Building A Biomedical Ontology On Breast Cancer

Abstract

Purpose: This thesis presents the purpose, needs, approaches and results of developing ontology on the breast cancer domain using UMLS as data source for improving discovery of medical knowledge. In the domain of biomedical research findings having vital knowledge are buried in publications in the form of unstructured texts. In the domain of cancer research, use of excessive technologies resulted in an eruption of knowledge related to cancers and their treatment. The research process is hampered by our inability to share technologies and data that have been developed by different organizations. This motivates representing information in semantic way by finding and storage of relational data into machine legible form. So, the prime objective of this research work is to contribute to the resolution of these problems by developing ontology on the most entities and relations containing amass of formalized background knowledge suitable for supporting reasoning in breast cancer domain. Methods: We undertook a study on existing domain ontologies, ontology implementation methods and end user interfaces for addressing the interdisciplinary nature of similar research. The study results were also extended with the publications to figure out the need for developing such an ontology using UMLS as a data source. We performed a thorough study of methodologies and evaluation approaches for designing ontology and data extraction approaches from UMLS. Results: In this research endeavour, we presented a summary of existing state of art for ontology on breast cancer. Our ontology contains integrated entities from well-organized source UMLS (Unified Medical Language System) and it is disseminated by NLM (National Library of Medicine). UMLS contains more than 170 biomedical dictionaries. The relations and facts amassed in the ontology are presented in a live web-browser interface http://115.167.72.12/ontology/ for end users and researchers. At present the ontology comprises facts on relations and semantic types for 106 categories of breast cancer including 14 main types and 92 subtypes. In addition to this the ontology contains 101 categories of broader relations with 27092 records and 254 categories of more specific relations covered with 10250 records, inclusive of publications from bio-medical giants.