HWE should therefore not be eliminated from the analysis, but flagged for further analysis after the association analyses are performed. Databases such as MySQL (see Table 1) can be very useful for joining association statistics with HWE statistics for easy reporting. It is also beneficial to examine HWE in controls separately, as disease-free controls should more closely follow the assumptions that lead to HWE than cases, and because some true associations are expected to be out of HWE. If multiple ethnicities are used in the same study it is necessary to test for HWE within each group separately. SNPs that are highly associated with the trait of interest that also show highly significant departures from HWE, especially in controls, should be closely scrutinized. Typically HWE deviations toward an excess of heterozygotes reflect a technical problem in the assay, such as non-specific amplification of the target region. On many GWAS platforms the quantitative allelic signals at a marker, i.e., the intensity plot for the SNP, can be used to screen for a technical origin of the HWE deviation: null alleles can produce multimodal genotype clusters in the heterozygote clusters and one of the homozygote clusters (Figure 8) or can produce an
