Evolutionary Computation for Gene Expression Analysis

DESCRIPTION (provided by applicant): The proposed research seeks to develop software that can classify cancers reliably and accurately using gene expression profiling. Published literature has indicated the potential for methods that utilize artificial neural networks for gene expression diagnostic classification and their importance to the research and health communities. The current effort will train and design multilayer feed forward and artificial neural networks using evolutionary optimization techniques for cancer cell class prediction and discovery as well as understanding of the dynamics of gene interaction in a cell. These methods for training and designing neural networks will provide an effective computer-based automatic system that can be used for gene expression profiling. Success in this research will provide the basis for developing a prototype neural network classification and profiling system in Phase II as a commercial application. This will increase the efficiency and efficacy of methods used for screening and diagnosing cancers and increase the likelihood of successful treatment. In Phase II, the prototype system will be made user-friendly so that microarray analysts will be able to interpret the results of the neural network with ease and have confidence in its performance. The commercial potential for such a system is significant especially with regard to cell classification of any kind from molecular data.