Project Summary A drug-drug interaction (DDI) occurs when one drug influences the level or activity of another drug. There is a big risk of DDI in Alzheimer’s disease (AD) patients. However, information about potential DDIs occur in AD patients is still lacking in clinical databases. In this proposal, we will develop and evaluate efficient computational models that can identify potential DDIs from the records in FDA Adverse Event Reporting System (FAERS), the largest drug post-marketing surveillance database in the world. Approximately 9 million ADE records from FAERS (from 2004 to 2019) will be analyzed. Several data mining algorithms, multi-item Gamma Poisson shrinkage (MGPS), Bayesian Confidence Propagation Neural Network (BCPNN), and association rule will be applied to detect DDI signals from these reports. The effects of some confounding factors (such as age, race, gender, and dosage) on DDIs will be also investigated using multivariate logistic regression. The DDIs identified by the computational model will be validated through a retrospective analysis of electronic health records (EHRs) of AD patients at the Byrd Alzheimer's Institute at the University of South Florida (USF). In addition, a web server, AD_DDI, will be developed to provide public access to the prediction model and results. The successful completion of this project will provide useful information for doctors or pharmacists to prescribe drugs for AD patients more appropriately.

Project Narrative In this proposal, we will develop and evaluate efficient computational models that can identify possible DDIs in Alzheimer’s disease (AD) patients from 9 million records of FDA Adverse Event Reporting System (FAERS). An online web server, AD-DDI, will be developed to provide public access for automatic retrieving, analyzing and displaying the possible DDIs identified from FAERS. The successful completion of this project will provide useful information for clinicians to prescribe drugs for AD patients appropriately.