Controlling Complexity in Biological Networks

Bradley Tonkes1, Janet Wiles2, John S. Mattick
1btonkes@itee.uq.edu.au, Institute for Molecular Bioscience, School of Information Technology and Electrical Engineering, University of Queensland; 2janetw@itee.uq.edu.au, School of Psychology, School of Information Technology and Electrical Engineering, University of Queensland

A major challenge for molecular biology is to find appropriate computational abstractions to study the complexity and control of regulatory networks responsible for autogenesis. Although complex behavior is easy to generate in any combinatorial system, controlling functional trajectories remains an open challenge. The most common framework to date for studying complexity in genetic systems uses random Boolean networks, which model the dynamics of epistasis but not functionality. However other models of network computation have been used as control systems, notably recurrent neural networks in computational neuroscience. Here we examine the complexity characteristics of an artificial neural network, initially as an abstract dynamical system and then as a functional system capable of regulating the ontogeny of C. elegans. We show that one solution to controlling complexity in genetic regulatory networks is to maintain their dynamics below the threshold of chaos and use inputs to the network to exploit transient dynamics in a simple (fixed point) attractor system. This approach provides a means to study how functionality can be encoded and controlled in regulatory networks that program the ontogenesis of multicellular organisms.