Using The Hybrid Model To Study The Association Of Genes With Disease

The case-control study is a tool often used in epidemiology. However, if there is a confouder, which is associated with both the case and the risk factor studied, confounding will be brought into the study and it will come to an incorrect conclusion of association. The main approach to solve this problem is to stratify the population studied through taking the confounder as the stratifying variable. In most cases, the confounder is unobservable, thus choosing a proper covariable is very important.In this paper, we take the multilocus genotype data as the covariable to the confounder and propose an approach using mixture models to detect the association between a candidate gene and a disease, while there is substructure in the population in a case-control study. In our approach, we make full use of the known information and stratify the population using genetic markers while considering that HardyWeinberg equilibrium holds in each subpopulation. EM algorithm is used in the course of estimating. Meanwhile, the association between the candidate gene and the disease is estimated. Additionally, we have tested our approach using simulated data based on allele frequencies in 12 short tandem repeat (STR) loci from four subpopulations, and compared our results with that of Satten et al. (2001). It can be concluded from the simulation that our approach is more accurate and more efficient. Besides, we consider the case when a part of multilocus genotype data are missing.