Translational bioinformatics for personalized medicine and integrative biology: Data integration, extraction, knowledge discovery, and visualization

This thesis focuses on developing a computational framework to support the Precision Medicine Initiative. The newly developed tools and algorithms use machine learning, text mining and visualization techniques for extracting salient information from heterogeneous sources such as scientific literature, clinical text, and –omics technologies to enhance clinical decision making and improve the quality of healthcare. Various advances in biomedical technologies have enhanced our ability to study disease processes at different molecular levels (genes, metabolites, histones, etc.). Similarly, technological advances in healthcare domain such as adoption of Electronic Health Record systems (EHRs) provide us a unique opportunity to develop a learning healthcare system where intelligent tools and algorithms can be utilized to extract information from clinical notes, patient medication records, laboratory results, etc. for early detection of medical risks and prevention of adverse drug events. The key novel contributions of this thesis are: a) development of novel full-text summarization algorithms that have been incorporated into a web application (CoReViz) for visualizing clinically relevant information and extracting relevant sentences from clinical text and scientific articles; b) development of novel association mining algorithms and graph summarization techniques incorporated into a web application (SEACOIN2.0) for interactive drill-down summarization and hypothesis generation to extend the functionality of PubMed; c) introduction of the concept of literature based Phenotype-wide Association Studies (Lit-PheWAS); d) development of an ensemble feature selection framework for biomarker discovery using agent-based modeling and stochastic optimization techniques.Ph.D