This review explores some of the techniques available to the computational chemist to maximise the success rate when attempting to identify pharmaceutically relevant ligands in a discovery environment. The authors introduce and discuss methods of constructing various types of compound collections and consider an alternative approach to measuring compound similarity. The discussion is illustrated with an example of a kinase collection constructed using these methods and the results obtained from it. As an example of predicting absorption, distribution, metabolism, excretion and toxicology liabilities, this review focuses on the CYP450 isozyme family. The field of high-throughput docking is touched on with further examples and discussion.