Biological network modelling

This study takes a network approach to understanding complex biological systems. The overall objective is to explore how the stability and flexibility of biological networks emerge from underlying structural and dynamical characteristics. The thesis is arranged as a journey into the complexity of biological network models. The starting point is qualitative structural network descriptions. The level of detail in the dynamical description of node properties is then gradually increased. Along this journey, new features, both structural and dynamical, are revealed as crucial for the function of biological networks. A set of constructional properties are defined: structural principles, structural complexity, interaction diversity, node diversity and network density. These constructional properties capture important aspects of the structural organization and dynamic mechanisms in biological networks. A set of functional properties are defined: structural robustness, structural cyclicity, dynamic stability and dynamic flexibility. These functional properties are systemic properties that are all related to the stability of biological networks. These two sets of properties are used to demonstrate how the construction of biological networks is crucial for their function. The general theory is applied to food web and neural network models, where the general network properties are given specific biological meanings. The studies within both fields have their system specific objectives. A simple food web model is developed for explicitly including a compartment for dead organic material (detritus). Several constructional properties are revealed as crucial for the structural robustness, the structural cyclicity and the dynamic stability of food webs. The pathways due to decomposing and recycling of detritus alter the constructional properties, and are crucial for food web function. Computational neural network models are developed for clinical applications. Possible mechanisms behind electroconvulsive treatment (ECT) and anaesthetics are modelled. Clinical observations are qualitatively reproduced. Several aspects of the constructional properties of neural networks are revealed as crucial for optimal stability and flexibility of neurodynamics