Table 2 Genomic prediction accuracy in NAM-H
Cross-validation prediction accuracy (P-value)Prediction accuracy (P-value)Difference in prediction accuracy (P-value)
Training panel (Testers)NAM-H (PHZ51)Ames-H (PHZ51, B47)Ames-H (PHZ51, B47)Ames × PHZ51 (PHZ51 only)Ames × B47 (B47 only)
DTS0.405 (2.2 × 10−11)0.331 (2.3 × 10−11)−0.013 (0.21)−0.028 (0.078)−0.068 (2.7 × 10−3)
PH0.394 (1.4 × 10−10)0.235 (9.6 × 10−8)+0.023 (0.021)+0.014 (0.42)−0.095 (8.5 × 10−5)
GY0.240 (2.2 × 10−9)−0.001 (0.96)−0.010 (0.33)−0.018 (0.29)+0.056 (0.014)
  • Prediction accuracy: average correlation between observed and predicted phenotypes over the 24 populations in NAM-H; cross-validation prediction accuracy: leave-one-population-out prediction accuracy within NAM-H; difference in prediction accuracy: difference in cross-panel prediction accuracy between models (DGBLUP vs. GBLUP model in Ames-H) or between training sets (Ames × PHZ51 or Ames × B47 vs. Ames-H). Significance of average prediction accuracies (nonzero mean) and estimated differences in prediction accuracy (nonzero difference, paired by NAM-H population) was assessed by Student’s t-tests. Ames, North Central Regional Plant Introduction Station association panel; Ames-H, hybrid Ames panel; NAM, U.S. Nested Association Mapping panel; NAM-H, hybrid NAM panel; GBLUP, genomic best linear unbiased prediction; DGBLUP, dominance GBLUP; DTS, days to silking; PH, plant height; GY, grain yield adjusted for DTS.