Abstract
Adverse drug reaction (ADR) is a big challenge in drug development process. The US Food and Drug Administration (FDA) collects a lot of ADR report cases. The first of goal of this thesis is that we develop a tool which integrates drug and drug target knowledge bases and builds up drug target networks whose topological analysis can reveal drug interaction complexity for every ADR report in FDA. We classify drugs using Anatomical Therapeutic Chemical Classification (ATC) code and drug targets using Drug Target Ontology (DTO). This tool can help not only the analysis and prediction on ADRs cases but also the drug target assessment in the early drug discovery process. We could employ this system grouping similar cases to improve the performances of statistical method or machine learning techniques which analyze ADR cases. Many studies reported that ADR-related proteins (ADRRPs) could cause ADRs. Studies on the molecular mechanisms of ADRs reported that some drug targets are ADR-related proteins (ADRRPs). The second goal of this thesis is to use Support Vector Machine method to classify the adverse drug reaction related proteins from drug targets. Our approach is to find out the features that are significant enough to classify those proteins. This method can help not only the analysis and prediction on ADRRPs from drug targets but also the evaluation on safety of drug targets in the early drug discovery process.