Table 4 Average predictive abilities of four methods for predicting 1000 metabolomic traits and pair-wise comparisons of the numbers of higher predicted metabolites between methods

From: Genomic selection using principal component regression

Method

Predictive ability

>PCR-CV

>PLSR-CV

>BLUP-HAT*

PCR-HAT

0.2218 ± 0.2157aA

523

856

211

PCR-CV

0.2233 ± 0.2191aAB

 

949

165

PLSR-CV

0.2033 ± 0.2231bA

  

6

BLUP-HAT

0.2453 ± 0.2233cB

   
  1. Class labels a, b, or c denote significant differences between methods at level of p < 0.05
  2. Class labels A, B, or C denote significant differences between methods at level of p < 0.01
  3. *Indicates the numbers of metabolites with higher predictive ability than BLUP-HAT for the methods listed in the first column. For example, of the 1000 metabolites, the PCR-HAT method has 211 metabolites with predictive ability higher than the BLUP-HAT method
Back to article page