Fig. 4

Accuracies of prediction with ssGBLUP including only causative variants. Predictions with ssGBLUP with an unweighted GRM with causative QTN only and a regular inverse with 5% blending by pedigree relationships (only QTN/5% \({\mathbf{A}}_{22}\)), as only QTN/5% \({\mathbf{A}}_{22}\) but with 1% blending by pedigree relationships (only QTN/1% \({\mathbf{A}}_{22}\)), as only QTN/1% \({\mathbf{A}}_{22}\) but with inversion by APY with the number of core animals equal to twice the number of QTN (only QTN/1% \({\mathbf{A}}_{22}\)/APY), as only QTN/1% \({\mathbf{A}}_{22}\)/APY but with blending of the identity matrix by 1% (only QTN/1% I/APY). Predictions with GRM weighted by true QTN effects were used with 1% pedigree relationship blending (only QTN/TRUE/1% \({\mathbf{A}}_{22}\)) and 1% identity matrix blending (only QTN/TRUE/1% \({\mathbf{I}}\)). The number of causative QTN is 100 or 1000