Informative Gene Discovery for Cancer Classification from Microarray Expression Data

Gene expression data analysis from microarray is a new advance of cancer diagnosis. However, the gene expression data often have high dimensionality and small sample size. These properties cause severe difficulties in classification. Gene selection is thus a crucial pre-processing step to filter out uninformative genes prior to the classification step. Our approach to perform gene selection is an information theoretic approach combining with sequential forward floating search. Experimental results show that our method is capable of efficiently finding a compact set of informative genes which can effectively discriminate different classes.