FCM for gene expression bioinformatics data

Abstract

Clustering analysis of data from DNA microarray hybridization studies is essential for evaluating and identifying biologically significant co-expressed genes. The K-means algorithm is one of the most widely used clustering technique. It attempts to solve the clustering problem by assigning each gene to a single cluster. However, in practice especially in case of Bioinformatics data, one gene can be found in many clusters simultaneously. To sort out this problem, Fuzzy C-means (FCM) clustering algorithm is applied to microarray data. Two pattern recognition data (IRIS and WBCD data) and thirteen microarray data is used to evaluate performance of K-means and Fuzzy C-means. Improvement of approx. 30 percent clustering accuracy is achieved in case of FCM compared to K-means algorithm. Extensive simulation results shows that the FCM clustering algorithm was able to provide the highest accuracy and generalization results compared to K-means clustering algorithm. © 2009 Springer Berlin Heidelberg.