Association Tests and Estimation of Haplotype Frequencies and of Penetrance-related Parameters in a Case-control Study
Shiori Furihata, Toshikazu Ito, and Naoyuki Kamatani
Japan Biological Information Research Center

We have developed an algorithm for association tests between an individual qualitative phenotype and specific haplotypes based on an Expectation-maximization method. The algorithm is implemented in the computer program PENHAPLO. Haplotype frequencies and diplotype-based penetrances are simultaneously estimated using the algorithm. In this study, we investigate the applicability of PENHAPLO in a case-control study. First, we show the meaning of the penetrance-related parameters estimated for a case-control study when using PENHAPLO. If prevalence of disease is known for a population, penetrances for specific haplotypes can be calculated using the estimated parameters. Simulations have confirmed that type I errors and the penetrance-related parameters under null hypothesis have been accurately estimated. The results are comparable to those analyzed using known-phase data, even when the loci are not within the haplotype block. We also compare the results obtained using PENHAPLO with those collected using an alternative method. Using this second method, the haplotype frequencies are first estimated, then association between case-control and specific haplotypes are tested using a contingency table. Under the alternative hypothesis, the statistical power calculated using PENHAPLO is greater than that calculated using the alternative method. For data with weak linkage disequilibrium, it was observed that type I errors under null hypothesis were severely underestimated when inferring haplotype frequencies and testing association were performed separately. It has been shown in this work that PENHAPLO is a useful tool for analysis of case-control studies concerning haplotype and qualitative phenotype association.