A decade of drug-likeness

Abstract

Applications of the concept of drug-likeness, which involve consideration of how properties of chemicals such as lipophilicity compare with those of approved drugs, have become widely accepted as a way to reduce attrition in the past decade. However, a recent analysis indicates that this acceptance is not completely reflected in the properties of potential drugs synthesized in recent years, perhaps increasing the likelihood of future failures.

Main

In the early 1990s, two complementary technologies — combinatorial chemistry and high-throughput screening (HTS) — were widely seen as having great potential to revolutionize drug discovery. Combinatorial chemistry would provide access to unprecedented numbers of small molecules, while HTS would allow these molecules to be rapidly screened against a burgeoning number of novel targets emerging from genomics.

Initially, however, the quality of the output from these technologies was much more limited than hoped. For example, at Pfizer in the mid-1990s, many compounds emerging from the company's efforts in HTS and combinatorial chemistry were proving unsuitable for further optimization into potential drug candidates owing to issues such as poor solubility and permeability. Seeking to understand why, Christopher Lipinski, a medicinal chemist, and colleagues investigated the properties of a large database of drugs and late-stage clinical candidates, reasoning that compounds with poor physicochemical properties would have been largely 'weeded out' by this stage in development. The resultant analysis, published a decade ago¹, introduced a set of guidelines — dubbed the 'rule of five' — for predicting the likelihood that a compound might have a high risk of poor absorption or permeability on the basis of simple physicochemical properties of the compound.

As originally presented, “the rule of 5 predicts that poor absorption or permeation is more likely when there are more than 5 hydrogen-bond donors, 10 hydrogen-bond acceptors, the molecular weight (MWT) is greater than 500 and the calculated Log P (CLogP) [a measure of lipophilicity] is greater than 5”¹. And importantly from the point of view of understanding why combinatorial chemistry and HTS weren't performing as hoped: “High throughput screening (HTS) leads tend to have higher MWT and Log P and lower turbidimetric solubility than leads in the pre-HTS era”¹.

The impact of the paper (which has now been cited more than 1,000 times) has been dramatic. As awareness grew of the importance of the concept of drug-likeness embodied by the rule-of-five guidelines, pharmaceutical companies completely re-evaluated their compound libraries, often discarding many of the molecules synthesized in the early years of combinatorial chemistry, and carefully monitored subsequent additions to the libraries. The resultant success in improving the quality of compound libraries, and the ease of application of the rule of five, have led to it becoming arguably the most popular filter included in computer programs to aid library design.

The paper has also stimulated considerable research related to the concept of drug-likeness, with various types of analysis providing support for the importance of key physicochemical properties when seeking to discover orally available drugs. For example, one analysis of approved drugs over two decades revealed that mean values of lipophilicity have not changed over time, indicating that such properties should be particularly controlled in drug discovery programmes².

Following in this vein, this month we feature our first Analysis article (see page 881), which are intended to draw novel and important conclusions from analysis of existing data. In this paper, using a database assembled from existing data on approved drugs and information on compounds published in recent patent applications from four major pharmaceutical companies, the authors examine the impact that the concept of drug-likeness has had on recent medicinal chemistry, as reflected in the patent applications.

Surprisingly, given the widespread acceptance of the concept of drug-likeness by medicinal chemists, the analysis indicates that the compounds synthesized more recently are larger and more lipophilic than historical oral drugs. And of particular concern could be the observed increases in lipophilicity, as this property has a key role in promoting binding to undesired targets, suggesting that the risk of attrition owing to side effects could also be increasing as a result. It therefore seems that after a decade in which the concept of drug-likeness has become widely accepted, the control of physicochemical properties of potential drugs — in particular, lipophilicity — still merits urgent attention to help ensure that this acceptance is translated more fully into practice.