Mihaljev, Tamara (2008)
Dynamics and evolution of random Boolean networks.
Technische Universität Darmstadt
Dissertation, Erstveröffentlichung
Kurzbeschreibung (Abstract)
Random Boolean networks are used as generic models for the dynamics of complex systems of interacting entities, such as social and economic networks, neural networks, and gene or protein interaction networks. The model studied in this thesis was introduced by S. Kauffman as a simple model for gene regulation. The system consists of N nodes, each of which receives inputs from K randomly chosen nodes. The state of a node is a Boolean function of the states of its input nodes. The functions are assigned to the nodes at random, and the states of all nodes in a network are updated in parallel. Asymptotic dynamical states of a network are represented by attractors in state space. Thus, their number and lengths are important features of the networks. The nodes in a network can be classified as frozen, irrelevant, or relevant, according to their dynamics on an attractor. The relevant nodes determine completely the number and the period of attractors. Although the random Boolean network model is simple, it shows a rich dynamical behavior with a phase transition between a frozen and a disordered phase and a very complex dynamics at the critical point between the phases. In this thesis dynamics and evolution of random Boolean networks are studied. The investigation of the dynamical properties of the model starts with the simplest realization of a random Boolean network, that is, with the network with one input per node. The topology of these networks is analyzed by generating networks through a growth process. Using probabilistic arguments and estimating the lower bounds, it is analytically proven that in this class of networks both, the mean number and the mean length of attractors grow faster than any power law with the size of the network. Next, the dynamics of critical networks with two inputs per node is studied and these studies are then generalized to networks with a larger number of inputs. Using methods from the theory of stochastic processes, the scaling behavior of the numbers of nonfrozen and relevant nodes is determined analytically. For all critical networks with K > 1 the same power-laws are found. The results obtained for the K = 1 networks are then used to show that in all critical random Boolean networks the mean number and length of attractors diverge faster than any power law with the network size. For the modeling of gene regulatory networks this means that the attractors are too long and too many to represent cellular differentiation, to which the model was originally applied. However, real networks are not random but are the result of evolutionary processes. Therefore, the evolution of populations of random Boolean networks under selection for robustness of the dynamics under small perturbations is investigated. The results of this study show that the fitness landscape contains a huge plateau of maximum fitness that spans the entire network space. It is found that the networks evolved on such a landscape are robust to changes in their structure, while being at the same time able to preserve their function under small environmental changes.
Typ des Eintrags: | Dissertation | ||||
---|---|---|---|---|---|
Erschienen: | 2008 | ||||
Autor(en): | Mihaljev, Tamara | ||||
Art des Eintrags: | Erstveröffentlichung | ||||
Titel: | Dynamics and evolution of random Boolean networks | ||||
Sprache: | Englisch | ||||
Referenten: | Drossel, Prof. Dr. Barbara ; Porto, Prof. Dr. Markus | ||||
Publikationsjahr: | 21 November 2008 | ||||
Ort: | Darmstadt | ||||
Verlag: | Technische Universität | ||||
Datum der mündlichen Prüfung: | 10 November 2008 | ||||
URL / URN: | urn:nbn:de:tuda-tuprints-11880 | ||||
Kurzbeschreibung (Abstract): | Random Boolean networks are used as generic models for the dynamics of complex systems of interacting entities, such as social and economic networks, neural networks, and gene or protein interaction networks. The model studied in this thesis was introduced by S. Kauffman as a simple model for gene regulation. The system consists of N nodes, each of which receives inputs from K randomly chosen nodes. The state of a node is a Boolean function of the states of its input nodes. The functions are assigned to the nodes at random, and the states of all nodes in a network are updated in parallel. Asymptotic dynamical states of a network are represented by attractors in state space. Thus, their number and lengths are important features of the networks. The nodes in a network can be classified as frozen, irrelevant, or relevant, according to their dynamics on an attractor. The relevant nodes determine completely the number and the period of attractors. Although the random Boolean network model is simple, it shows a rich dynamical behavior with a phase transition between a frozen and a disordered phase and a very complex dynamics at the critical point between the phases. In this thesis dynamics and evolution of random Boolean networks are studied. The investigation of the dynamical properties of the model starts with the simplest realization of a random Boolean network, that is, with the network with one input per node. The topology of these networks is analyzed by generating networks through a growth process. Using probabilistic arguments and estimating the lower bounds, it is analytically proven that in this class of networks both, the mean number and the mean length of attractors grow faster than any power law with the size of the network. Next, the dynamics of critical networks with two inputs per node is studied and these studies are then generalized to networks with a larger number of inputs. Using methods from the theory of stochastic processes, the scaling behavior of the numbers of nonfrozen and relevant nodes is determined analytically. For all critical networks with K > 1 the same power-laws are found. The results obtained for the K = 1 networks are then used to show that in all critical random Boolean networks the mean number and length of attractors diverge faster than any power law with the network size. For the modeling of gene regulatory networks this means that the attractors are too long and too many to represent cellular differentiation, to which the model was originally applied. However, real networks are not random but are the result of evolutionary processes. Therefore, the evolution of populations of random Boolean networks under selection for robustness of the dynamics under small perturbations is investigated. The results of this study show that the fitness landscape contains a huge plateau of maximum fitness that spans the entire network space. It is found that the networks evolved on such a landscape are robust to changes in their structure, while being at the same time able to preserve their function under small environmental changes. |
||||
Alternatives oder übersetztes Abstract: |
|
||||
Freie Schlagworte: | Networks, Kauffman model, Random Boolean networks, genetic regulation, attractors, universal scaling, dynamics, evolution | ||||
Sachgruppe der Dewey Dezimalklassifikatin (DDC): | 500 Naturwissenschaften und Mathematik > 530 Physik | ||||
Fachbereich(e)/-gebiet(e): | 05 Fachbereich Physik 05 Fachbereich Physik > Institut für Festkörperphysik (2021 umbenannt in Institut für Physik Kondensierter Materie (IPKM)) |
||||
Hinterlegungsdatum: | 24 Nov 2008 08:34 | ||||
Letzte Änderung: | 26 Aug 2018 21:25 | ||||
PPN: | |||||
Referenten: | Drossel, Prof. Dr. Barbara ; Porto, Prof. Dr. Markus | ||||
Datum der mündlichen Prüfung / Verteidigung / mdl. Prüfung: | 10 November 2008 | ||||
Export: | |||||
Suche nach Titel in: | TUfind oder in Google |
Frage zum Eintrag |
Optionen (nur für Redakteure)
Redaktionelle Details anzeigen |