Percentage change method for microarray gene expression based classification of glioma

Abstract

With the invention of microarray technology there is considerable improvement in the survival rate of cancer patients. Microarray gene expression data has a very large dimension and small sample size. In this paper we have suggested number of methods to reduce the size of microarray data. In the first method, we propose to threshold and calculate the percentage change in gene expression values followed by feature extraction using discrete wavelet transform and classification using neural network. In second method we propose to calculate the mean and standard deviation of percentage change of every cancer sample and use them as two inputs to the neural network for classification. In the third method, we propose to perform the classification using two gene expression values having maximum percentage change.