Statistical Implications of Unobserved Parental Genotypes in Estimating Linkage Disequilibrium

• Activity Number: 183
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Bevan Emma Huang*+ and Colin Cavanagh
• Companies: CSIRO Mathematics, Informatics, and Statistics/Food Futures National Research Flagship and CSIRO Plant Industry/Food Futures National Research Flagship
• Keywords: linkage disequilibrium ; association mapping ; multiallelic ; bias ; genomic structure ; latent data

Abstract

Estimation of linkage disequilibrium (LD) plays important roles both in characterizing the genetic history of populations and in designing genomewide association studies. Numerous estimators have been devised for biallelic loci and have been extended to multiallelic markers. However, not all of these estimators perform equally well for both types of data. We consider their behavior when observed genetic data are biallelic surrogates for unobserved multiallelic genotypes which identify a limited number of genetic founders. This occurs in practice for complex experimental crosses. We estimate LD in two ways: first, based directly on the biallelic data; and second, based on imputing the fully informative multiallelic markers which identify each founder. We compare the performance of common LD measures for these two options through simulations and practically relevant multiparent experimental cross data. Characterizing the relationship between the two approaches as the design approaches the structure of a typical association mapping population will demonstrate the danger in ignoring the unobserved parental alleles for certain LD measures.