Haplotypes were reconstructed using the accelerated expectation-maximization algorithm implemented in Haploview for genes that include at least one marker whose P-value was less than 0.01 in the single-locus association test. Haplotype association analysis was performed for each of the reconstructed haplotypes (against the rest) using the Fisher's exact test. Multi-locus genotype patterns (MLGP) of the markers in the blocks identified by Haploview were determined. For a set of a given number m SNPs, n observed m locus genotype pattern frequencies were stored in an n x 2 contingency table where the two columns corresponded to cases and controls. A Fisher's exact test for differences in MLGP frequencies between the two groups was performed. In addition to single-locus effects, MLGP analysis also evaluates epistatic interaction influencing the outcome. Since we are not interested in the effects of rare genotype patterns, in this analysis, pattern frequencies below 5% in cases and controls each were pooled into a single “rare” class.
