Table 3:

Analytically derived and simulated metrics for different designs with different $MS{E}_{\mathrm{\Delta }\mathrm{\Delta }{G}^{FEP}}$ variability and ∆G^True distributions.

		a∆GTrue from standard normal
		$\mathit{M}\mathit{S}{\mathit{E}}_{\mathit{\Delta }\mathit{\Delta }{\mathit{G}}^{\mathit{F}\mathit{E}\mathit{P}}}=\mathbf{1.0}$			$\mathit{M}\mathit{S}{\mathit{E}}_{\mathit{\Delta }\mathit{\Delta }{\mathit{G}}^{\mathit{F}\mathit{E}\mathit{P}}}=\mathbf{2.25}$
Pairs	Design	$\mathit{M}\mathit{S}{\mathit{E}}_{\stackrel{\wedge }{\mathit{\Delta }\mathit{G}}}$c	Median ρd	Accuracye	$\mathit{M}\mathit{S}{\mathit{E}}_{\stackrel{\wedge }{\mathit{\Delta }\mathit{G}}}$	Median ρ	Accuracy
20	Random	4.02	0.60	30.5%	9.00	0.46	20.4%
	D-optimal	3.83	0.60	31.5%	8.61	0.46	22.8%
	A-optimal	1.42	0.70	35.3%	3.19	0.55	24.6%
30	Random	1.66	0.77	40.9%	3.73	0.63	27.8%
	D-Optimal	1.36	0.81	49.2%	3.12	0.68	34.6%
	A-Optimal	1.08	0.79	45.3%	2.43	0.66	34.3%
50	Random	0.98	0.87	55.0%	2.21	0.77	43.2%
	D-Optimal	0.92	0.89	58.3%	2.09	0.79	47.2%
	A-Optimal	0.79	0.88	57.5%	1.78	0.78	43.9%
		b ∆GTrue from uniform between −12.0 and −6.5
20	Random	4.02	0.77	28.1%	9.00	0.65	22.6%
	D-optimal	3.83	0.78	30.2%	8.61	0.65	23.7%
	A-optimal	1.42	0.85	32.9%	3.19	0.74	24.8%
30	Random	1.66	0.89	38.0%	3.73	0.80	30.2%
	D-Optimal	1.36	0.91	41.7%	3.12	0.84	32.3%
	A-Optimal	1.08	0.91	41.1%	2.43	0.83	30.7%
50	Random	0.98	0.94	49.0%	2.21	0.89	37.8%
	D-Optimal	0.92	0.95	49.4%	2.09	0.91	39.3%
	A-Optimal	0.79	0.94	49.0%	1.78	0.90	39.4%

^a∆G^True values were generated from a normal distribution with µ = 0 and σ = 1

^b∆G^True values were generated from a continuous uniform distribution between −12.0 and −6.5.

^c$MS{E}_{\stackrel{\wedge }{\mathrm{\Delta }G}}$ was analytically derived from equation (3) for optimal designs.

^dSpearman’s rank correlation (ρ) between the ∆G^True and estimated $\stackrel{\wedge }{\mathrm{\Delta }G}$ values of 20 ligands.

^eProbability of correctly identifying the ligand with the lowest absolute free energy.