MCMC for Co-Ancestry in Pedigrees and Populations

• Activity Number: 223
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Elizabeth Thompson*+
• Companies: University of Washington
• Keywords: Identity by descent (IBD) ; IBD graph ; Markov inheritance ; Latent variables ; Ancestral recombination graph (ARG)

Abstract

The descent of genome from one generation to the next underlies the genetically mediated within-species similarities among organisms and among populations. This identity by descent (IBD) of genome from common ancestors can be summarized as an IBDgraph expressing the sharing of genome among observed individuals. More generally ancestry across the genome can be expressed as an ancestral recombination graph (ARG). However, the IBDgraph or ARG cannot be directly observed. The latent structure provided by descent of genome, and the resulting complex patterns of dependence, have been widely addressed by MCMC sampling, conditionally on genetic marker data. Within defined pedigrees, the structure and descent are tightly constrained, leading to challenges in obtaining good samples of IBD. As genetic data increase, IBD can be inferred without pedigree information. Using a population model, new MCMC methods permit sampling of either the ARG or IBD among a set of gametes, and hence the IBDgraph among the individuals who carry them, and pedigree-based and population-based IBD can be combined in the IBDgraph. Chris Glazner, Mary Kuhner, and Chaozhi Zheng contributed to parts of this work.