Whole-genome evaluation of complex traits using SNP, haplotype, or QTL information
Abstract:
Whole-genome
technologies provide rich data for
dissection of complex traits. While gene discovery is still largely
limited, the data at hand can be successfully used for evaluation of
genetic merit. The aim of this work was to demonstrate the value of
different sources of information (pedigrees, Single Nucleotide
Polymorphisms – SNP, haplotypes, or Quantitative Trait Loci –
QTL) for genetic evaluation of non-phenotyped individuals in a
typical animal breeding scenario via simulation. In the first step a
coalescent simulation was used to create a base population with
structured chromosomes that were in the second step dropped and
recombined through the pedigree of 10 generations with 50 sires per
generation, 10 dams per sire, and 2 offspring per dam. Phenotypic
values were simulated with different genetic architectures (QTL
effects were sampled from Gaussian or gamma distribution and minor
allele frequency less than 0.3) and heritability of 0.25. Genotypic
data was available for all individuals from generation 4 onwards,
while phenotypic data was available for individuals in generations 4
and 5. Genetic evaluation was based on linear mixed models with
relationship matrix between individuals. This matrix was built using
pedigree, SNP, haplotype, or QTL data. Haplotypes of different length
were considered (from 5 to all the way up to 2000 SNP) with an option
to account for similarities between haplotypes while building
relationship matrices. The accuracy of different methods was assessed
by correlation between true and evaluated additive genetic values for
individuals in generations 6, 8 and 10. Average accuracy over ten
replications for Gaussian trait over generations was between 0.45 to
0.10 for pedigree data, 0.50 to 0.35 for SNP and haplotype data and
0.6 to 0.4 for QTL data. In the case of long haplotypes accuracies
dropped considerably, but accounting for similarities between
haplotypes prevented this drop. In the case of gamma trait accuracies
were slightly higher in generation 6 and dropped faster in the later
generations in the case of pedigree, SNP, and haplotype data due to
recombinations. On the other hand accuracies were substantially
higher with QTL data and quite stable over generations (from 0.75 to
0.65) though still far from perfect (even though QTL genotypes are
known), due to estimation errors. Results demonstrate the value and
limitations of genotypic information for the evaluation of additive
genetic merit in animal populations.
