Prediction Model For The Fluoroquinolone Pharacokinetics Based On Physicochemical Properties

Pharmacokinetic prediction can provide drug researchers with evidence for betterdecision-making during the drug discovery and development. As one of the three mostwidely used antibiotics agents, development of new fluoroquinolones is always a hotissue. This paper is to build the prediction models for the fluoroquinolonepharmacokinetics based on physicochemical properties, with the purpose to guild itsnew drug development.In this paper, pharmacokinetic data of 21 fluoroquinolones were widely selectedfrom Phase I study literatures, as well as their physicochemical properties both fromliteratures and specific software calculation. Multiple linear regression was then use tobuild the pharmacokinetic models for plasma protein-binding (PPB), bioavailability(BA), peak time (tmax), biological half-time (t1/2), peak concentration (Cmax), areaunder concentration-time curve extrapolated to infinity (AUC∞), apparent volume ofdistribution (Vd/F) and apparent total clearance(CL/F), respectively. During the modelbuilding, matrix scatter plot, curve estimation and correlations were employed tofigure out the significant physicochemical properties. Meanwhile, statistical andresidual plots were used for the hypothesis verification. Finally, Leave-One-Out wasapplied to validate the model prediction capability.The AUC∞model was successfully built in this paper withAUC∞=360.138+0.132MV-84.032pKa2+4.541(pKa2)2,R2=0.906. Correlation betweenpredicted and experimental values was 0.952 and the prediction capability was morethan 80% validated by Leave-One-Out.The model may be useful for a quick AUC∞prediction for new fluoroquinolonesin the early stage of drug development based on MV and pKa2. This can prevent drugswith poor pharmacokinetics being selected into the clinical studies, so as to save timeand effort, reduce risk and eventually promote the new drug development.