Characterization of Type II Diabetes through biomedical data fusion
Advisor Information
Hesham Ali
Location
UNO Criss Library, Room 112
Presentation Type
Oral Presentation
Start Date
6-3-2015 1:00 PM
End Date
6-3-2015 1:15 PM
Abstract
Recent advancements in biomedical engineering has resulted in an impressive collection of biomedical instruments. This, in turn, has led to the ability to generate incredible amount of “big” biomedical data with huge volume and high levels of heterogeneity and veracity. The lack of advanced data integrating and analytics tools has been limiting our ability to take full-advantage of the massive amount of raw data currently available. This shortcomings is particularly exemplified by the naïve integration procedures implemented in protein-protein interaction (PPI) networks. Traditionally, PPI networks are created from a single source, usually literature mining. We propose a data fusion approach to integrate multiple sources through data fusion techniques. Our proposed approach is based on the need for an in-depth understanding of the data sources and interrelationship between the biological elements abstained from each source, in addition to increasing the specificity in which the PPI networks are constructed. Toward that goal, we compare the knowledge overlap and enrichment between protein-protein databases, microarray data, and gene-ontology relationships. In analyzing the results, the most prominent theme discovered was the exclusivity that each source contributed, even among the various PPI databases. We then implemented a data fusion algorithm which incorporates the inherent domain knowledge within each source. The obtained results show that the networks obtained from the proposed algorithm carry more biological significant knowledge and have higher signal to noise ratios.
Characterization of Type II Diabetes through biomedical data fusion
UNO Criss Library, Room 112
Recent advancements in biomedical engineering has resulted in an impressive collection of biomedical instruments. This, in turn, has led to the ability to generate incredible amount of “big” biomedical data with huge volume and high levels of heterogeneity and veracity. The lack of advanced data integrating and analytics tools has been limiting our ability to take full-advantage of the massive amount of raw data currently available. This shortcomings is particularly exemplified by the naïve integration procedures implemented in protein-protein interaction (PPI) networks. Traditionally, PPI networks are created from a single source, usually literature mining. We propose a data fusion approach to integrate multiple sources through data fusion techniques. Our proposed approach is based on the need for an in-depth understanding of the data sources and interrelationship between the biological elements abstained from each source, in addition to increasing the specificity in which the PPI networks are constructed. Toward that goal, we compare the knowledge overlap and enrichment between protein-protein databases, microarray data, and gene-ontology relationships. In analyzing the results, the most prominent theme discovered was the exclusivity that each source contributed, even among the various PPI databases. We then implemented a data fusion algorithm which incorporates the inherent domain knowledge within each source. The obtained results show that the networks obtained from the proposed algorithm carry more biological significant knowledge and have higher signal to noise ratios.