Algorithms Based on Density and Shared Neighbors for Functional Modules Identification in PPI Networks

Abstract

Accumulating evidence suggests that biological systems are composed of interacting, separable, functional modules-groups of vertices within which connections are dense but between which they are sparse. Identifying these modules is essential to understand the organization of biological systems. However, the most existing deterministic algorithms only find ldquodenserdquo clusters. Actually, the modules are of differing sizes, densities and shapes. In this paper, we take into account the property of diversity of module topological structure, propose an efficient algorithm relying on density and shared neighbors for detecting overlapping modules in PPI (protein-protein interaction) networks. Our algorithm first finds the skeleton of the modules, SNCS(Shared Neighbor Connected Subgraph), then constructs the modules by expanding the leaf vertices of SNCS based on shared neighbors. Furthermore, since the PPI network is noisy and still incomplete, some methods treat the PPI networks as weighted graphs in which each edge (e.g., interaction) is associated with a weight representing the probability or reliability of that interaction for preprocessing and purifying PPI data. Thus, we extend our method into weighted networks which takes into account the link weights in a more delicate way by incorporating the subgraph intensity. We test our method on PPI networks. Our analysis of the yeast PPI network suggests that most of these modules have well biological significance in the context of protein localization, function annotation.