Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
Creating new drugs is extremely difficult, and very expensive1. There are an estimated 1060 drug-like molecules in chemical space — the theoretical space spanned by all possible molecules and compounds2. However, only relatively few will bind to a particular target site in the body and thus be medically useful. Therefore, identifying a molecule that acts on a specific target and is also suitable for use as a drug (for example, in terms of how it travels to its target and is broken down) is a highly complex task. Writing in Nature, Sadybekov et al.3 present an innovative approach to identifying potential drug molecules from exceptionally large chemical libraries, and demonstrate its performance in a screen for inhibitors of three target proteins.