A Statistical Network Pharmacology Approach for Early Detection of Adverse Drug Events Project Summary Early detection of adverse drug events (ADEs) in the post market phase is essential for protecting the public from significant morbidity and mortality. The broad, long-term objectives of this project are to develop tools and techniques that enable scientists to discover ADEs earlier and more reliably. Post-market spontaneous reporting system (SRS) of ADEs serves as a cornerstone of pharmacovigilance and a series of drug safety signal detection methods play an important role in providing drug safety insights. However, existing methods are developed to generate safety signals for drugs with enough reports in SRS, but few methods can be used to generate signals for newly approved drugs with few or even no safety reports in SRS. Also, few methods formulate the signal detection problem under a rigorous hypothesis test framework, and no method exploits drug label and drug property information. The goal of this project is to develop novel statistical learning methods to tackle those challenges and detect ADEs in an early and actionable manner: I. Develop an integrative label propagation framework to re-rank drug safety signals based on multiple drug similarity networks for early detection of ADEs. We hypothesize that ADEs of newly approved drugs can be detected in a timely fashion by incorporating multiple drug similarity networks. We will compute original drug safety signals via common signal detection algorithms. Then, we will construct drug similarity networks based on multiple data sources (e.g., chemical structures, targets, indications). Finally, we will generate enhanced drug safety signals by propagating original signals on multiple drug similarity networks. The proposed method enriches SRS with multiple drug similarity networks, alleviating issues of insufficient cases for newly approved drugs and paving the way for early detection of ADEs. II. Develop a Bayesian hypothesis testing approach which facilitates early detection of ADEs, while controlling the false positive rate. We hypothesize that the proposed approach has increased power to detect ADEs comparing with frequentist approaches, especially when the sample size is small (i.e., in a short period after a new drug’s approval). Additionally, we hypothesize that the proposed approach is able to control the false positive rate. Specifically, prior distribution of the new drug’s ADE risk can be estimated by using the existing drugs’ risks, similarity scores between the new drug and the existing drugs, and drugs’ label information. Subsequently, the prior distribution and the observed data can be utilized to derive the posterior probability of the null hypothesis, which shall be used to detect ADE in a timely fashion.