Novel Integrative Bioinformatics Approaches to Biomedical Ontology Practice for Translational Informatics.

Abstract

This study presents novel ontology-based approaches to solve biomedical problems raised from basic research to clinical data analysis. An ontology-based method that integrates basic and clinical data representation and analyses is critical for advanced translational informatics. This thesis demonstrates these capabilities in two case studies: 1) the Cell Line Ontology (CLO), and 2) the Ontology of Adverse Events (OAE). The Cell Line Knowledgebase (CLKB) was first created to analyze its nomenclature for data integration. CLO was built on CLKB based on the Open Biomedical Ontologies (OBO) Foundry principles. Cell lines in CLO are associated with terms from other ontologies such as Cell Type Ontology, and NCBI Taxonomy. Currently CLO contains over 36,000 cell lines. The OWL-based CLO is machine-readable and supports automated reasoning. CLO structure also allows for information storage of genetic signature that assists the process of cell line authentication. CLO consortium is initiated with collaborations of the European Bioinformatics Institute, The Bioassay Ontology group at the University of Miami, and the Riken Cell Bank in Japan. OAE was developed to support adverse event (AE) data standardization and analysis. OAE was used to analyze the CDC Vaccine AE reports (VAERS) of trivalent inactivated influenza vaccination from 1990-2011 (TIV) and trivalent live attenuated influenza vaccination (LAIV) from 2003-2011. 37,621 and 3,707 AE reports were associated with TIVs and LAIV respectively. Background noise filtering, the Proportional Reporting Ratio, and Chi-square analyses identified 48 TIV-enriched and 68 LAIV-enriched AEs as statistically significant. These AEs were then classified with OAE. TIV-enriched AEs were clustered in neurological and muscular processing, while LAIV-enriched AEs was associated with inflammatory responses and respiratory system disorders. Higher reporting rate of severe AEs (i.e. Guillain-Barre Syndrome (GBS), and paralysis) were identified with TIV. Post-vaccination GBS reports were examined to conclude the lower risk of severe AEs in LAIV recipients. Potential genes interaction networks regulating the enriched AEs were modeled with a natural language processing method run on PubMed abstracts. In total, 130 and 223 genes were found to interact with TIV and LAIV respectively. New potential hypotheses have been generated to elucidate the mechanism of the GBS formation.