Table 4.

Offset error metrics of binding affinity predictions for the separate OAH and OAMe datasets

Method	OAH				OAMe
	RMSEo	${\mathrm{R}}_{\mathrm{o}}^2$	mo	τ o	RMSEo	${\mathrm{R}}_{\mathrm{o}}^2$	mo	τ o
Null1a	1.7 (1.6 ± 0.6)	–	–	–	1.1 (1.1 ± 0.3)	–	–	–
Null2b	1.7 (1.9 ± 0.5)	0.4 (0.4 ± 0.3)	0.8 (0.4 ± 4.5)	0.4 (0.3 ± 0.3)	3.1 (3.1 ± 0.8)	0.7 (0.6 ± 0.2)	−1.5 (−1.6 ± 1.0)	−0.8 (−0.6 ± 0.2)
APR-OPC	0.7 (0.7 ± 0.2)	0.9 (0.7 ± 0.3)	1.1 (1.2 ± 2.7)	0.6 (0.5 ± 0.3)	1.6 (1.6 ± 0.6)	0.9 (0.9 ± 0.2)	2.2 (1.9 ± 0.7)	0.9 (0.7 ± 0.2)
APR-TIP3P	0.8 (0.8 ± 0.2)	1.0 (0.9 ± 0.2)	1.4 (1.5 ± 0.9)	0.6 (0.5 ± 0.3)	0.9 (0.9 ± 0.3)	0.9 (0.9 ± 0.1)	1.6 (1.5 ± 0.4)	0.9 (0.7 ± 0.2)
BEDAM	4.8 (4.7 ± 0.9)	0.0 (0.1 ± 0.2)	−0.5 (0.2 ± 18.5)	−0.1 (0.0 ± 0.3)	3.9 (3.8 ± 0.7)	0.4 (0.5 ± 0.3)	2.7 (2.1 ± 2.4)	0.3 (0.3 ± 0.4)
DFT/TPSS-cc	5.7 (5.6 ± 1.3)	0.2 (0.4 ± 0.3)	1.7 (3.7 ± 9.1)	0.5 (0.4 ± 0.3)	4.8 (4.7 ± 0.8)	0.7 (0.7 ± 0.2)	4.1 (3.9 ± 2.4)	0.3 (0.3 ± 0.4)
DFT/TPSS-nc	4.4 (4.3 ± 0.7)	0.5 (0.6 ± 0.3)	2.2 (2.0 ± 5.8)	0.3 (0.3 ± 0.4)	6.5 (6.4 ± 1.3)	0.4 (0.5 ± 0.3)	3.9 (2.7 ± 4.3)	0.3 (0.3 ± 0.4)
DLPNO-CCSD(T)c	7.0 (6.9 ± 1.3)	0.5 (0.5 ± 0.3)	3.3 (1.8 ± 9.9)	0.3 (0.3 ± 0.4)	7.6 (7.4 ± 1.7)	0.3 (0.4 ± 0.3)	3.7 (2.2 ± 5.4)	0.2 (0.2 ± 0.4)
Metadynamics	0.7 (0.9 ± 0.3)	0.9 (0.7 ± 0.3)	1.2 (1.1 ± 1.9)	0.6 (0.4 ± 0.3)	1.2 (1.6 ± 0.5)	0.3 (0.3 ± 0.3)	0.6 (0.8 ± 1.1)	0.3 (0.3 ± 0.3)
MMPBSA-GAFF	1.8 (1.8 ± 0.3)	0.5 (0.5 ± 0.3)	1.0 (1.3 ± 5.4)	0.5 (0.4 ± 0.4)	3.9 (3.7 ± 1.2)	0.4 (0.5 ± 0.3)	−1.7 (−1.2 ± 1.9)	−0.3 (−0.3 ± 0.4)
MovTyp-1d	0.8 (0.7 ± 0.2)	0.8 (0.7 ± 0.3)	1.0 (0.9 ± 2.4)	0.5 (0.4 ± 0.4)	3.1 (2.9 ± 1.0)	0.7 (0.6 ± 0.3)	−1.5 (−1.3 ± 0.9)	−0.5 (−0.4 ± 0.4)
MovTyp-2d	1.2 (1.2 ± 0.2)	0.8 (0.7 ± 0.3)	1.2 (1.4 ± 4.2)	0.9 (0.8 ± 0.2)	3.5 (3.4 ± 1.0)	0.5 (0.5 ± 0.3)	−1.7 (−1.5 ± 1.3)	−0.5 (−0.4 ± 0.4)
TI-ps	0.8 (1.3 ± 0.4)	0.8 (0.7 ± 0.2)	1.0 (1.6 ± 3.6)	0.7 (0.5 ± 0.3)	3.0 (3.1 ± 1.1)	0.2 (0.3 ± 0.3)	−0.8 (−0.4 ± 1.8)	−0.2 (−0.1 ± 0.4)
