Design of Robust and Efficient Topology Using Enhanced Gene Regulatory Networks

Abstract

Biological networks are characterized by their inherent robustness against component failures. Gene regulatory networks (GRNs) are biological networks with graph properties contributing to their innate functional robustness. In this paper, we first propose a three tier topological characterization to study the graph properties of GRN, namely scale free out-degree distribution, low graph density, and abundance of subgraphs, called motifs. We then present a novel edge rewiring mechanism, consisting of edge addition and deletion algorithms, to remedy its vulnerability against failure of well-connected nodes while preserving its graph properties. We discuss the preferential attachment growth-based edge addition and greedy edge deletion. We then highlight the drawbacks of greedy edge deletion and propose an improved dynamic edge deletion algorithm, employing nonlinear least squares optimization to preserve in-degree distribution of original GRN. We prove that dynamically rewired GRNs preserve the optimal number of a three-node motif called feed forward loops, which render topological robustness to GRN. Our graph experiments reveal that dynamically rewired GRN subgraphs possess higher robustness when compared against greedily rewired GRN, original GRN and Erdös-Rényi random graphs of similar graph density. Finally, we perform wireless sensor network (WSN) simulations using OMNET++, to demonstrate that rewired GRN-based WSNs yield significantly higher packet delivery at comparable latency.