| Popis: |
Over the past 10 years, multiple executable modelling formalisms for molecular biology have been developed in order to address the growing need for a system-level understanding of complex biological phenomena. An important class of these formalisms are biology-inspired process algebras, which offer-among other desirable properties - an almost complete separation of model specification (syntax) from model dynamics (semantics). In this thesis, the similarity between this separation and the genotype-phenotype duality in evolutionary biology is exploited to develop a process-algebraic approach to the study of evolution of biochemical systems. The main technical contribution of this thesis is the continuous π-calculus (cπ), a novel process algebra based on the classical π-calculus of Milner et. al. Its two defining characteristics are: continuous, compositional, computationally inexpensive semantics, and a exible interaction structure of processes (molecules). Both these features are conductive to evolutionary analysis of biochemical systems by, respectively, enabling many variants of a given model to be evaluated, and facilitating in silico evolution of new functional connections. A further major contribution is a collection of variation operators, syntactic model transformation schemes corresponding to common evolutionary events. When applied to a cπ model of a biochemical system, variation operators produce its evolutionary neighbours, yielding insights into the local fitness landscape and neutral neighbourhood. Two well-known biochemical systems are modelled in this dissertation to validate the developed theory. One is the KaiABC circadian clock in the cyanobacterium S. elongatus, the other is a mitogen-activated protein kinase cascade. In each case we study the system itself as well as its predicted evolutionary variants. Simpler examples, particularly that of a generic enzymatic reaction, are used throughout the thesis to illustrate important concepts as they are introduced. |
