Research Project
6070441
DESCRIPTION: (Applicant's Description) In order to better understand the vast quantity of microarray data currently being generated in a number of laboratories, new computational and analysis tools are needed. Pattern analysis of thousands of simultaneously expressed genes will provide insight into novel and progressive disease treatments. With the use of information technology tools, substantial opportunities exist for improving the ability to identify genetic anomalies. These tools will be critical in advancing the automation and interpretation of experimental results. We will develop an innovative computer-based analytical tool called MicroExplore that will bring together multiple methods for analysis of microarray data. Each method has unique properties that produce conceptually different results. In the first phase, we propose to compare conceptual clustering, hierarchical agglomerative and k-means algorithms. Evaluation of the quality of data will be assessed through a combination of factors that measure how well-known functional groupings are reflected in the output. Efficiency and scalability will be measured through timed runs of MicroExplore on a set of expression datasets of various sizes. Our study will focus on the complex dataset of lymphoid gene expression belonging to an intramural NCI laboratory. In the second phase, a novel hybrid approach will be designed to utilize the results of multiple clustering algorithms and configurations to achieve a ranked set of best overall clusters. Attempts to integrate external data sources will add new dimensions to the analysis tool and provide for powerful predictions. Improving the scientist's ability to accurately and efficiently identify which candidate genes would make good therapeutics will be a fundamental step for the advancement of cancer research and scientific discovery. MicroExplore will be a valuable resource for the continuance of gene discovery and characterization. This platform will have strong potential for the enhancement of clinical data analysis helping to characterize or profile the molecular changes found in normal, precancerous and malignant tissue samples.
