Linear interaction energy approximation for binding affinities of nevirapine and HEPT analogues with HIV-1 reverse transcriptase

Abstract

The binding affinities and binding mechanism of nevirapine and HEPT analogues with HIV-1 reverse transcriptase (HIV-1RT) have been studied in this paper with a recently proposed linear interaction energy method based on a surface generalised Born continuum solvent model (LIE–SGB). Various LIE–SGB fitting schemes are presented and discussed on this relatively large binding set, consisting of a total of more than 50 ligands. For a training subset of 40 ligands (20 nevirapine and 20 HEPT analogues), the LIE–SGB method gives an RMS error (RMSE) of 0.89 kcal/mol with a correlation coefficient of 0.74. The leave-one-out cross validation results also show an encouraging RMSE of 1.00 kcal/mol, with a correlation coefficient of 0.69. The further blind tests on seven mostly high-potent candidates (not originally in the training subset) also show reasonable accuracies. The binding mechanism of this HIV-1RT binding set is found to be mainly driven by van der Waals interactions (i.e. a good geometrical fit is important) and the net loss of ligand cavity energies (i.e. the burial of solvent accessible surface area is favourable). The net electrostatic interactions, however, are found to be anti-binding. A secondary amide indicator is found to be necessary for nevirapine analogues to account for the deficiencies in quantum partial charges, and it is also shown that it represents a critical -type hydrogen bond between the secondary amide fragment of nevirapine analogues and the phenyl ring of Tyr188A residue, which explains their otherwise surprising binding affinities. More quantum charge fittings, with the protein environment included, gave similar results, indicating that an out-of-plane polarisability might be needed to fully capture this -type hydrogen bond in classical force fields. Finally, six new ligands are designed for optimal binding based on our predictions for further experimental validations.