Table 2 Predictive ability, root mean square error (RMSE), and prediction bias for growth (BWSel and ADG) and carcass-related (REA, BF, and RF) traits in Nellore cattle.

From: Variable selection strategies for genomic prediction of growth and carcass related traits in experimental Nellore cattle herds under different selection criteria

Trait1

Predictive ability 2

Root mean square error

Prediction bias3

GBLUP

BayesB

Enet

RD (%) BayesB

RD (%) Enet

GBLUP

BayesB

Enet

GBLUP

BayesB

Enet

BWSel

0.69

0.69

0.75

0.0%

8.7%

12.90

18.94

10.1

1.08

0.95

1.01

ADG

0.679

0.679

0.761

0.0%

12.1%

0.03

0.05

0.02

0.97

1.07

1.01

REA

0.728

0.753

0.828

3.4%

13.7%

2.51

3.16

1.97

0.98

1.06

0.99

BF

0.651

0.66

0.745

1.4%

14.4%

0.37

0.5

0.29

0.97

0.95

0.99

RF

0.742

0.777

0.8

4.7%

7.8%

0.22

0.37

0.17

1.06

0.94

1.01

  1. 1 BWSel, Body weight at selection; ADG, average daily gain; REA, rib eye area obtained by ultrasound; BF, subcutaneous backfat thickness obtained by ultrasound; and RF, rump fat thickness obtained by ultrasound.
  2. 2Predictive ability was assessed by Pearson’s correlation (r) between phenotypes adjusted for fixed effects (\(\:{y}_{i}^{*}\)) and predicted adjusted phenotype (\(\:{\widehat{\text{y}}}_{\text{i}}^{\text{*}}\)). RD - relative difference (RD) in prediction accuracy assessed as \(\:RD\:\left(\%\right)=\frac{({r}_{m1}-{\:r}_{GBLUP})}{{r}_{GBLUP}}\times\:100\), where \(\:{r}_{m1}\) is the predictive ability using BayesB or ENet and \(\:{r}_{GBLUP}\) is the predictive ability using the GBLUP approach.
  3. 3Slope of linear regression of phenotypes adjusted for fixed effects (\(\:{y}_{i}^{*}\)) on predicted adjusted phenotype (\(\:{\widehat{\text{y}}}_{\text{i}}^{\text{*}}\)).
Back to article page