Kinetic model construction using chemoinformatics

Kinetic models of chemical processes not only provide an alternative to costly experiments; they also have the potential to accelerate the pace of innovation in developing new chemical processes or in improving existing ones. Kinetic models are most powerful when they reflect the underlying chemistry by incorporating elementary pathways between individual molecules. The downside of this high level of detail is that the complexity and size of the models also steadily increase, such that the models eventually become too difficult to be manually constructed. Instead, computers are programmed to automate the construction of these models, and make use of graph theory to translate chemical entities such as molecules and reactions into computer-understandable representations. This work studies the use of automated methods to construct kinetic models. More particularly, the need to account for the three-dimensional arrangement of atoms in molecules and reactions of kinetic models is investigated and illustrated by two case studies. First of all, the thermal rearrangement of two monoterpenoids, cis- and trans-2-pinanol, is studied. A kinetic model that accounts for the differences in reactivity and selectivity of both pinanol diastereomers is proposed. Secondly, a kinetic model for the pyrolysis of the fuel "JP-10" is constructed and highlights the use of state-of-the-art techniques for the automated estimation of thermochemistry of polycyclic molecules. A new code is developed for the automated construction of kinetic models and takes advantage of the advances made in the field of chemo-informatics to tackle fundamental issues of previous approaches. Novel algorithms are developed for three important aspects of automated construction of kinetic models: the estimation of symmetry of molecules and reactions, the incorporation of stereochemistry in kinetic models, and the estimation of thermochemical and kinetic data using scalable structure-property methods. Finally, the application of the code is illustrated by the automated construction of a kinetic model for alkylsulfide pyrolysis.

http://ift.tt/2oPwUnm