Machine Learning Techniques and Chi-Square Feature Selection for Cancer Classification Using SAGE Gene Expression Profiles
Recently developed Serial Analysis of Gene Expression (SAGE) technology enables us to simultaneously quantify the expression levels of tens of thousands of genes in a population of cells. SAGE is better than Microarray in that SAGE can monitor both known and unknown genes while Microarray can only measure known genes. SAGE gene expression profiling based cancer classification is a better choice since cancers may be due to some unknown genes. Whereas a wide range of methods has been applied to traditional Microarray based cancer classification, relatively few studies have been done on SAGE based cancer classification. In our study we evaluate popular machine learning methods (SVM, Naive Bayes, Nearest Neighbor, C4.5 and RIPPER) for classifying cancers based on SAGE data. In order to deal with the high dimensional problem, we propose to use Chi-square for tag/gene selection. Both binary classification and multicategory classification are investigated. The experiments are based on two human SAGE datasets: brain and breast. The results show that SVM and Naive Bayes are the top-performing SAGE classifiers and that Chi-square based gene selection can improve the performance of all the five classifiers investigated.
KeywordsSupport Vector Machine Near Neighbor Machine Learn Technique Sequential Minimal Optimization Maximum Margin Hyperplane
Unable to display preview. Download preview PDF.
- 2.Ng, R.T., Sander, J., Sleumer, M.C.: Hierarchical Cluster Analysis of SAGE Data for Cancer Profiling. BIOKDD, 65–72 (2001) Google Scholar
- 4.Buckhaults, P., et al.: Identifying Tumor Origin Using a Gene Expression-based Classification Map. Cancer Res. 63, 4144–4149 (2003)Google Scholar
- 7.Statnikov, A., et al.: A Comprehensive Evaluation of Multicategory Classification Methods for Microarray Gene Expression Cancer Diagnosis. Bioinformatics Advance Access (September 16, 2004)Google Scholar
- 8.Golub, T., et al.: Molecular Classification of Cancer: Class Discovery and Class Prediction by Gene Expression Monitoring. Science 286 (October 15, 1999)Google Scholar
- 11.Cohen, W.W.: Fast Effective Rule Induction. Machine Learning. In: Proceedings of the Twelfth International Conference (ML 1995) (1995)Google Scholar
- 12.SAGEMap (2005), http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL4
- 13.Schneider, K.-M.: A Comparison of Event Models for Naive Bayes Anti-Spam E-Mail Filtering. In: Proceedings of the 10th Conference of the European Chapter of the Association for Computational Linguistics, Budapest, Hungary, pp. 307–314 (April 2003) Google Scholar
- 14.Statnikov, A., Aliferis, C.F., Tsamardinos, I., Hardin, D.P., Levy, S.: A Comprehensive Evaluation of Multicategory Classification Methods for Microarray Gene Expression Cancer Diagnosis. Bioinformatics (2004)Google Scholar
- 15.Androutsopoulos, I., et al.: Learning to Filter Spam E-mail: A Comparison of a Naive Bayesian and a Memory-based Approach. In: Proceedings of the Workshop on Machine Learning and Textual Information Access, pp. 1–13 (2000)Google Scholar
- 16.John, G.H., Langley, P.: Estimating Continuous Distributions in Bayesian Classifiers. In: Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence, pp. 338–345. Morgan Kaufmann, San Mateo (1995)Google Scholar
- 18.Witten, I., Frank, E.: Data Mining –Practical Machine Learning Tools and Techniques with Java Implementation. Morgan Kaufmann, San Francisco (2000)Google Scholar
- 19.Quinlan, R.: C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers, San Mateo (1993)Google Scholar