Drug-likeness rules

Drug development would be a lot more successful and much less costly if it were simple to tell which molecules would behave like drugs. In fact, given the number of drugs on the market, it would seem like it should be easy to design new compounds that behave like drugs. But even though the search for drug-likeness predictability has gone on for at least 100 years, medicinal chemistry remains as much an art as a science.

The gold standard is Lipinski's Rule of 5, which was proposed in 1995 (see "Rule of 5," A2). The promulgation of the rule sparked renewed interest in drug-likeness models, which have proceeded along two paths: the physicochemical properties of molecules and computer models based on data from biological assays. While both have a long way to go, the combination of increased volumes of biological and chemical data, and more powerful computing capabilities, should result in continued improvement in models...