Posted by Nodus Labs | April 9, 2012
What Makes a Network Robust?
A very interesting paper by Kitano “Biological Robustness” (published in 2004 in Nature Genetics journal) offers a very insightful summary on what biological robustness is and how it emerges in networks. Kitano demonstrates how small-world architecture and the existence of an interconnected core control center both increase the robustness of a system.
In short, Kitano shows that robustness is the ability of a system to perform its functions despite internal and external perturbations. He shows that the four main prerequisites for robustness are: systems feedback control, modularity, redundancy, and decoupling. Feedback control allows to integrate information from a range of inputs and produce the required response. Modularity ensures that the system can maintain several states simultaneously and perturbations in one state do not affect the others. Redundancy ensures that in case of a partial failure the functions can still be carried out by the remaining parts. Decoupling separates low-level perturbations from high-level functionalities.
Kitano uses an airplane as an example of a robust system. The core control system is responsible for the feedback mechanisms. Several units duplicating each others’ tasks are built in for redundancy. Various functional units are isolated to ensure modularity.
Therefore, polysingularity is the key quality of biological robustness: the network is structured to allow for simultaneous existence of several possible states that are distinct enough to support the differences, but connected enough to enable global cooperation.
These qualities allow the system to maintain its stability even when there are perturbations that temporarily shift it from its attractor state (shown on the image below). When perturbations are too strong, the system shifts to a different attractor state, which can deal with perturbations more efficiently.
Kitano proposes that robustness is the result of evolution and that’s why it’s present in so many biological systems. He also expands his idea to show how robustness comes at the price of flexibility (it’s difficult to change the state of the system) and how it’s susceptible to unexpected perturbations it has not been designed to cope with. He also shows how viral diseases “success” is due to their takeover of the core organism’s functions that are built to be robust. And how the best strategy of coping with them may not be the one of eradication, but instead taking over the control over the robust dynamics. In case of HIV virus, for example, this could be an introduction of a mutated virus that easily joins the robust dynamics but keeps the virus’ activity at a dormant level.