TI-raw	1.7 (1.9 ± 0.5)	0.5 (0.5 ± 0.3)	1.0 (1.7 ± 4.6)	0.6 (0.4 ± 0.3)	2.7 (2.8 ± 1.1)	0.1(0.4 ± 0.3)	−0.6 (−0.2 ± 1.7)	−0.2 (−0.1 ± 0.4)
HBAR	1.7 (1.7 ± 0.5)	0.2 (0.4 ± 0.3)	0.3 (0.7 ± 4.8)	0.5 (0.4 ± 0.3)	3.0 (2.9 ± 1.0)	0.2 (0.4 ± 0.3)	−0.9 (−0.4 ± 1.5)	−0.1 (−0.1 ± 0.4)
HBAR-ps	1.3 (1.3 ± 0.3)	0.4 (0.6 ± 0.2)	0.4 (0.8 ± 2.2)	0.6 (0.5 ± 0.3)	3.0 (2.9 ± 1.0)	0.2 (0.4 ± 0.3)	−0.9 (−0.4 ± 1.5)	−0.1 (−0.1 ± 0.4)
HBAR-ps1	1.2 (1.2 ± 0.4)	0.5 (0.6 ± 0.2)	0.5 (1.0 ± 2.0)	0.6 (0.6 ± 0.3)	3.2 (3.2 ± 1.0)	0.2 (0.4 ± 0.3)	−1.1 (−0.6 ± 1.6)	−0.2 (−0.2 ± 0.4)
HBAR-ps2	1.5 (1.5 ± 0.4)	0.3 (0.4 ± 0.2)	0.3 (0.7 ± 1.8)	0.5 (0.4 ± 0.3)	1.9 (2.0 ± 0.5)	0.0 (0.3 ± 0.3)	−0.1 (0.2 ± 1.1)	−0.1 (0.0 ± 0.4)
SOMD-1	0.8 (1.0 ± 0.4)	0.9 (0.7 ± 0.3)	1.3 (1.1 ± 1.6)	0.5 (0.4 ± 0.4)	0.7 (0.8 ± 0.3)	0.8 (0.7 ± 0.3)	1.1 (1.0 ± 0.5)	0.5 (0.5 ± 0.3)
SOMD-2	0.9 (1.0 ± 0.3)	0.9 (0.8 ± 0.3)	1.4 (1.1 ± 1.6)	0.5 (0.4 ± 0.4)	0.7 (0.8 ± 0.3)	0.8 (0.7 ± 0.3)	1.1 (1.0 ± 0.5)	0.6 (0.5 ± 0.3)
SOMD-3	1.0 (1.1 ± 0.3)	0.9 (0.7 ± 0.3)	1.3 (1.0 ± 1.7)	0.3 (0.3 ± 0.4)	0.7 (0.8 ± 0.3)	0.8 (0.7 ± 0.3)	1.1 (1.0 ± 0.5)	0.6 (0.5 ± 0.3)
SOMD-4	9.5 (9.5 ± 1.5)	0.4 (0.4 ± 0.3)	3.6 (3.5 ± 30.8)	0.5 (0.4 ± 0.3)	10.0 (9.9 ± 1.5)	0.0 (0.3 ± 0.3)	−1.4 (−0.5 ± 7.1)	0.1 (0.1 ± 0.4)
TI/BAR	1.0 (1.2 ± 0.4)	0.6 (0.6 ± 0.2)	0.6 (1.1 ± 2.5)	0.9 (0.6 ± 0.3)	–	–	–	–
MMPBSA-OPLS	–	–	–	–	3.4 (3.2 ± 1.1)	0.7 (0.6 ± 0.3)	−1.7 (−1.4 ± 1.0)	−0.6 (−0.4 ± 0.3)

The values before the parentheses are error metrics computed from the original data offset by the MSE. The mean values and uncertainties inside the parentheses were obtained through resampling the original data using replacement. The offset error metrics RMSE_o, ${\mathrm{R}}_{\mathrm{o}}^2$, m_o and τ_o are provided

^aNull1 model with a constant value 0.0 kcal/mol and no computational uncertainties

^bNull2 model based on the linear regression equation ΔG = −1.11 × number of heavy atoms + 5.06 kcal/mol (based on SAMPL4 OA data); SEM was set to 1.0 kcal/mol

^cComputational uncertainties were not reported

^dComputational uncertainties were not reported