Skip to main content
ARS Home » Research » Publications at this Location » Publication #208314

Title: Methods to explain genomic estimates of breeding value

item Vanraden, Paul
item Tooker, Melvin

Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 3/7/2007
Publication Date: 7/8/2007
Citation: Van Raden, P.M., Tooker, M.E. 2007. Methods to explain genomic estimates of breeding value. Journal of Dairy Science. 90(Suppl. 1):374(abstr. 413).

Interpretive Summary:

Technical Abstract: Genetic markers allow animal breeders to locate, estimate, and trace inheritance of many unknown genes that affect quantitative traits. Traditional models use pedigree data to compute expected proportions of genes identical by descent (assumed the same for all traits). Newer genomic models use thousands of marker genotypes to obtain actual fractions of DNA shared by any two individuals. Full sibs, for example, may actually share 45% or 55% of their DNA rather than the 50% expected in the traditional relationship matrix. The actual percentage of shared genes has a standard deviation (SD) equal to 50% divided by the square root of twice the number of independent loci affecting the trait. This SD does not decrease below 3.5% even with very large numbers of loci on 30 chromosome pairs because loci on the same chromosome are linked rather than independent. Accounting for these small differences in the relationship matrix and tracing individual genes can greatly increase reliability, especially if the number of genotyped individuals is large. Reliability from parent average is <50% and is the upper limit for individuals without phenotypic data or progeny in traditional models. With genotypes and phenotypes for full sibs, genomic models can increase the reliability to 62% with 100 full sibs or 95% with 1000 full sibs. If no sibs have both genotypes and phenotypes but 100 full sibs of each parent do, reliability can increase to 55%. Less gain is provided from each distant relative, but the number of distant relatives may be very large. “Unrelated” individuals actually share many unknown common ancestors born prior to the known pedigree file and thus can provide additional information. More markers are required to estimate and trace genetic effects for distant rather than close relatives because the shared DNA segments are shorter. More markers are also required to locate and estimate genetic effects when more loci affect a trait. Another useful concept is the proportion of genes in common that affect a particular trait, called a quantitative trait loci (QTL) relationship matrix. Genetic evaluations should be more accurate if genomic relationships replace the traditional relationships computed from pedigrees.