Submitted to: Interbull Annual Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: August 23, 2007
Publication Date: December 31, 2007
Citation: Van Raden, P.M. 2007. Genomic Measures of Relationship and Inbreeding. Interbull Annual Meeting Proceedings. Interbull Bulletin 37:33-36.
Interpretive Summary: New tools let scientists measure how related individuals are by examining pedigrees, finding out the amount of DNA that individuals have in common, comparing their outward characteristics, or a combination of all those methods. Statistical models that include genomic information can predict genetic effects more accurately than traditional models that use pedigrees to compute expected proportions of genes identical by descent. Newer genomic models use thousands of marker genotypes to determine actual fractions of DNA shared by any two individuals. Full siblings, for example, may share 45% or 55% of their DNA rather than the 50% expected using traditional methods. The accuracy of genetic merit estimates based on solely on pedigree data is less than 50%. However, in simulations, accuracy of genetic estimates increased to 77% by including genomic data for 100 full siblings and to 97% for 1,000 full siblings. Increased accuracy of genetic evaluations that include genomic relationships will allow dairy producers to make better breeding choices.
Genetic similarity can be defined in several ways using pedigree data, genotypic data, phenotypic data, or combinations of those data. Models that include genomic relationships can predict genetic effects more accurately than those that use expected relationships from pedigrees. Relationship matrices can estimate the expected fraction of genes identical by descent, the actual fraction of DNA shared, or the fraction of alleles shared for loci that affect a particular trait. Each may be a valid answer to the question “Are two individuals related?” Several options are available for including genomic relationships in genetic evaluations. Use of exact fractions of shared genes in the genomic relationship matrix can provide more accurate predictions than use of the expected fractions in the additive genetic relationship matrix. Non-linear models can change the weights on individual markers to match actual fractions of shared alleles of quantitative trait loci in the relationship matrix for a specific trait more closely. Full siblings may actually share 45 or 55% of their DNA rather than the expected 50%. Accounting for those small differences in the relationship matrix and tracing individual genes can greatly increase reliability, especially if the number of genotyped individuals is large. In simulations with 50,000 markers, reliability increased to 77% by including genomic data for 100 full siblings and to 97% for 1,000 full siblings.