Figure 1:

Published drug repositioning strategies described in this review are organized according to their primary mode of inference. The dashed arrows connect high-level informational aspects of drugs and diseases with the methods that incorporate these types of information in their approach. Methods are generally categorized as focusing largely on either (i) ‘Direct’ inference, where established or directly measured biomolecular or chemical properties are used to infer therapeutic relationships between drugs and diseases, (ii) ‘Indirect’ inference, where related or higher level data or representations of drugs and diseases is used to infer therapeutic relationships between drugs and diseases or (iii) ‘Simulation’, where therapeutic interactions are inferred through simulation of interactions between drugs and diseases rather than through direct or indirect measurement of their salient properties. We predict that newer methods will move toward integrating multiple forms of therapeutic inference incorporating many forms of both drug- and disease-based data and knowledge to enable the discovery of new uses for drugs—as some of the methods described in this review have taken steps toward.