Date of Award
3-21-2013
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Computer Science
First Advisor
Xiaowei Xu
Abstract
The pharmaceutical industry has been experiencing diminishing returns in terms of newly approved drugs over a decade. Limited knowledge of target space and the safety issues that come along with newly discovered compounds have forced pharmaceutical companies, government agencies, and academia to seek more practical solutions. Since there have been several cases of repurposing, research interests have emerged for existing drugs whose pharmacokinetics properties are already known and investigation of their potential to treat other cases has become a new direction in pharmacology. Recently, several in silico methodologies have been developed for drug repositioning. In this study, the same purpose has been addressed from a graphical model perspective to blend all phenotypic information rather than a single criterion to create more comprehensive knowledge. In the proposed framework, a Latent Dirichlet Allocation (LDA) model, which is basically a Bayesian network, was constructed based on the observed outcomes of each drug. The model benefits the side and therapeutic effects of drugs, which were collected from the U.S. Food and Drug Administration (FDA) drug labels and Side Effect Resource (SIDER) respectively, by treating the molecular targets as latent variables. Resulting probabilistic associations provided a grouping of drugs according to their safety concerns and therapeutic catalogs. Furthermore, a new drug-drug similarity measure was defined based on the same associations by which a network analysis was performed and alternative drugs were suggested for some diseases. Lastly, therapeutic uses were predicted with some supportive evidence using the same probabilistic values. As the first attempt to focus on FDA drug labels for drug repositioning, this work indicates a promising direction for future studies which could shorten the drug discovery pipeline.
Recommended Citation
Bisgin, Halil, "A Graphical Approach to Drug Safety and Drug Repositioning" (2013). Theses and Dissertations. 419.
https://research.ualr.edu/etd/419
