Agunloye, Emmanuel; Petsagkourakis, Panagiotis; Yusuf, Muhammad; Chamberlain, Thomas; Muller, Frans; Bourne, Richard; Galvanin, Federico; (2023) Application of a novel cloud-based platform for kinetics model identification in pharmaceutical processes. Presented at: Pharmaceutical Manufacturing Forum, London, UK.

Preview

Text Poster Emmanuel Agunloye.pdf - Published Version Download (852kB) | Preview

Abstract

The pharmaceutical industry has recently implemented a new framework to incorporate industry 4.0 technologies into drug manufacturing to guarantee quality and accelerate commercialization. These technologies include cloud computing for speeding up information sharing among research collaborators, across companies and with health authorities worldwide, artificial intelligence, digitalization and mathematical modelling as accelerators for drug discovery, and the use of emerging data-driven and physics-based modelling technologies in pharmaceutical development and manufacturing for advanced process design, monitoring and control in continuous systems. In accordance, we present in this work a novel cloud-based platform driven by optimal experimental design software deployed from University College London to remotely control experimentation in a smart flow reactor system situated at University of Leeds. Employing hybrid modelling, the software initially designs experiments using model-free DoE techniques such as Latin hypercube sampling (LHS) to obtain process information and thereafter designs experiments using model-based design of experiments (MBDoE) techniques to identify reaction kinetics. The model is then validated online using statistical tests to achieve a probabilistic description of model reliability across the experimental design space. We have demonstrated this platform on pharmaceutical-related processes including homogenous amide formation, and heterogeneous hydrogen borrowing. In the homogeneous amide formation, following initial process understanding and experimentation driven by LHS, the platform tested 2 alternative kinetic models representing a single-forward chemical reaction and a reversible chemical reaction, respectively. While rejecting the former as inadequate in describing the process system, the cloud-based platform proceeded with the latter to design a new experiment, the most informative experiment in the design space, which by updating the experimental conditions ensured a precise estimation of kinetic model parameters. In heterogeneous hydrogen borrowing, a synthesis protocol being explored for new drug discovery, the platform via sequential parameter estimation and MBDoE for model discrimination, reduced 12 initially tested candidate kinetic models to 2 models with identifiable parameters and tested the latter models in silico for distinguishability.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item