Recursive Long Range Phasing and Long Haplotype Library Imputation: Building a Global Haplotype Library for Holstein cattle

Long range phasing (LRP) is a fast and accurate rule based method which uses information from both related and unrelated individuals by invoking the concepts of surrogate parents and Erdös numbers (Kong et al., 2008). Recursive long range phasing and long haplotype imputation (RLRPLHI; Hickey et al., 2009) is an extended LRP algorithm with increased robustness partially due to the extra long haplotype imputation step (LHI) which is based on a the construction of a library of long haplotypes (e.g. 10cM) for a dataset. The LHI part of the algorithm is computationally much less intensive and less error prone compared to the LRP part, and it can easily incorporate prior information from other phasing methods, both laboratory and in silico. After building the initial haplotype library it may be possible to phase other genotyped individuals from the same population simply via the LHI step. The application of such an approach may be useful where small numbers of extra individuals are regularly added to a data set of previously phased individuals, especially where very many animals are genotyped making even a fast phasing method feasible for only small subsets of the data. It may also be useful through incorporation into the LRP part of RLRPLHI in order to improve the accuracy and speed of RLRPLHI. Another application of a long haplotype library could be to use it in combination with segregation analysis (Kerr and Kinghorn, 1996) to impute dense genotype or sequence data in ungenotyped individuals in a sparsely genotyped pedigree. The objectives of this research were to evaluate LHI for phasing genotyped individuals, and segregation analysis and long haplotype imputation (SALHI) for imputing genotypes in ungenotyped individuals, using simulated genotypic data in a national and global Holstein pedigree.