TABLE 3

Cross-validation (CV) correlation between predicted and observed phenotypes, obtained in a 10-fold CV conducted for grain yield (GY) records of 599 wheat lines genotyped with 1279 molecular markers

Modela
Trait–environmentPM-RKHSPM-RKHSM-BLPM-BLBLUPb
Correlation
GY-E10.4480.6010.6080.5180.5420.480
GY-E20.4170.4940.4970.4930.5010.488
GY-E30.4170.4450.4780.4030.4490.355
GY-E40.4490.5240.5240.4570.4950.464
% change (relative to P)
GY-E134.235.715.621.07.1
GY-E218.519.218.220.117.0
GY-E36.714.6−3.47.7−14.9
GY-E416.716.71.810.23.3
  • Six models were fitted to GY measured in four environments (E1, E2, E3, and E4).

  • a Models were pedigree model (P), molecular marker model using reproducing kernel Hilbert space (M-RKHS) regression, pedigree plus molecular marker model using reproducing kernel Hilbert space regression (PM-RKHS), molecular marker regression model using the Bayesian LASSO (M-BL), pedigree plus molecular marker model regression using the Bayesian LASSO (PM-BL), and best linear unbiased prediction (BLUP) using marker genotypes.

  • b Values of genetic variances used to compute BLUP ranged from 0.8065 to 0.9141.