were inserted evenly, and the true breeding value
was the sum of the QTL effects. Markers were placed
at 1-cM intervals throughout the genome. The
effects of marker haplotypes were estimated for each
interval (50 000 effects in total). Then using the
genotype of the animal and the estimated haplotype
effects, an estimated breeding value (EBV) was calculated
as the sum of the haplotype effect estimates
corresponding to the genotype (haplotypes) of the
animal. This EBV will be denoted as GEBV, for genome-
wide EBV. The estimated haplotype effects are
assumed to be general population estimates and not
specific to any one animal. Haplotype effects are
assumed to be additive and epistasis among SNPs is
assumed not to exist. Thus, GEBV could be calculated
for resulting progeny as long as they were genotyped
and marker haplotypes determined. The
remarkable features of this approach were that the
correlation of GEBV with true breeding values was
0.78–0.85 (for typical heritability values) and that
animals could obtain a GEBV at birth with 0.80
accuracy. Usually in dairy cattle, females seldom
reach this level of accuracy, and bulls take 6 years or
more to reach this accuracy in their EBVs.
A similar simulation study by Kolbehdari et al.
(unpublished data) verified the results of Meuwissen
et al. (2001) using different heritabilities, and either
evenly or randomly spaced QTL. Randomly spaced
QTL gave better results than evenly spaced QTL.
Correlations between GEBV and true breeding values
of around 0.80 were found. Therefore, assuming that
GEBV with high levels of accuracy are achievable at
an early age, the question is how to take advantage of
these properties. How can traditional progeny testing
schemes be modified (or replaced) so as to make faster
genetic change? How much faster can it be? The
purpose of this paper was to look at these questions.