To date most individual GWA studies do not have enough power to detect true associations at the conservative significance levels necessary to distinguish false positives from false negatives. This point has typically been made by referencing the large number of tests conducted in a GWA study and the consequent severe adjustment of the p-value threshold in order to control experiment-wide Type I error rate. Empirical estimates of the threshold needed to preserve the genome-wide Type I error rate in studies of European-ancestry subjects using current genotyping arrays range from 5×10−7 to 1×10−8. [8–11] These thresholds are different in other populations; for example, they are even lower in African or African-American samples, due to the greater genetic diversity in these populations. Even these stringent thresholds take into account only the complexity of the genetic architecture, and they do not adjust for the potential complexity of the phenotypic architecture, i.e. when targeting multiple phenotypes.
