Profiling of Disease-Associated Proteins Neighborhood Networks

Abstract

Discovering disease-associated proteins in the human Protein-Protein Interaction (PPI) network is an important problem. Several algorithms have been proposed to discover these proteins from an initial set of known disease-associated proteins. A majority of these algorithms depend on the connectedness of disease-associated proteins in the PPI network. In this paper, We performed an experimental study to understand the interaction between disease-associated proteins and their neighbors. We observe that most disease-associated proteins have common neighbors in the PPI network. We performed comparative analysis on the interaction of disease proteins in disease-associated proteins neighborhood networks and networks consisting of only disease proteins. Our experimental results show that for all disease that we considered in this study, have a higher number of disease proteins in the largest connected component in the disease-associated proteins neighborhood network. We develop a score metric to capture the reachability of disease-associated proteins via common neighborhood proteins that not necessarily associated with a disease. We observe that a high number of disease proteins are part of the largest connected component in the neighborhood network with fewer disease proteins are disconnected forming separately connected components. we also develop a reachability metric to profile diseases based on the connectedness of associated proteins based on the number of components and the density of disease proteins in a disease-associated protein neighborhood network.