Editorial: Computational methods for drug repurposing
Philippe Sanséau, Jacob Koehler · Briefings in Bioinformatics · 2011
Repurposing marketed drugs or compounds in development for alternative indications is not new in the pharmaceutical industry. It is in fact quite common but it has been mostly done by serependity over the years. It is only recently that more systematic approaches based on computational analyses are being used. Faced with scientific and economical challenges the possibility of finding new indications for drugs is an attractive proposition for the industry. Drug repurposing could be applied at many phases of drug discovery and development but has a greater potential when the drug has already been tested for safety. Recently there has been a significant growth of publications for novel, more systematic and nonobvious approaches for computational drug discovery. In parallel some companies specialized in computational drug repurposing have also developed. This special issue of Briefing in Bioinformatics publishes a number of papers covering multiple aspects of computational drug repurposing. Butte et al. introduce the field and review recent developments in computational drug repurposing. Computational repurposing may be target based or disease based. Target based approaches infer similar binding sites by assessing the compound–protein interactions, whereas disease-based approaches associate drugs to new indications, by comparing the characteristics and similarities of different diseases. The use of chemical similarities is a common target-based approach for drug repurposing. In this case the structural similarity of binding sites of known and new targets is compared for inferring new indications. Schroeder et al. reviews tools and success stories that follow this approach. Swamidass describes a target-based approach, which analyses results from high-throughput screens in novel ways for identifying new drug–target interactions. This approach exploits public data sets of high-throughput screens of drug like small molecules for predicting new indications. Moriaud et al. describe an approach that analyzes protein structures and their binding sites to predict new proteins and off-target interactions for known compounds. The authors developed a new approach, MED-SuMo, that scans the entire protein surface and thus is less reliant on prior knowledge of protein docking sites. The authors validate their approach on a number of known molecules. Most, if not all approaches for computational drug repurposing are heavily reliant on different data sets. The Rare Disease Repurposing Database (RDBD), described by Cote et al. is a promising novel resource provided by the FDA. This database publishes a comprehensive list of over 200 known products that have the potential to be repurposed to treat rare diseases. This resource is rich in information, and offers sponsors a new tool for finding special opportunities to develop niche therapies for rare disease patients. The paper by Jegga et al. is also motivated by repurposing known drugs toward rare diseases. In this publication, the authors review and discuss the intrinsic difficulties of computational approaches for systematic drug repurposing. They provide an interesting discussion on intrinsic difficulties and issues in this field, and review a number of promising approaches that were recently published. Whereas most of the abovementioned computational drug repurposing methods rely on different types of structured data sets, Persidis et al. review how literature mining and ontology modeling can be used for extracting relevant data from free texts. Such data can subsequently be mined and visualized for identifying novel indications for existing drugs. This special issue on computational drug repurposing concludes with a Letter to the Editor by Nikodemus on the stability and ranking of predictors from random forest variable importance measures. It is an exciting time for computational drug repurposing with more molecular and textual data being available combined the current challenges faced by the pharmaceutical industry to deliver products of value for patients and payers. However, scientific, regulatory or legal challenges (such as patents) should not be underestimated. Most of the approaches described in the accompanying papers are very recent and adoption of these methods is still ongoing. We also expect methods to continue to evolve rapidly with possibly only a limited number of approaches becoming mainstream. Ultimately we expect computational drug repurposing to become more generalized although it is likely to still take a few more years before the true potential and impact of these methods materialize.