. 2018 Oct 29;8(12):3903–3913. doi: 10.1534/g3.118.200710

Table 2. Average prediction accuracies (r) for four optimized subsets of TPs and full set across genomic prediction models.

Training Size (TP)	G-BLUP			Bayes-A			Bayes-B			Bayesian		Lasso
Training Size (TP)	CBSD3s	CBSD6s	CBSDRs	CBSD3s	CBSD6s	CBSDRs	CBSD3s	CBSD6s	CBSDRs	CBSD3s	CBSD6s	CBSDRs
TP100	0.27^ns	0.23^ns	−0.10^ns	0.26^ns	0.22^ns	−0.19^ns	0.30*	0.23^ns	−0.03^ns	0.33*	0.19^ns	−0.07^ns
TP200	0.27^ns	0.28^ns	−0.03^ns	0.26^ns	0.26^ns	−0.29^ns	0.27^ns	0.26^ns	0.07^ns	0.34*	0.22^ns	0.06^ns
TP400	0.32*	0.19^ns	−0.01^ns	0.32*	0.18^ns	−0.19^ns	0.32*	0.17^ns	−0.09^ns	0.36*	0.14^ns	−0.08^ns
TP800	0.31*	0.26^ns	0.06^ns	0.29^ns	0.25^ns	−0.13^ns	0.29^ns	0.23^ns	−0.04^ns	0.31*	0.17^ns	−0.01^ns
TP922	0.30*	0.25^ns	0.05^ns	0.24^ns	0.21^ns	0.11^ns	0.30*	0.26^ns	−0.09^ns	0.31*	0.15^ns	−0.04^ns

CBSD3s = Cassava brown streak disease severity scored at three months, CBSD6s = Cassava brown streak disease severity scored at six months, CBSDRs = Cassava brown streak disease root severity scored at 12 months; TP100, TP200, TP400, TP800 and TP922 = Optimized training populations of size 100, 200, 400, 800 and a full set of 922 clones, ns = non-significant prediction accuracies (r), * accuracy significantly different from zero (P ≤ 0.05).