The Multi-Node Topological Overlap Measure For Gene Neighborhood Analysis

Defining the neighborhood of an initial set of nodes is an important task in network analysis. For example, we show that the neighborhood of an initial set of brain cancer related genes is highly enriched with other cancer genes as well. It remains an active area of research to define a biologically meaningful concept of neighborhood in gene or protein networks. In gene networks, genes with high topological overlap have been found to have an increased chance of being part of the same biological pathway. Since our main interest lies in gene and protein networks, we propose a generalization of the topological overlap matrix to define the neighborhood of a set of genes. A pair of nodes in a network is said to have high topological overlap if they are both strongly connected to the same group of nodes. We generalize the standard pairwise topological overlap measure to multiple nodes. The resulting neighborhoods are comprised of tightly interconnected nodes. We provide empirical evidence that a neighborhood surrounding an initial set of 2 nodes can be far more informative than the neighborhood of a single node. Using a cancerand a yeast network applications, we provide empirical evidence that the multiTOM approach yields biologically meaningful results and compares favorably to alternative approaches. Our approach is implemented in the freely available multiTOM software package.