An efficient gene selection method for high-dimensional microarray data based on sparse logistic regression

Gene selection in high-dimensional microarray data has become increasingly important in cancer classification. The high dimensionality of microarray data makes the application of many expert classifier systems difficult. To simultaneously perform gene selection and estimate the gene coefficients in the model, sparse logistic regression using L1-norm was successfully applied in high-dimensional microarray data. However, when there are high correlation among genes, L1-norm cannot perform effectively. To address this issue, an efficient sparse logistic regression (ESLR) is proposed. Extensive applications using high-dimensional gene expression data show that our proposed method can successfully select the highly correlated genes. Furthermore, ESLR is compared with other three methods and exhibits competitive performance in both classification accuracy and Youdens index. Thus, we can conclude that ESLR has significant impact in sparse logistic regression method and could be used in the field of high-dimensional microarray data cancer classification.