TY - JOUR
T1 - Predictive ability of genome-assisted statistical models under various forms of gene action
AU - Momen, Mehdi
AU - Mehrgardi, Ahmad Ayatollahi
AU - Sheikhi, Ayyub
AU - Kranis, Andreas
AU - Tusell, Llibertat
AU - Morota, Gota
AU - Rosa, Guilherme J.M.
AU - Gianola, Daniel
N1 - Funding Information:
Authors acknowledge the Ministry of Science, Research and Technology of Iran for financially supporting the visit of MM to the University of Wisconsin-Madison. This study was partially supported by the Wisconsin Agriculture Experiment Station under hatch grant 142-PRJ63CV to DG.
PY - 2018/12/1
Y1 - 2018/12/1
N2 - Recent work has suggested that the performance of prediction models for complex traits may depend on the architecture of the target traits. Here we compared several prediction models with respect to their ability of predicting phenotypes under various statistical architectures of gene action: (1) purely additive, (2) additive and dominance, (3) additive, dominance, and two-locus epistasis, and (4) purely epistatic settings. Simulation and a real chicken dataset were used. Fourteen prediction models were compared: BayesA, BayesB, BayesC, Bayesian LASSO, Bayesian ridge regression, elastic net, genomic best linear unbiased prediction, a Gaussian process, LASSO, random forests, reproducing kernel Hilbert spaces regression, ridge regression (best linear unbiased prediction), relevance vector machines, and support vector machines. When the trait was under additive gene action, the parametric prediction models outperformed non-parametric ones. Conversely, when the trait was under epistatic gene action, the non-parametric prediction models provided more accurate predictions. Thus, prediction models must be selected according to the most probably underlying architecture of traits. In the chicken dataset examined, most models had similar prediction performance. Our results corroborate the view that there is no universally best prediction models, and that the development of robust prediction models is an important research objective.
AB - Recent work has suggested that the performance of prediction models for complex traits may depend on the architecture of the target traits. Here we compared several prediction models with respect to their ability of predicting phenotypes under various statistical architectures of gene action: (1) purely additive, (2) additive and dominance, (3) additive, dominance, and two-locus epistasis, and (4) purely epistatic settings. Simulation and a real chicken dataset were used. Fourteen prediction models were compared: BayesA, BayesB, BayesC, Bayesian LASSO, Bayesian ridge regression, elastic net, genomic best linear unbiased prediction, a Gaussian process, LASSO, random forests, reproducing kernel Hilbert spaces regression, ridge regression (best linear unbiased prediction), relevance vector machines, and support vector machines. When the trait was under additive gene action, the parametric prediction models outperformed non-parametric ones. Conversely, when the trait was under epistatic gene action, the non-parametric prediction models provided more accurate predictions. Thus, prediction models must be selected according to the most probably underlying architecture of traits. In the chicken dataset examined, most models had similar prediction performance. Our results corroborate the view that there is no universally best prediction models, and that the development of robust prediction models is an important research objective.
UR - http://www.scopus.com/inward/record.url?scp=85051703874&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85051703874&partnerID=8YFLogxK
U2 - 10.1038/s41598-018-30089-2
DO - 10.1038/s41598-018-30089-2
M3 - Article
C2 - 30120288
AN - SCOPUS:85051703874
VL - 8
JO - Scientific Reports
JF - Scientific Reports
SN - 2045-2322
IS - 1
M1 - 12309
ER -