Algorithm for Clustering Gene Expression Data with Outliers Using Minimum Spanning Tree

Microarrays enable biologists to study genome-wide patterns of gene expression in any given cell type at any given time and under any given set of conditions. Identifying group of genes that manifest similar expression pattern is important in the analysis of gene expression in time series data. In this paper multidimensional gene expression data is represented using Minimum Spanning Tree (MST). A key property of this representation is that each cluster of the expression data corresponds to one sub tree of the Minimum Spanning Tree, which converts a multidimensional clustering problem to a tree partitioning problem. Each node represents one gene, and every edge is associated with a certain level of pheromone intensity, densities and the co-expression level between two genes. MST-based clustering method is presented for finding cluster in gene expression time series data using new dissimilarity measure namely DMk. It is effective in classifying DNA sequences with similar biological characteristics and discovering the relationship among the sequences.