The octanol/water partition coefficient, P (logP), is a hydrophobicity index and is one of the determining factors of the pharmacokinetics of chemical compounds. LogP values obtained from in silico software, open chemistry databases, and in vitro liquid chromatography retention factors may vary. Some chemicals (boscalid, etoxazole, and permethrin) have up to four-order-magnitude differences in in silico/in vitro P values. This study aimed to evaluate the effects of logP values of these three compounds, along with bisphenol A, 1,2-dibromobenzene, tetrabromobisphenol A, trazodone, and triazolam, on the input parameters and output plasma/hepatic concentration-time profiles of simple physiologically based pharmacokinetic (PBPK) models. Although the blood-to-plasma concentration ratios (~0.9-0.6) were slightly affected by variations in logP values, logarithmic plasma unbound fraction values and liver-to-plasma partition coefficients (Kp,h) were, respectively, inversely and linearly correlated with logP values (Kp,h was stable at ~6.7 for logP > 4). LogP was among the input parameters for previously established machine learning systems; consequently, the resulting logarithmic intrinsic clearance values were correlated with logP values in the range 2-8. However, the bioavailability, absorption rate constants, and volumes of distribution were not affected. PBPK-modeled plasma and hepatic maximum concentrations and areas under the concentration-time curves after virtual oral doses were mostly within ~0.5- to ~2-fold ranges, except for substances with low in vitro logP values, e.g., etoxazole and permethrin. These results suggest that in silico logP values are generally suitable for pharmacokinetic modeling; nevertheless, caution is needed for compounds with low in vitro logP values of ~2.
Keywords: input parameter; logP; output exposure; pharmacokinetic model.