Genomic prediction based on data from three layer lines using non-linear regression models

Table 4 Performance comparison of seven prediction methods in seven training scenarios for line W1

The predictive correlation is computed as the correlation coefficient of the predicted value and phenotype of line W1; GBLUP: genome-enabled best linear unbiased prediction; RRPCA: ridge regression principal component analysis; MTGBLUP: multi-trait GBLUP; Poly: polynomial kernel based linear models; RBF: radial basis function kernel based linear models; RR/Poly/RBF-PCA: the model with the features reduced by PCA.
Approximated SE across the genomic prediction models and training data sets ranged from 0.045-0.060.
¹Results are presented by Calus et al. [22].

ISSN: 1297-9686