Model-based evaluation and data requirements for parallel kinetic experimentation and data-driven reaction identification and optimization

Recently there has been growing interest in implementing the high-throughput approach to access the dynamics of chemical processes across different fields. With an ever-increasing amount of data provided by high-throughput experimentation, the development of fully-integrated workflows becomes crucial. These workflows should combine novel experimental tools and interpretation methods to convert the data into valuable information. To design feasible data-driven workflows, it is necessary to estimate the value of information and balance it with the number of experiments and resources required. Basing this kind of workflow on actual physical models appears to be a more feasible strategy as compared to data-extensive empirical statistical methods. Here we show an algorithm that constructs and evaluates kinetic models of different complexity. The algorithm facilitates the evaluation of the experimental data quality and quantity requirements needed for the reliable discovery of the rates driving the corresponding chemical models. The influence of the quality and quantity of data on the obtained results was indicated by the accuracy of the estimates of the kinetic parameters. We also show that this method can be used to find correct reaction scenarios directly from simulated kinetic data with little to no overfitting. Well-fitting models for theoretical data can then be used as a proxy for optimizing the underlying chemical systems. Taking real physical effects into account, this approach goes beyond: we show that with the kinetic models, one can make a direct, unbiased, quantitative connection between kinetic data and the reaction scenario. Chemfit uses kinetic data to facilitate the discovery of rates within chemical models, the quantification of chemical model accuracy, and the exploration of data quality and quantity requirements needed for reliable kinetic parameter discovery.