Thesis (MSc (Biochemistry))--University of Stellenbosch, 2009. / A Python package called PySUNDIALS has been developed which provides an interface to the suite of nonlinear di erential/algebraic equation solvers (SUNDIALS) using ctypes as a foreign function interface (FFI). SUNDIALS is a C implementation of a set of modern algorithms for integrating and solving various forms of the initial value problem (IVP). Additionally, arbitrary root nding capabilities, time dependent sensitivity analysis, and the solution of di erential and algebraic systems are available in the various modules provided by SUNDIALS. A signi cant focus of the project was to ensure the python package conforms to Python language standards and syntactic expectations. Multiple examples of the SUNDIALS modules (CVODE, CVODES, IDA and KINSOL) are presented comparing PySUNDIALS to C SUNDIALS (for veri cation of correctness), and comparing PySUNDIALS to various other comparable software packages. The examples presented also provide benchmark comparisons for speed, and code length. Speci c uses of the features of the SUNDIALS package are illustrated, including the modelling of discontinuous events using root nding, time dependent sensitivity analysis of oscillatory systems, and the modelling of equilibrium blocks using a complete set of implicit di erential and algebraic equations. PySUNDIALS is available as open source software for download. It is being integrated into the systems biology software PySCeS as an optional solver set, on an ongoing basis. A brief discussion of potential methods of optimization and the continuation of the project to wrap the parallel processing modules of SUNDIALS is presented.

L’objectif de cette thèse est de proposer une nouvelle méthode de construction de modèles métaboliques dans le contexte de la génomique comparée. Nous avons développé un outil, ab-pantograph, permettant l’inférence de modèles métabolique se basant sur la logique abductive. Pour ce faire, nous avons introduit une représentation logique de modèles métaboliques minimaux enzymatiques, puis à partir d’un modèle métabolique dit de référence, nous avons dérivé un modèle minimal enzymatique explicite accompagné d’association de gènes. Enfin, en couplant ce modèle métabolique au génome d’un organisme cible, nous inférons par abduction un modèle enzymatique pour cet organisme cible accompagné d’un ensemble d’associations de gènes, modèle que l’on veut congruent à celui que l’on aurait pu obtenir en ayant toutes les informations pour l’organisme cible.L’outil proposé, ab-pantograph, a été développé en utilisant la programmation logique par contraintes et Hyprolog. / The objective of this thesis is to propose a new method of constructing metabolic models in the context of comparative genomics. We have developed a tool, abpantograph, allowing the inference of metabolic models based on the Abductive logic. To do this, we have introduced a logical representation of minimal enzymatic metabolic models and from a metabolic model called reference, we derived an explicit enzymatic minimal model accompanied by gene association. Finally, by coupling this metabolic modele with the genome of a target organism, we infer abductively a model enzyme for this target organism accompanied by a set of gene associations, pattern one wants congruent to that which is could have obtained by having all the information to the target organism. The proposed tool, ab-pantograph, has been developed using constraint logic programming and Hyprolog.