Each of these groups was then analyzed with the computer program STRUCTURE to identify possible sub-structure in the sample of 286 unrelated individuals [6]. This method assumes that the sample contains a mixture of subpopulations and that within each subpopulation there is Hardy-Weinberg and linkage equilibrium between markers. This program identifies subpopulations of individuals who are genetically similar through a Markov chain Monte Carlo (MCMC) sampling procedure [7]. The STRP and SNP sets were run for 10,000 iterations after 10,000 burn-in replicates assuming a two-cluster solution. Although we tested various numbers of clusters, the solutions involving more than two clusters merely divided the European Americans into subgroups; for maximum parsimony, we used only the two cluster solution. The STRP results were used as a gold standard for comparison.
