Table 2. Accuracy of the TM region, EL2 and ligand (LIG) after MD refinement based on the difference in RMSD compared to the initial model (ΔRMSD = RMSD_{MD refined}−RMSD_{Initial model}).

OPLS protocol
Model qualitya		Selection	ΔRMSD/Åb	ΔRMSDBest/Åc	ΔRMSDMin/Åd
All		LIG	0.54	–0.02	–0.90
		TM	0.71	0.51	0.01
		EL2	0.72	–0.06	–1.3
Good	0.0–2.5 Å	LIG	1.3	1.0	–0.34
	0.0–1.5 Å	TM	0.86	0.61	0.04
	0.0–4.0 Å	EL2	1.4	0.56	–0.65
Medium	2.5–5.0 Å	LIG	0.31	–0.43	–1.3
	1.5–2.5 Å	TM	0.52	0.33	–0.02
	4.0–5.0 Å	EL2	0.67	0.02	–1.5
Bad	> 5.0 Å	LIG	–1.5	–2.2	–2.7
	> 2.5 Å	TM	0.26	0.03	–0.28
	> 5.0 Å	EL2	0.19	–0.96	–2.1
CHARMM protocol
Model qualitya		Selection	ΔRMSD/Åb	ΔRMSDBest/Åc	ΔRMSDMin/Åd
All		LIG	–0.10	–0.63	–1.2
		TM	0.49	0.22	–0.14
		EL2	0.84	–0.02	–1.4
Good	0.0–2.5 Å	LIG	0.13	–0.27	–0.91
	0.0–1.5 Å	TM	0.53	0.34	–0.02
	0.0–4.0 Å	EL2	2.0	0.60	–0.45
Medium	2.5–5.0 Å	LIG	0.17	–0.60	–1.1
	1.5–2.5 Å	TM	0.37	–0.04	–0.26
	4.0–5.0 Å	EL2	0.80	–0.61	–1.9
Bad	> 5.0 Å	LIG	–1.6	–4.0	–5.8
	> 2.5 Å	TM	0.64	0.24	–0.14
	> 5.0 Å	EL2	–0.11	–1.7	–3.1

^a Based on the RMSD values of the GPCR Dock models.

^b Calculated based on the median of the centroids representing the five largest clusters from each MD refinement, i.e. the third best result of each MD refinement is used to calculate the ΔRMSD.

^c Calculated based on the best RMSD value obtained from the centroids representing the five largest clusters from each MD refinement.

^d Calculated based on the minimum RMSD identified in all 1500 snapshots generated for each model.