CIS: A nonparametric clustering algorithm for gene expression data

The rapid development of the microarray technology brings about a great challenge to conventional clustering methods. The sparsity of data (sample), the high dimensionality of feature (gene) space and many irrelevant or redundant features all make it difficult to find correct clusters in gene expression data by using conventional clustering methods directly. In this paper, we present CIS, an algorithm for clustering biological samples using gene expression microarray data. Different from other approaches, CIS iterates between two processes, reclustering genes (not filtering genes) and clustering samples. Inspired by the policy of refining progressively, CIS repeatedly partition the set of initial genes with the new-generated sample clusters as features, and then partition samples with the new-generated gene clusters as features over again to identify the significant sample clusters and relevant genes. The method is applied to two gene microarray data sets, on colon cancer and leukemia. The experiment result show that CIS works well on both two datasets. We partition the two sample sets by eight and twentynine genes respectively, thus both acquire the accuracy about 90%. All these indicate that the CIS might be a promising approach for gene expression data analysis when domain knowledge is absent.