Author: Mengshi Zhou

Publisher:

Published: 2020

Total Pages: 199

ISBN-13:

Neuropsychiatric disorders (NPDs) such as Alzheimer's disease (AD) lead to enormous societal burdens. Current treatments of NPDs have limitations, and the drug development process is at a standstill. In this dissertation, we proposed integrated computational approaches to facilitate NPDs drug discovery. First, we developed data-driven system approaches to identify FDA-approved drugs that may target NPDs-related genes. We developed a novel network-based model for drug-target interaction (DTI) prediction by preserving context-specific drug-side effect relationships. Our new network model improved DTI prediction comparing to the traditional similarity-based network model. Furthermore, we extended our model by modeling 855,904 phenotypic and genetic relationships among 24,600 biomedical entities and constructed a DTI prediction system (TargetPredict). Next, TargetPredict was used to identify the FDA-approved drugs that may target AD-associated genes. The AD drugs identified by TargetPredict were associated with lower risks of AD and dementia in electronic health record (EHR) data of 17 million patients over 65 years old. Second, we developed a phenome-driven computational drug repositioning approach to identify NPDs treatments without known NPDs-related genes. Our approach hypothesized that similar drugs treat the same diseases. We applied the approach to Drug addiction (DA), assuming that new treatments share similar phenotypes and common targets with drugs that cause or treat DA. Our method could prioritize FDA-approved and not-yet-approved DA drugs. The top-ranked DA drug candidates may play a beneficial role regarding remission from dependence in 326,340 opioid-dependent patients' EHR data. The pathway-enrichment analysis supported this clinical observation. Third, we performed an EHRs-based retrospective case-control study of 56 million adults (age ≥ 18 years) to study the relationships between tumor necrosis factor (TNF)-mediated systemic inflammation and AD, and the potential of using anti-TNF drugs as AD treatments. We found the co-morbid inflammatory disease involving TNF was associated with an increased risk of AD, and this could be mitigated from the treatment of a TNF blocking agent. In conclusion, our studies, including computational drug target prediction, drug repositioning, and retrospective clinical corroboration, can rapidly identify anti-NPDs drug candidates. Those drug candidates will allow biomedical researchers to conduct hypothesis-driven functional studies in experimental models for NPDs.