During drug discovery, it is important to optimise the affinity for the biomolecular target and also the properties of molecules that influence absorption, distribution, metabolism, excretion, and toxicity (ADMET). The goal is to improve the properties of a lead compound and select highly ‘developable’ candidates. Efficient pharmaceutical property profiling operations are now run in parallel to potency screening during lead optimization and provide data for compound prioritization and for the selection of compounds for in vivo studies. The main components of a profiling strategy suitable for early discovery are the measurement of selected physicochemical properties together with in vitro screening for metabolic stability. In addition, enzyme or cellular assays may be deployed for investigation of cellular permeability (including active and passive transport), plasma protein binding, toxicity, and the potential for drug-drug interactions. The four principal physicochemical parameters measured are lipophilicity, dissociation constant, permeability through artificial membranes, and aqueous solubility. An objective of property-based design is the identification of structure property relationships for ADMET that can suggest structural modifications that will also promote or retain high affinity for the biomolecular target. Sometimes this is not possible and new lead molecules may need to be identified to provide new starting points. A sub-optimal in vitro profile may be acceptable as long as there is a reasonable probability of achieving an adequate in vivo profile for clinical studies. In these cases, the application of composite models that can relate in vitro to in vivo behavior is important. Full ‘physiologically based pharmacokinetic’ (PBPK) models can be used; however, there are also simpler approaches available that may be adequate and more easily applied for ranking compounds within a series. Reliable in vitro—in vivo correlation is still very difficult because of the multiplicity of mechanisms involved. For the future, there is a need to develop better criteria for making mechanistically based classifications to develop differentiated models that are appropriate for different types of compound. It is also difficult to correct models for non-specific binding of drugs to membranes and proteins and we need better lipophilicity measures for accurately estimating the affinity of drugs for tissues. PBPK modeling leads to the exciting concept of the ‘virtual human’; however, PBPK models suitable for drug discovery applications are still in their infancy and it will be some time before their promise is fulfilled.