The development of DNA microarray technology has facilitated in-depth research into cancer classification, and has made it possible to process thousands of genes simultaneously. Since not all genes are crucial for classifying cancer, it is necessary to select informative genes which are associated with cancer. Many gene selection methods have been investigated, but none is perfect. In this paper we investigate methods of finding optimal informative genes for classification of gene expression profiles. We propose a new gene selection method based on the forward selection method with regression analysis in order to find informative genes which predict cancer. The genes selected by this method tend to have information about the cancer that does not overlap with the other genes selected. We have measured the sensitivity, specificity, and recognition rate of the selected genes with the $k$-nearest-neighbour classifier for the colon cancer dataset and the lymphoma dataset. In most cases, the proposed method produces better results than gene selection based on other feature selection methods, yielding a high accuracy of 90.3% for the colon cancer dataset and 72% for the lymphoma dataset.
All Science Journal Classification (ASJC) codes
- Computer Science Applications
- Computational Theory and Mathematics
- Applied Mathematics