Sparse single-step method for genomic evaluation in pigs

Table 2 Correlations between EBV from alternate core groups and EBV from the full single-step model for DanAvl Duroc

Scenario	Cor all	Cor genotyped	PCG iterations	Sparsity of \(({\mathbf{G}}_{\varvec{ }}^{ - 1} - {\mathbf{A}}_{22}^{ - 1}\))
NormalG^a	1	1	301	0.0 %
Random10^b	0.993	0.981	309	76.9 %
Unrelated10^c	0.968	0.944	412	77.4 %
Offspring10^d	0.996	0.985	298	78.5 %
OffspringRandom10^e	0.997	0.989	287	78.1 %
Random30^b	0.999	0.997	306	46.5 %
Random50^b	1.000	0.999	284	23.7 %
Old10^f	0.947	0.939	405	77.7 %
Young10^g	0.963	0.934	370	76.1 %
NormalA^h	0.965	0.901	320	–

Correlations were calculated for all animals (Cor all) and genotyped animals (Cor genotyped)
Number of PCG iterations and sparsity of the matrix involved in the single-step formula \(({\mathbf{G}}_{ }^{ - 1} - {\mathbf{A}}_{22}^{ - 1}\))
All correlations were significantly different from each other (p < 0.05)
^aNormalG is the usual single-step procedure without sparse approximations
^bRandom10, Random30, Random50 are the sparse single-step, where a random subset of animals (10, 30, 50 %) were treated as core
^cUnrelated10 is 10 % animals chosen as core by minimizing the degree of relatedness between core animals
^dOffspring10 is 10 % animals chosen based on the number of genotyped offspring
^eOffspringRandom10 is, for old animals (excluding last year of birth) 10 % animals chosen based on the number of genotyped offspring, whereas for young animals (last year of birth) 10 % of the animals were chosen at random
^fOld10 is the sparse single-step, where the 10 % oldest animals were treated as core
^gYoung10 is the sparse single-step, where the 10 % youngest animals were treated as core
^hNormalA is where genotypes are ignored completely

ISSN: 1297-9686