Organizational principles of the Reactome human BioPAX model using graph theory methods

We investigate the Reactome human BioPAX model, in terms of the various node and edge types, i.e. protein, complex and reaction nodes, activation and inhibition edges. We uncover topological features of such large-scale heterogeneous biological networks using graph theory methods, calculating various centrality distributions, network assortativity and partitioning using k-core decomposition and modularity. The network is characterized by a fat tail power law degree distribution and a centralized organization of proteins that follow a steep power law PageRank distribution. We highlight the differences in individual distributions for each node type and calculate statistically significant shifts from the original distribution. We also discover a power law scaling of the clustering coefficient, with a steeper slope compared with metabolic networks indicating hierarchical modular organization. Applying k-core decomposition reveals a strong peripheral component, offering potential for communication on the periphery, not relying heavily on central nodes. Overall, we extend the organizational models typically applied to metabolic networks in order to address the properties of directed signalling networks, uncovering key organizational principles in cell signalling.