Network-based approach to prediction and population-based validation of in silico drug repurposing
By Feixiong Cheng, Rishi J. Desai, Diane E. Handy, Ruisheng Wang, Sebastian Schneeweiss, Albert-László Barabási & Joseph Loscalzo
Although investment in biomedical and pharmaceutical research and development has increased significantly over the past 20 years, the annual number of new treatments approved by the US Food and Drug Administration (FDA) has not significantly increased1. Among the reasons for this shortcoming in contemporary drug development are a lack of well-established predictive pharmacokinetics/pharmacodynamics approaches, and concerning safety and tolerability profiles for new chemical entities from preclinical studies to clinical trials2. In addition to these well recognized explanations, another important factor limiting more effective drug development may be continued adherence to the classical (one gene, one drug, one disease) hypothesis. Focusing on just single targets results in failure to anticipate off-target toxicity, unintended beneficial effects, or multiple target interactions leading to suboptimal efficacy3,4. Without full knowledge of the broader network context of the molecular determinants of disease and drug targets in the protein–protein interaction network (human interactome), investigators cannot develop meaningful approaches for efficacious treatment of complex diseases5.
Novel approaches, such as network-based drug-disease proximity, that shed light on the relationship between drugs (drug targets) and diseases [molecular (protein) determinants in disease modules]6,7,8 can serve as a useful tool for efficient screening of potentially new indications for approved drugs with well-established pharmacokinetics/pharmacodynamics, safety and tolerability profiles, or previously unidentified adverse events9,10,11,12. However, in order to prioritize the repurposed candidates or suggest novel interventions based on drug-disease associations identified by network-based approaches, rigorous validation is mandatory. Since network-based drug repurposing focuses on drugs that are already approved and are used in clinical practice, such hypothesis testing is possible using large-scale patient-level data collected during routine healthcare. Such data are regularly used to generate actionable evidence regarding effectiveness, harm, use, and value of medications to supplement evidence generated in randomized controlled trials; these trials that lead to drug approval are typically limited in scope owing to a relatively modest study sample size, comparatively short follow-up time, and frequent underrepresentation of the most relevant populations13. The unique strengths of routine healthcare data that make them ideal for validating hypotheses generated by network-based predictions include their provision of large patient populations useful for detecting small differences, and the availability of a large number of patient factors recorded without any recall bias, including demographics, comorbid conditions, and medication use, that allow for high-dimensional covariate adjustment to minimize confounding14,15,16.
In this study, we developed a systems pharmacology-based platform that quantifies the interplay between disease proteins and drug targets in the human protein–protein interactome with state-of-the-art pharmacoepidemiologic methods for hypothesis validation using longitudinal data with over 220 million patients. We followed this analysis with in vitro assays to test potential drug mechanisms. As proof of the utility of the overall approach, we focused on cardiovascular (CV) outcomes given their high prevalence in the population, as an exemplary set of diseases with which to identify associations between drugs used for non-cardiac indications and CV outcomes. We demonstrate that an integrated approach incorporating network proximity together with large-scale patient longitudinal data and in vitro experimental assays offers an effective platform by which to identify and validate novel associations that can be used to minimize unanticipated adverse drug effects and optimize drug repurposing. These results suggest that this integrative approach can be generalized to other drugs/disease combinations.
