Scalable Methods for the Inference of Identity by Descent

Abstract

Identity by descent (IBD) describes the shared inheritance of DNA and underlies genetic similarity between individuals. Estimated IBD graphs describing the IBD relationships among individuals have many uses in statistical genetics. An important application is to detect, through linkage analysis, the location of genes that cause genetic diseases. Accurate estimation of IBD graphs among large groups of individuals is essential. IBD is typically estimated either among individuals in small family pedigrees or among distantly related individuals sampled from a population with an unknown pedigree relationship. Both pedigree and population approaches require different modeling assumptions and are applied to different study designs. In this thesis, scalable methods of estimating IBD that combine both pedigree and population estimation methods are developed for family based studies. IBD is estimated between the founders of the family pedigrees to incorporate more information about shared inheritance without additional data collection. A combined IBD model is developed for sib-pair studies, and is demonstrated on siblings from a 50 generation simulated population. A merging algorithm is also developed to combine pedigree and population IBD estimates for family based studies with larger component pedigrees. The merging method is demonstrated on pedigrees from the simulated population and on an Alzheimer’s disease family study. Combined IBD models are shown to increase power and resolution for locating genes that cause genetic diseases.