Improved prediction of HIV-1 protease-inhibitor binding energies by molecular dynamics simulations

Table 2 All-atom root mean square deviation (RMSD) of the protein-ligand complexes relative to their corresponding x-ray structures.

PDB code	Complex all-atom RMSD (Å) (picosecond time scales)
	0.01	0.10	1.00	10.00
1gno	0.22	0.32	1.11	3.20
1hbv	0.16	0.35	0.95	2.84
1hef	0.16	0.34	0.90	2.71
1heg	0.16	0.32	0.91	2.80
1hih	0.18	0.35	0.90	2.93
1hiv	0.20	0.36	0.85	2.98
1hps	0.18	0.38	0.95	2.95
1hpv	0.17	0.33	0.92	2.96
1hvi	0.14	0.35	0.92	2.99
1hvj	0.17	0.33	0.91	3.06
1hvk	0.12	0.33	0.94	2.97
1hvl	0.21	0.33	0.91	2.93
1hvr	0.18	0.36	0.92	2.92
1hvs	0.24	0.35	0.92	2.88
1hte	0.20	0.37	0.90	2.89
1htf	0.10	0.32	0.93	3.06
1htg	0.14	0.31	0.76	2.86
1pro	0.26	0.38	0.94	2.94
1sbg	0.20	0.36	0.93	2.86
2upj	0.23	0.41	0.90	2.88
4phv	0.14	0.35	0.90	2.93
4hvp	0.20	0.35	0.93	2.95
5hvp	0.18	0.31	0.94	2.93
8hvp	0.18	0.35	0.92	2.90
9hvp	0.19	0.35	0.92	2.90
Complex (average)	0.18	0.35	0.92	2.92
Flap (average)	0.14	0.54	0.95	3.30

The correlations between experimentally-determined and calculated binding energy significantly improved after MD simulation, and were inversely influenced by the all-atom RMSD of the complex as well as the protease flap region (i.e., as the all-atom RMSD increased with longer MD simulations, the correlations were reduced from their peak). The best correlation of 0.87 was observed at 0.1 ps. Similar results were observed regardless of the starting seed used.