Diagnosis of acute myeloid and lymphoblastic leukemia using gene selection of microarray data and data mining algorithm

Microarray technology represents the expression of thousands of genes simultaneously. Microarray analysis may not be possible without statistical analysis and artificial intelligence methods. The aim of this paper is to diagnose acute leukemia using microarray data and data mining algorithms. The expression of 7129 genes of 72 patients with leukemia was used in this study. Then, by the selection of important genes based on correlation coefficient, information gain, gain ratio and fisher score criteria and by the use of linear discriminat, support vector machine, k nearest neighbor, naïve Bayes, Bayes net, nearest mean, logistic regression, multilayer perceptron neural network and J48 decision tree methods on the selected genes, acute myeloid and lymphoblastic leukemia were attemted to be diagnosed. The methods of nearest mean, support vector machine, k nearest neighbor, naïve Bayes, and multilayer perceptron neural network are able to detect acute myeloid and lymphoblastic leukemia using 39 selected genes by the gain ratio with 100 percent accuracy. Moreover, support vector machine method using 87 selected genes by information gain and support vector machine method using 133 selected genes by information gain are able to detect acute myeloid and lymphoblastic leukemia with 100 percent accuracy. The results of this study showed that gene selection and data mining algorithm are able to diagnose leukemia with high accuracy. Therefore, appropriate decisions can be made using these methods about the how of the diagnosis and treatment of patients.