Extraction of informative genes from microarray data

Topon Kumar Paul, Hitoshi Iba
2005 Proceedings of the 2005 conference on Genetic and evolutionary computation - GECCO '05  
Identification of those genes that might anticipate the clinical behavior of different types of cancers is challenging due to availability of a smaller number of patient samples compared to huge number of genes, and the noisy nature of microarray data. After selection of some good genes based on signal-to-noise ratio, unsupervised learning like clustering and supervised learning like k-nearest neighbor (kNN) classifier are widely used in cancer researches to correlate the pathological behavior
more » ... f cancers with the gene expression levels' differences in cancerous and normal tissues. By applying adaptive searches like Probabilistic Model Building Genetic Algorithm (PMBGA), it may be possible to get a smaller size gene subset that would classify patient samples more accurately than the above methods. In this paper, we propose a new PMBGA based method to extract informative genes from microarray data using Support Vector Machine (SVM) as a classifier. We apply our method to three microarray data sets and present the experimental results. Our method with SVM obtains encouraging results on those data sets as compared with the rank based method using kNN as a classifier.
doi:10.1145/1068009.1068081 dblp:conf/gecco/PaulI05 fatcat:tcu6v6oacjajne6bjb5yr5rdom