A novel gene selection method using GA/SVM and fisher criteria in Alzheimer´s disease

Identification of those genes which cause diseases can develop the process of diagnosis and the treatment of diseases. In this paper, a gene selection method based on genetic algorithm (GA) and support vector machines (SVM) is presented. At first, Fisher criteria is utilized in order to do filtration for those genes which are noisy and redundant in high dimensional microarray data. Then, GA/SVM model is used for selection of various subsets of maximally informative genes with the use of different training sets. The frequency of appearance of each gene in various subsets of genes is analyzed. Therefore, the last subset contains those genes which are highly informative. In fact, Fisher and GA/SVM methods have been merged in order to take benefit from a filtering method as well as an embedded method. The proposed method is tested on DNA microarray gene expression data of Alzheimer´s disease. The results show that the proposed method has a good selection and classification performance, which can yield 100% classification accuracy using only 15 genes. From biological point of view, at least 8 (53%) of these genes are Alzheimer associated genes. Thus, these genes not only can serve as predictors of the disease, but also can use as a means to find new candidate genes.