Biclustering of Microarray Data Employing Multiobjective GA

Abstract

In genetic research, microarray technology is rapidly growing and gaining importance because of its capacity of measuring multiple genes simultaneously. Biclustering of microarray data is an efficient data mining technique to gain knowledge regarding the functional behaviour of multiple genes under a set of experimental states. In this work, sequential GA using weighted sum approach is first implemented to derive good quality biclusters. The multiple objective functions are mapped to single objective function using weighted sum approach; however, the primary challenge lies in deriving the accurate weight values. To overcome this drawback of sequential GA, NSGA-II is adopted for solving multiobjective optimization problem. To further improve the performance of NSGA-II, an adaptive feature is incorporated in NSGA-II. All the three approaches were experimented on yeast Saccharomyces Cerevisiae data set and efficiency of individual approaches are discussed.