TABLE 3.

Genomic relatedness matrices for different genomic prediction models for all markers or haplotypes.

Models	Description	Relatedness matrices	Use1
GBLUP	Genomic best linear unbiased prediction	$G={\displaystyle \frac{(M-P){(M-P)}^{\prime }}{2{\sum }_{i-1}^m{p}_i(1-p_i)}}$	All markers
GHBLUP	Haplotype based GBLUP	$G_H={\displaystyle \frac{M_H{M}_H^{\prime }}{Q_H}}$	All markers
GHBLUP|GA	Haplotype based GBLUP given genome annotation	$G_{{\mathrm{H}}_{\text{GA}}}=\frac{M_{{\mathrm{H}}_{\text{GA}}}{M}_{{\mathrm{H}}_{\text{GA}}}^{\prime }}{Q_{{\mathrm{H}}_{\text{GA}}}}$	Genomic classes
CM	Categorical marker effect model	$S=({\displaystyle \frac{{\sum }_{q-1}^Q{\mathrm{\phi }}_{jik}}{m}})ij$	All markers
CE	Categorical epistasis model	E = 0.5 × mS#(mS + 1n×n)/m2	All markers
CHM	Haplotype based CM	$S_H=({\displaystyle \frac{{\sum }_{q-1}^Q{\mathrm{\phi }}_{jiq}}{Q_H}})ij$	All markers
CHE	Haplotype based CE	$E_H={\displaystyle \frac{0.5\times Q_H{S}_H\mathrm{\#}(Q_H{S}_H+1_{n\times n})}{Q_H^2}}$	All markers
CHM|GA	CHM given genome annotation	$\tilde{S}=({\displaystyle \frac{{\sum }_{q-1}^Q{\mathrm{\phi }}_{jiq}}{Q_{{\mathrm{H}}_{\text{GA}}}}})ij$	Genomic classes
CHE|GA	CHE given genome annotation	$\tilde{E}=0.5\times Q\tilde{S}\mathrm{\#}(Q\tilde{S}+1_{n\times n})/Q_{{\mathrm{H}}_{\text{GA}}}^2$	Genomic classes

¹Refers to the whole genome-wide SNP; genomic classes refer to IGR, gene, exon, CDS, and UTR class.