Agunloye, Emmanuel; Galvanin, Federico; Petsagkourakis, Panagiotis; Labes, Ricardo; Chamberlain, Thomas W; Muller, Frans L; Bourne, Richard; (2024) Distinguishing alternative kinetic models for hydrogen borrowing within the model-based design of experiment framework for model discrimination. Presented at: ChemEngDayUK 2024, London, UK.

Preview

Text ChemEngDayUK2024_Poster_Agunloye.pdf - Published Version Download (663kB) | Preview

Abstract

Hydrogen borrowing is a catalytic synthetic protocol gaining popularity for producing a variety of new drug entities as the synthetic protocol can elongate the entity molecular structure when repeated in several cycles. While various kinetic models can be developed to describe the hydrogen borrowing cycles, available experimental data constrain the model space to a few kinetic models with estimable parameters. The best model can then be selected using model-based design of experiments for model discrimination (MBDoE-MD), hindered only where models are indistinguishable. Focussing on a synthesis case study that can be described using two hydrogen borrowing cycles, we employ in this work a software system called “SimBot”, which within a cloud-based platform controls a remotely located smart flow reactor system for physical experimentation and comprises modules for kinetic model development, parameter estimation and MBDoE-MD. Among six candidate kinetic models developed for hydrogen borrowing, Simbot identified only two candidate kinetic models as statistically acceptable after employing available experimental data by assessing the models adequacy using the χ^2 test and parameter identifiability using Fisher information matrix-based identifiability criteria. Using in-silico experiments manipulating inlet flow conditions and temperature only, these two models (Model 1, Model 2) could not be distinguished within the MBDoE framework for model discrimination as their corresponding discrimination probabilities were close to 0.50. However, exploiting the model structural differences: Model 1 being zeroth order with respect to the catalyst amount while Model 2 being first order, the SimBot software allowed to identify the catalyst amount as key model discrimination driver to be used in further kinetic experiments. In silico testing showed that assuming 1, 2 and 3% decrease in catalyst amount, Model 1 differed from Model 2, and these experiments would allow to achieve discrimination probabilities of 0.52, 0.78 and 0.99, respectively. Future validation experiments in the automated platforms will be needed to confirm the adequacy of Model 2 on representing reaction kinetics in the hydrogen borrowing system.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as CSVJSONXML

Archive Staff Only

View Item