March – 2003 ^April

Aggregation of data in relational systems is the critical factor that limits query performance in business intelligence and data warehousing. Today it is solved by purely relational aggregation mechanisms. However, relational tools suffer from performance and optimization problems, stemming from the very nature of the relational data model. Relational tools are purely two-dimensional (relations), while aggregated data need a multi-dimensional handling. A novel approach will be presented. It is based on transparent integration of multidimensional, non-relational algorithms, into relational environment, improving aggregation performance by over two levels of magnitude.

Current genomic research has generated an immense volume of data and a tremendous increase in the number of gene-related publications. This wealth of information presents a major data analysis challenge. The ultimate goal is to understand the complex biological interrelationships among all discovered genes and proteins. Meeting this goal requires scanning the abundant literature about each gene and plenty of human expertise. As several research groups have recently noted, automated systems for extracting relevant information from the literature can complement existing techniques, speed up analysis, and greatly enhance our understanding of genetic processes.

We present a new approach, based on probabilistic information retrieval, which uses the literature to establish functional relationships among genes on a genome-wide scale. Experiments applied to documents discussing yeast genes, and a comparison of the results to well-established gene function, demonstrate the effectiveness of our approach.