Reliability Assessment of Microarray Data Using Fuzzy Classification Methods: A Comparative Study

Microarrays have become the tool of choice for the global analysis of gene expression. Powerful data acquisition systems are now available to produce massive amounts of genetic data. However, the resultant data consists of thousands of points that are error-prone, which in turn results in erroneous biological conclusions. In this paper, a comparative study of the performance of fuzzy clustering algorithms i.e. Fuzzy C-Means, Fuzzy C-medoid, Gustafson and Kessel, Gath Geva classification, Fuzzy Possibilistic C-Means and Kernel based Fuzzy C-Means is carried out to separate microarray data into reliable and unreliable signal intensity populations. The performance criteria used in the evaluation of the classification algorithm deal with reliability, complexity and agreement rate with that of Normal Mixture Modeling. It is shown that Kernel Fuzzy C-Means classification algorithms appear to be highly sensitive to the selection of the values of the kernel parameters.