Gene Selection Based On SVM

The advent of DNA micro-array make it possible to perform gene diagnosis and gene treatment. Gene selection is one of the major challenge of gene-chip technology , for gene diagnosis where only a gene subset is enough for diagnosis of diseases, for resolution of curse of dimensionality which occurs especially in DNA microarray dataset where there are more than thousands of genes and only a few number of experiments(sample). This paper studied the method of gene selection and four parts work are studied as follows:(l) present a gene selection method by training linear SVM(Support Vector machine) classifier and testing them with cross validation for finding gene subset which is optimal/suboptimal for diagnosis of binary disease classes. Genes are selected with linear SVM classifier incrementally for the diagnosis of binary disease. (2) present a gene selection method for multi-class disease by training SVM/MLP classifier with method-l,we get many gene subset ,then the union of them is used as initialized gene subset for diagnosis of all related disease classes.(3)present a gene selection method for multi-class disease based on Combined vector. Based the methods above, get many contribute vector, the combined vector of them is used to selected gene subset for diagnosis.(4)present a gene selection method based on Contribute Space. Construct a contribute Space with many contribute vector got as in method 3.Genes are selected in Contribute Space. For real DNA microarray data with 2308 genes and only 64 labeled samples belonging to 4 disease classes, only 6 genes are selected to be diagnostic genes; with 7129 genes and only 72 labeled samples belonging to 2 disease classes, only 11 genes are selected.