The Power of Fourier and Wavelet Transforms of Genetic Data in Genotype-phenotype Association Tests under both Monogenic and Polygenic Inheritance

Genetic association tests seek to determine if a candidate gene or region contributes to phenotypic variation. Usually, the causative polymorphism within a region is not among the set of genotyped loci. To be detected, the causative locus must be highly correlated with at least one genotyped locus. In these cases, combining genotype information across a region should increase the power to detect causative loci. Nevertheless, a multiple marker approach can increase the degrees of freedom of a test. Approaches involving basis-function transforms (Fourier and wavelet) have been proposed to decrease degrees of freedom. We define a powerful score statistic, and we compare its power to detect associations using untransformed (raw genotype) data and transformed data (both Fourier and wavelet). Our results indicate that over a broad set of genetic architectures, association tests with raw genotype data have more power than tests with transformed data. In scenarios of polygenic inheritance, associations between traits and loci not closely linked to causative loci occur at high frequencies.