Maker Gene Identification: a Multiple Kernel Support Vector Machine Approach

Recently, gene expression profiling using DNA microarray technique has been shown as a promising tool to improve the diagnosis and treatment of cancer. Support vector machine has been successfully used to classify cancer tissue based on gene expression data. Besides performance, the ability to discover underlying principles will be a crucial point in the medical field. In this paper, we present a novel marker gene identification method based on multiple kernel support vector machine (MK-SVM). The main strength of this technique is the detection of gene groups that are strongly associated with specific types of cancer and maybe useful to the diagnosis and treatment. It achieves this by employing a two phases' framework. Firstly, a 1-norm based regularized cost function is used to enforce sparsity and obtain gene subset. Secondly, a support vectors based rule extraction algorithm is implemented to determine the final marker genes. The ALL-AML Leukemia dataset is used to demonstrate the promising performance of this approach.