A bio-inspired optimization for inferring interactive networks: Cockroach swarm evolution

Most diseases are the result of interactions between gene products. Inferring the quantitatively dynamic interactive mechanisms of underlying systems has become a central part of the ongoing revolution in biology. Biological systems are sparely connected and always noise contaminated. The connection cannot be neglected even the strength is down to 0.001. Therefore, the used computational technologies are not only good in exploitation and exploration but also robust to noise. The development of effective top-down modeling technologies to largely increase the gap between redundant and true connections is a significant challenge. Cockroaches are found in nearly all habitats and survive in extreme environments and when there is food scarcity. We previously introduced cockroaches' competition behavior to improve GAs' exploitation for the parameter estimation of biological systems. However, cockroaches always work together for foraging. Competition occurs only during food is extremely scarcity. In this study, we further mimick cockroaches' cooperated-based swarm behavior, wherein their competitive behavior and migration are event-induced. The proposed cockroach-inspired swarm evolution (CSE) was tested through dry-lab experiments under noise environment. We successfully identified 80 connections (62 redundant and 18 true connections) with truncated interactive strengths smaller than 10−14. Only one to three pruning steps for noise-free systems and five steps for noise-contaminated systems are needed even in a wide searching space. The estimated parameters are almost perfectly equal to the true ones.