Various Partial Charge Schemes on 3D-QSAR Models for P-gp Inhibiting Adamantyl Derivatives

Abstract

We developed three-dimensional quantitative structure activity relationship (3D-QASR) models for 17 adamantyl derivatives as P-glycoprotein (P-gp) inhibitors. Eighteen different partial charge calculation methods were tested to check the feasibility of the 3D-QSAR models. Best predictive comparative molecular field analysis (CoMFA) model was obtained with the Austin Model 1-Bond Charge Correction (AM1-BCC) atomic charge. The 3D-QSAR models were derived with CoMFA and comparative molecular similarity indices analysis (CoMSIA). The final CoMFA model (<TEX>$q^2$</TEX> = 0.764, <TEX>$r^2$</TEX> = 0.988) was calculated with an AM1-BCC charge and electrostatic parameter, whereas the CoMSIA model (<TEX>$q^2$</TEX> = 0.655, <TEX>$r^2$</TEX> = 0.964) was derived with an AM1-BCC charge and combined steric, electrostatic, hydrophobic and HB-acceptor parameters. Leave-five-out (LFO) cross-validation was also performed, which yielded good correlation coefficient for both CoMFA (0.801) and CoMSIA (0.656) models. Robustness of the developed models was checked further with 1000 run bootstrapping analyses, which gave an acceptable correlation coefficient for CoMFA (BS-<TEX>$r^2$</TEX> = 0.997, BS-SD = 0.003) and CoMSIA (BS-<TEX>$r^2$</TEX> = 0.996, BS-SD = 0.018).