Estimating relationships and relatedness from dense genome-wide data

Relationship and relatedness estimation from genetic markers is relevant to many areas, including genealogical research, genetic counselling, forensics, linkage analysis and association analyses in genetic epidemiology. Traditionally unlinked genetic markers (microsatellites) are used. But the problems which can be solved by such markers are limited. Linked genetic markers are not only available in much larger numbers, but also provide extra information which is not available from unlinked markers. It is desirable to exploit the increasing availability of dense genome-wide single nucleotide polymorphisms (SNP) data for estimating relationships and relatedness. While Method of Moments (MoM) methods and other non-pedigree approaches only give a degree of pairwise relatedness, a pedigree likelihood approach can distinguish exact relationships. The pedigree likelihood approach also has advantages in that extra individuals can be considered jointly and extra data such as Y-chromosomal and mitochondrial SNPs can be incorporated with autosomal SNPs easily. In this thesis I firstly confirm that the increase in information obtained from large sets of linked markers substantially increases the number of problems that can be solved with pedigree approach. Furthermore, when two distant relatives do share genome segments through identity by descent (IBD), we usually have greater power to distinguish more distant relatives from unrelated pairs than was previously believed. Data on extra individuals always improve discriminatory power, but the position of the extra individuals in the pedigree dictates the extent of this increase of power. Linkage Disequilibrium (LD) is an issue for pedigree likelihood approach and it needs to be dealt with. MoM methods are easy to use and are generally robust to the effect of LD, but they are only accurate for relatives up to second cousins. I propose using pedigree likelihood approach to estimate pairwise relatedness and find we can greatly improve the accuracy in detecting distant relatives.