Direct Estimation of Haplotype Frequencies in Genetic Association Studies

• Activity Number: 12
• Type: Topic Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Stefan Boehringer*+
• Companies: Leiden University Medical Center
• Keywords: genetic association ; haplotype ; likelihood ; multipoint ; algorithm ; families

Abstract

Current genome wide association studies (GWASs) are dominated by individual single nucleotide polymorphism (SNP) analysis. It is well recognized that haplotype analyses have advantages over marker by marker analyses in many situations. However, broad application of haplotype analysis is hampered by the lack of efficient algorithms. This study suggests a new algorithmic approach based on a reparametrized and modified likelihood as compared to a standard EM formulation. Starting with closed form solutions for the estimation of allele frequencies, covariances between alleles can be estimated independently. This idea can be generalized to covariances between different haplotypes leading to a completely non-iterative approach. Asymptotic behavior is characterized and simulations show that the accuracy of the new algorithm is identical to that of the EM. Computation time is vastly reduced.