Center for Bioinformatics and Molecular Biostatistics

Admixture mapping is potentially a powerful method for mapping genes for complex human diseases, when the disease frequency due to a particular disease susceptible gene is different between founding populations of different ethnicity. The method tests for genetic linkage by detecting association of the allele ancestry with the disease. Since the markers used to define ancestral populations are not fully informative for the ancestry status, direct test of such association is not possible. In this paper, we develop a hidden Markov model (HMM) framework for estimating the unobserved ancestry haplotypes across a chromosomal region based on marker haplotypes. The HMM efficiently utilizes all the marker data to infer the latent ancestry states at the putative disease locus. In this modeling framework, we consider a likelihood based approach for detecting genetic linkage based on case-control data. We evaluate by simulations how several factors affect the power of admixture mapping, including sample size, ethnicity relative risk, marker density and the different admixture dynamics. Our simulation results indicate correct type 1 error rates of the proposed likelihood ratio test and great impact of marker density on the power. In addition, simulation results indicate that the methods work well for the admixed populations derived from both hybrid-isolation and continuous gene-flowing models.