Assessing and testing for population stratification and its impact on power of family-based association studies in the United States

The case-control design for genome-wide association studies has gained popularity as an alternative to family-based studies amongst geneticists because it is a very powerful method for detecting an association between a disease trait and a marker. Furthermore, it is often easier to recruit cases than to recruit probands along with their entire families. However, case-control studies are susceptible to increased false positive error rates in the presence of population stratification. When a study population is comprised of individuals with many different genetic ancestries (population stratification), the validity of genome-wide association studies may be compromised. Stratification into broad racial categories has been used to reduce this problem. We use 128 alleles from over 200 populations with U.S. Census 2000 ancestry data and provide estimates of the amount of residual population stratification remains after using this approach. Geographic maps with the distribution of population stratification (measured by Wright's F ST ) are created, and the proportion of genetic variation at the county, state and regional level are calculated for each of the broad racial categories (White, European-derived White, Asian, Hispanic and Native Hawaiian Pacific Islander). The Transmission Disequilibrium Test (TDT) has gained popularity because it has the desired Type I error rate in the presence of population stratification. We derive a method to calculate the true frequency of the mating types without assuming Hardy-Weinberg Equilibrium (HWE), showing that stratification results in an increase in the observed frequency of matings between individuals of similar genotypes compared to that expected using the product of the respective genotype frequencies that comprise the mating type and also results in a deficiency of heterozygotes compared to that expected assuming HWE. This causes considerable variation in the sample sizes required for adequate power within a genetic model for a fixed risk allele frequency--implying that two similarly sized studies may have different abilities to detect the same risk allele. We use the above phenomena to construct the population stratification test (PST). Power calculations show the PST is more powerful than the HW test. The PST detects stratification at the marker locus in question, and does not require genotyping additional markers