In Silico Prediction of Oral Bioavailability

Abstract

Oral dosing is the preferred route of administration for most drugs due to ease of administration and patient compliance. Orally delivered pharmacologically active compounds must have favorable absorption and clearance properties and satisfactory metabolic stability to provide adequate systemic exposure to elicit a pharmacodynamic response. Once ingested, the drug must undergo absorption, distribution, and hepatic metabolism prior to reaching the systemic circulation and becoming bioavailable (see 5.02 Clinical Pharmacokinetic Criteria for Drug Research). A bioavailable compound is then able to bind to target receptors and exert a pharmacological effect. Poor bioavailability limits oral administration of drugs, while high clearance limits desirable dosing regimens. <br/> In order to assess oral pharmacokinetic parameters such as bioavailability, it is necessary to administer the drug both intravenously and by the oral route. Bioavailability is then extracted from the combined intravenous and oral concentration-time data. Due to practical, ethical, legal, and safety constraints, pharmacokinetic characterization is often attended to late in the developmental process. Clearly, if pharmacokinctic input is to be available early in drug discovery and development, more viable predictive methodologies are necessary. <br/> The quantitative structure-activity relationship (QSAR) approach developed last century has recently been adapted to construct quantitative structure-pharmacokinetic relationships (QSPkRs). In addition to exploring physicochemical parameters including partitioning, solubilitv, and charge characteristics, increasing attention is being paid to the extraction of useful information from the molecular structure of a compound for QSPkRs. <br/> QSPkR modeling of bioavailability must rake into consideration structural influences on absorption and metabolism of a compound, as well as the variabilitv inherent in human/biological systems. Various computational techniques have been developed, some of which utilize experimental data for model generation while others rely on theoretically derived parameters including those that relate to molecular physicochemical properties. In silico methods depend heavily on the nature and quality of data used to construct models. Model validity, therefore, is influenced by issues of data variability, diversity, and experimental quality. These concepts and the theory behind them will now be discussed in relation to the current status of in silico bioavailabilitv modeling.