Table 4.

Median and standard deviation of prediction accuracy [r(ĝ, g)] of genomic prediction in PIN184 obtained with the prediction method GBLUP under the genetic model M2, for all SolCAP SNPs, with a reduced relationship between training and validation set across 100 cross-validation runs.

Clusters	Training set	Validation set	PIN184
	n	n	rAUDPC	MCR	PM	TSC
PREDICTION OF 50% OF THE GENOTYPES IN ONE CLUSTER BASED ON ALL OTHER CLUSTERS—F3A
4	116–154	15–34	0.35 ± 0.17	0.31 ± 0.23	0.34 ± 0.16	0.66 ± 0.13
8	147–177	3–18	0.32 ± 0.35	0.35 ± 0.46	0.47 ± 0.32	0.51 ± 0.33
PREDICTION OF 50% OF THE GENOTYPES IN ONE CLUSTER BASED ON ALL OTHER CLUSTERS
PLUS THE OTHER 50% FROM THE SAME CLUSTER—F3B
4	116–154	15–34	0.41 ± 0.19***	0.40 ± 0.30***	0.37 ± 0.16***	0.73 ± 0.17*
8	147–177	3–18	0.50 ± 0.47**	0.48 ± 0.55ns	0.47 ± 0.25ns	0.57 ± 0.30ns

The studied traits were relative area under disease progress curve (rAUDPC), plant maturity (PM), maturity corrected resistance (MCR), and tuber starch content (TSC). Prediction accuracies for scenario F3B marked by ^ns or ^*, ^**, ^*** are non significantly or significantly (P < 0.05, 0.01, 0.001) higher than those of the corresponding comparison in scenario F3A.