Deriving meaningful rules from gene expression data for classification

摘要

Although a fuzzy rule based system offers interpretability, its application in gene expression data analysis becomes difficult due to the very high dimensional nature of the data. Here we propose an interesting scheme of combining fuzzy modeling with neural networks for designing fuzzy rule based classifiers for gene expression data analysis. A neural system is used for selecting a set of informative genes. Considering only these selected set of genes, we cluster the expression data with a fuzzy clustering algorithm. Each cluster is then converted into a fuzzy if-then rule, which models an area in the input space. These rules are tuned using a gradient descent technique to improve the classification performance. The rule base is tested on a leukemia data set containing two classes and it is found to produce good results. We propose some simple criteria to simplify membership functions and the rules. Our rule extraction scheme can be automated. Unlike other classifiers, it produces human interpretable rules which are not expected to give poor generalization because fuzzy rules do not respond to areas not represented by the training data. The last two properties are very important for problems like diagnosis of cancer.