Limits...
Comparison of analyses of the QTLMAS XII common dataset. I: Genomic selection.

Lund MS, Sahana G, de Koning DJ, Su G, Carlborg O - BMC Proc (2009)

Bottom Line: The best BLUP models as well as the best Bayesian models gave unbiased predictions.The BLUP models are clearly sensitive to the assumed SNP variance, because they do not estimate SNP variance, but take the specified variance as the true variance.On the other hand fixed effect type models are unlikely to provide any gain over traditional pedigree indexes for selection.

View Article: PubMed Central - HTML - PubMed

Affiliation: Aarhus University, Faculty of Agricultural Sciences, Department of Genetics & Biotechnology, Research Centre Foulum, DK-8830, Box 50, Tjele, Denmark. Mogens.lund@agrsci.dk

ABSTRACT
A dataset was simulated and distributed to participants of the QTLMAS XII workshop who were invited to develop genomic selection models. Each contributing group was asked to describe the model development and validation as well as to submit genomic predictions for three generations of individuals, for which they only knew the genotypes. The organisers used these genomic predictions to perform the final validation by comparison to the true breeding values, which were known only to the organisers. Methods used by the 5 groups fell in 3 classes 1) fixed effects models 2) BLUP models, and 3) Bayesian MCMC based models. The Bayesian analyses gave the highest accuracies, followed by the BLUP models, while the fixed effects models generally had low accuracies and large error variance. The best BLUP models as well as the best Bayesian models gave unbiased predictions. The BLUP models are clearly sensitive to the assumed SNP variance, because they do not estimate SNP variance, but take the specified variance as the true variance. The current comparison suggests that Bayesian analyses on haplotypes or SNPs are the most promising approach for Genomic selection although the BLUP models may provide a computationally attractive alternative with little loss of efficiency. On the other hand fixed effect type models are unlikely to provide any gain over traditional pedigree indexes for selection.

No MeSH data available.


Design of the simulation study. 1Data provided to participants. 2400 individuals sampled randomly in each generation from population of 1500. 3True breeding values known only to organisers for validation. Numbers in parenthesis is the number of parents for the next generation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2654490&req=5

Figure 1: Design of the simulation study. 1Data provided to participants. 2400 individuals sampled randomly in each generation from population of 1500. 3True breeding values known only to organisers for validation. Numbers in parenthesis is the number of parents for the next generation.

Mentions: The pedigree was simulated in three parts as illustrated in Figure 1. First a historic population was simulated with 50 generations without records (Gh1 to Gh50), followed by 4 generations (Gt1 to Gt4; the training set), with both genotype and phenotype records, and finally the last 3 generations (Gv1 to Gv3; the validation set) were only genotypes and true breeding values were generated. The historic population was created by 100 founder individuals (50 males and 50 females) in generations Gh1. For each of the subsequent 50 generations, 50 males and 50 females were produced by randomly sampling parents from the previous generation. The base generation of the recorded pedigree (Gt1) had 15 males and 150 females. The parents for these were sampled randomly from individuals in Gh50. Each male was mated to 10 females and each mating pair produced 10 offspring. This created a fullsib-halfsib design, in which each male had 100 progeny and each female had 10 progeny and a total of 1500 individuals per generation. In the following 5 generations (Gt2 to Gt4 and Gv1 to Gv2), 15 males and 150 females were selected randomly to be parents of the next generation and the same mating design was repeated. In the last 3 generations (Gv1 to Gv3) 400 of the 1500 individuals were selected randomly to be genotyped. The resulting 1200 individuals constitute the validation set.


Comparison of analyses of the QTLMAS XII common dataset. I: Genomic selection.

Lund MS, Sahana G, de Koning DJ, Su G, Carlborg O - BMC Proc (2009)

Design of the simulation study. 1Data provided to participants. 2400 individuals sampled randomly in each generation from population of 1500. 3True breeding values known only to organisers for validation. Numbers in parenthesis is the number of parents for the next generation.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2654490&req=5

Figure 1: Design of the simulation study. 1Data provided to participants. 2400 individuals sampled randomly in each generation from population of 1500. 3True breeding values known only to organisers for validation. Numbers in parenthesis is the number of parents for the next generation.
Mentions: The pedigree was simulated in three parts as illustrated in Figure 1. First a historic population was simulated with 50 generations without records (Gh1 to Gh50), followed by 4 generations (Gt1 to Gt4; the training set), with both genotype and phenotype records, and finally the last 3 generations (Gv1 to Gv3; the validation set) were only genotypes and true breeding values were generated. The historic population was created by 100 founder individuals (50 males and 50 females) in generations Gh1. For each of the subsequent 50 generations, 50 males and 50 females were produced by randomly sampling parents from the previous generation. The base generation of the recorded pedigree (Gt1) had 15 males and 150 females. The parents for these were sampled randomly from individuals in Gh50. Each male was mated to 10 females and each mating pair produced 10 offspring. This created a fullsib-halfsib design, in which each male had 100 progeny and each female had 10 progeny and a total of 1500 individuals per generation. In the following 5 generations (Gt2 to Gt4 and Gv1 to Gv2), 15 males and 150 females were selected randomly to be parents of the next generation and the same mating design was repeated. In the last 3 generations (Gv1 to Gv3) 400 of the 1500 individuals were selected randomly to be genotyped. The resulting 1200 individuals constitute the validation set.

Bottom Line: The best BLUP models as well as the best Bayesian models gave unbiased predictions.The BLUP models are clearly sensitive to the assumed SNP variance, because they do not estimate SNP variance, but take the specified variance as the true variance.On the other hand fixed effect type models are unlikely to provide any gain over traditional pedigree indexes for selection.

View Article: PubMed Central - HTML - PubMed

Affiliation: Aarhus University, Faculty of Agricultural Sciences, Department of Genetics & Biotechnology, Research Centre Foulum, DK-8830, Box 50, Tjele, Denmark. Mogens.lund@agrsci.dk

ABSTRACT
A dataset was simulated and distributed to participants of the QTLMAS XII workshop who were invited to develop genomic selection models. Each contributing group was asked to describe the model development and validation as well as to submit genomic predictions for three generations of individuals, for which they only knew the genotypes. The organisers used these genomic predictions to perform the final validation by comparison to the true breeding values, which were known only to the organisers. Methods used by the 5 groups fell in 3 classes 1) fixed effects models 2) BLUP models, and 3) Bayesian MCMC based models. The Bayesian analyses gave the highest accuracies, followed by the BLUP models, while the fixed effects models generally had low accuracies and large error variance. The best BLUP models as well as the best Bayesian models gave unbiased predictions. The BLUP models are clearly sensitive to the assumed SNP variance, because they do not estimate SNP variance, but take the specified variance as the true variance. The current comparison suggests that Bayesian analyses on haplotypes or SNPs are the most promising approach for Genomic selection although the BLUP models may provide a computationally attractive alternative with little loss of efficiency. On the other hand fixed effect type models are unlikely to provide any gain over traditional pedigree indexes for selection.

No MeSH data available.