Identification Of Differentially Expressed Gene Sets Based Cluster

This paper briefly gives the background knowledge of microarray at first, and gives the concept of identification of differentially expressed genes, which can be divided into individual gene and gene set research and they are both focused by lots of researchers. Nowadays, more and more studies showed that the phenotype always relates with lots of genes, and there are relationships among genes expression, and it is common that the number of genes in the gene set is greater than the number of samples. So in this paper we address a method called GSCL bases the result of cluster to analysis gene set. The process of GSCL can be described for three parts. Firstly, we use cluster to divide the genes into several classes, which called gene subsets. Secondly, by defining the gene subsets information, we give the weight vector of these subsets. Thirdly, after quickly reducing the dimension by weight vector, we use t-test to identify whether the gene set is differently expressed or not. After applying GSCL to simulation data and experiment data (include ALL/AML and p 53 datasets), we have the result:the method appears perform well in terms of control of power under the models considered in the simulation; when we apply the method to ALL/AML, we have the same conclusion as the formal researchers; when we apply the method to p 53 dataset, and we also compare with other methods (two kind of sum of squared t-test, Hotelling's T:, N-statistics), find some common but also some complementing differentially expressed gene sets, which can be considered latterly in biological.