methods section. The flowchart also shows the number of regions containing highly significant localized haplotype clusters after each filtering step; full details are given in the Results section.

Figure 2

Single genotyped marker (filled circle), single imputed marker (open circle) and localized haplotype cluster (cross) p-values in 4 regions showing strong evidence of association between trait and localized haplotype clusters Allelic and genotypic test p-values are shown for single markers, and allelic test p-values are shown for localized haplotype clusters. All p-values were calculated using Fisher's exact test. . The candidate region, delimited by the dashed vertical lines, contains all markers that are in moderate to strong linkage disequilibrium with the signal based on pairwise linkage disequilibrium data and recombination hotspot data from phase 2 of the HapMap. Genes in each delimited candidate region are listed the upper right corner of the plot. Positions are in NCBI Build 35 coordinates.