Research Related to Data Warehousing, Database Management, Data Mining and Use of Information

The integration of diverse databases, from EMR to experimental phenotype and genotype databases, reveals several avenues of research.

The Federation of Existing Databases

The federation of existing databases, a central BIC task, will suggest new research problems in database management. While management of individual databases is a well-studied problem, an integration problem of this scale presents new challenges in terms of redundancy, inconsistency, uncertainty and security. For example, one question is how to deal with inevitable redundancy and inconsistency of patient information across different clinical databases in quantifying uncertainty of queries over an integrated database.

An Integrated Database

An integrated database will enable new research in knowledge discovery. The development of methods for discovering patterns and knowledge buried in large amounts of data (data mining) is a fast-growing research area in computer science. Rather than testing known hypotheses as in traditional research, data mining discovers possible hypotheses not known a priori but supported by patterns and correlations hidden in large databases. These hypotheses may provide insight to existing clinical results and spark new research directions, which would otherwise take years to discover.

An Integrated Database

The search for significant patterns located within high-dimensional clinical and experimental data is the focus of ongoing efforts in computer science. For example, the method has been applied to microarray datasets to identify co-regulated genes and thereby suggest the structure of gene networks. Another application is the discovery of correlations between phenotypic and genotypic data to identify genes linked to particular diseases. UK provides a unique opportunity to carry out such research due to its ability to collect both phenotypic and genotypic data across large populations in clinical trials.

Information Visualization

Computer Science and College of Engineering research strengths in information visualization will enhance the utility of the High-Performance Analytical Data Warehouse infrastructure. The scale and complexity of the biomedical informatics system challenge not only data management and analysis but also the presentation of information carried in the data. An effective information visualization approach will provide immediate comprehension of data to researchers and clinicians allowing them to interactively explore and link data across multiple databases.