Molecular properties that influence oral drug-like behavior

Pharmaceutical companies are constantly racing to discover the next therapeutic blockbuster. The consensus in the industry is to focus on compounds that are by some measure drug-like, but in order to do this effectively a number of questions must be answered. For example, how should drug-like be defined and how might this definition be used to enhance drug discovery? Has the field moved beyond Lipinski's seminal 'rule-of-five' observations? This review offers a working definition of oral drug-likeness, describes various approaches used in its characterization and discusses its appropriate use. We will focus primarily on the use of computed molecular properties that attempt to predict the oral drug-like behavior of compounds and propose guidelines for the use of observations and trends established in existing datasets to support drug discovery efforts. In particular, this review will demonstrate how trends in simple properties of the data can be used prospectively to compare and prioritize groups of compounds, chemical libraries and different chemical series with greater reliability than for predicting drug-likeness of single compounds. It is the authors' belief, however, that properties or descriptors that will completely separate drug-like space from non-drug-like space are unlikely to be found; the focus should instead lie on overall distributions of drug-like and non-drug-like compounds in the property space that tend to overlap significantly.