Sampling genotype configurations in large complex pedigree

Efcient genotype samplers are needed for Bayesian and maximumlikelihood analysis of complex genetic problems implemented via Markov Chain Monte Carlo (MCMC) algorithms. The examples of such analysis include polygene mapping in complex pedigrees and prediction of total genetic value using genome-wise dense marker maps. For large complex pedigree sampling from desired probability is impossible. We present a simple method to sample genotype congurations for large pedigree from approximate probability. The sampler uses combination of exact (simple peeling) and iterative methods (iterative peeling) to approximate target probability. Two techniques were applied to reduce computational burden: genotype elimination and set-recoding of alleles. The new sampler was evaluated on a large complex pedigree using simulated data sets for various experimental designs and degree of marker polymorphism. The pedigree used in simulation was real bovine pedigree of 907 903 animals born between 1960 and 2005 derived from Belgian dairy and dual-purpose cattle database. Four types of experimental designs were considered: (i) genotyping sires only, (ii) genotyping dams only, (iii) genotyping half of the dams but no sires, and (iv) genotyping half of the sires and half of the dams. For hypothetical single nucleotide polymorphism the new sampler reached 100%, 100%, 44% and 89% of maximum efciency for the four experimental designs respectively. For microsatellite polymorphism the efciency of the sampler reached 100%, 100%, 32% and 76%, respectively. To exemplify the use of the new sampler it was applied to estimate genes shared identical by descent (IBD). The calculation of genes shared IBD among relatives is an important component of gene mapping in complex diseases and quantitative traits. The convergence diagnostic methods gave indirect evidence of irreducibility of the new sampler and showed its good mixing performance.