G. Gottipati, J. Venitz; Virginia Commonwealth University, Richmond, VA

BACKGROUND: The aim was to develop and validate QSPKR models for predicting biologically relevant human in vivo PK properties based on molecular and physicochemical properties (PC) for AAR.
METHODS: In vivo systemic PK properties, i.e., CLtot, Vdss and fe following IV administration of AAR to healthy humans were compiled from the literature and corrected for fu (from in vitro plasma protein binding studies) to obtain biologically relevant PK variables, namely Vdssu, CLtotu, CLrenu and CLnonrenu. Pertinent PC properties were obtained from SciFinder Scholar. Univariate relationships were explored (r2 ≥ 0.3, p < 0.05) after log-transformation (except fu) by linear regression (LR), and final QSPKR model building was performed by stepwise multiple LR in JMP Pro 10.0.1. Cross-validation (CV) was done using the Leave-Out-One method in RStudio v0.96.330; CV-explained variance (q2) of ≥ 0.4 was considered acceptable.
RESULTS: The final PK database consisted of seven AAR. LogD7.4 - which was highly correlated with molecular weight, MW - showed a negative association with fu (r2 = 0.83, n = 7) and positive associations with Vdssu (r2 = 0.97, n = 7), CLtotu (r2 = 0.71, n = 7) and CLnonrenu (r2 = 0.79, n = 7). Final, validated QSPKR models gave acceptable predictions (q2 = 0.45 - 0.94) for fu, Vdssu and CLnonrenu.
CONCLUSION: Lipophilicity was found to be the major determinant affecting the systemic disposition of AAR - suggesting the important role of nonspecific hydrophobic interactions with plasma proteins, biological membranes, extravascular tissues and drug metabolizing enzymes. However, since logD7.4 was found to be highly correlated with MW, their individual contributions cannot be distinguished.