Finding Teams in Graphs and Its Application to Spatial Gene Cluster Discovery

摘要

Analysis of the Diversity and Dynamics of Genomes", Bielefeld University, Bielefeld, Germany Gene clusters are sets of genes in a genome with associated functionality. Often, they exhibit close proximity to each other on the chromosome which can be beneficial for their common regulation. A popular strategy for finding gene clusters is to exploit the close proximity by identifying sets of genes that are consistently close to each other on their respective chromosomal sequences across several related species. Yet, even more than gene proximity on linear DNA sequences, the spatial conformation of chromosomes may provide a pivotal indicator for common regulation and/or associated function of sets of genes. We present the first gene cluster model capable of handling spatial data. Our model extends a popular computational model for gene cluster prediction, called S-teams, from sequences to general graphs. In doing so, δ-teams are single-linkage clusters of a set of shared vertices between two or more undirected weighted graphs such that the largest link in the cluster does not exceed a given threshold 5 in any input graph. We apply our model to human and mouse data to find spatial gene clusters, i.e., gene sets with functional associations that exhibit close neighborhood in the spatial conformation of the chromosome across species.