Modela
| Training sample | Testing sample |
---|
Phenotyped fish | Genotyped fish | Effective SNPs |
\(h^{2}\)
b
| Genotyped fish | Predictive abilityc
| Biasd
|
---|
P-BLUP | 7893 | 0 | 0 | 0.37 | 0 | 0.34 | 0.86 |
ssGBLUP | 7893 | 1473 | 35,636 | 0.33 | 930 | 0.63 | 0.99 |
wssGBLUP2 | 7893 | 1473 | 35,623 | 0.33 | 930 | 0.67 | 0.71 |
wssGBLUP3 | 7893 | 1473 | 35,623 | 0.33 | 930 | 0.65 | 0.65 |
BayesB | 1473 | 1473 | 35,636 | 0.23 | 930 | 0.71 | 1.16 |
-
aThe estimated breeding values (EBV) were estimated with a pedigree-based animal model (P-BLUP); and the genomic EBV (GEBV) were estimated with three genomic selection (GS) models: single-step GBLUP (ssGBLUP), weighted ssGBLUP (wssGBLUP) and Bayesian method BayesB. The wssGBLUP2 and wssGBLUP3 corresponds to iteration 2 and 3, respectively
-
bFor the GS models, \(h^{2}\) is the proportion of phenotypic variance explained by the markers. For the P-BLUP model, \(h^{2}\) is the trait narrow-sense heritability estimated from pedigree and phenotypic records
-
cThe predictive ability of EBV \(\left( {PA_{EBV} } \right)\) or GEBV \(\left( {PA_{GEBV} } \right)\) was defined as the correlation of mid-parent EBV or GEBV with MPP from each PTF: \(PA_{EBV} = CORR\left( {MPP, \;Midparent\;EBV} \right)\); \(PA_{GEBV} = CORR\left( {MPP, \;Midparent\;GEBV} \right)\)
-
dThe bias of EBV \(\left( {Bias_{EBV} } \right)\) or GEBV \(\left( {Bias_{GEBV} } \right)\) was defined as the regression coefficient of performance MPP on predicted mid-parent EBV or GEBV: \(Bias_{EBV} = REGRES\left( {MPP, \;Midparent\; EBV} \right); \;Bias_{GEBV} = REGRES\left( {MPP, \;Midparent \;GEBV} \right)\)