(- / 0)

Recording date:

Duration:

Session:

Speaker:

Co-Authors:

Back

Abstract:

Keywords

Python tool, bioprocess digitalization, hybrid mechanistic/ML modeling, Industry 4.0

Abstract

Hybrid models combine state-of-the-art machine learning algorithms (ML) with mechanistic models in the same model structure and is gaining prominence in developing digital twins for Biopharma 4.0 because it offers a flexible balance between prior knowledge and data availability. They have been widely applied in process systems engineering for predictive modeling, process monitoring, model predictive control and design of experiments as well as systems biology problems (Agharafeie et al., 2023). For example, in the last years substantial work in (bio)systems approaches that merge ML/AI and traditional engineering methods were developed, namely the concept of “general bioprocess hybrid model” to describe bioprocess dynamics (Pinto et al., 2022, Teixeira et al., 2007) and for the field of systems biology (von Stosch et al., 2010, Pinto et al., 2023). Nevertheless, the use of such hybrid  modeling techniques has usually been limited. Only a small number of experts (research groups of the in-house tools) have the knowledge and methods to develop such models worldwide. Additionally, a specific open-source and user-friendly software tool to develop a hybrid model is lacking. To  overcome the current limitations, including the dependency on proprietary software, this work aims to provide an open-source, modular and easy-to-use (automated) framework for hybrid models building, simulating and analysis. The outcome is HYBpy, a user-friendly Python tool designed to accelerate the development of decision hybrid models, which offers access to a generalized step- by-step pipeline and intuitive user interfaces. HYBpy provides an added-value for the community interested in hybrid modeling without any programming knowledge or background on hybrid model development, to  improve (bio)process understanding and support relevant process interpretation tasks. We demonstrated the utility of HYBPlat through benchmark case studies. The source code is available at the GitHub repository (https://github.com/joko1712/HYBpy-a-Python-tool-for-hybrid-modeling).

Acknowledgement

This work was supported by Fundação para a Ciência e Tecnologia (FCT) and the Associate Laboratory for Green Chemistry (LAQV) financed by national funds from FCT/MCTES (LA/P/0008/2020 DOI 10.54499/LA/P/0008/2020, UIDP/50006/2020 DOI 10.54499/UIDP/50006/2020, UIDB/50006/2020 DOI 10.54499/UIDB/50006/2020). JP acknowledges PhD grant SFRD/BD14610472019, FCT. RSC acknowledges the contract CEECIND/01399/2017. The authors also wish to acknowledge the European Union’s Horizon BioLaMer project under grant agreement 101099487.

References

  • AGHARAFEIE, R., RAMOS, J. R. C., MENDES, J. M. & OLIVEIRA, R. 2023. From Shallow to Deep Bioprocess Hybrid Modeling: Advances and Future Perspectives. Fermentation-Basel, 9.
  • PINTO, J., MESTRE, M., RAMOS, J., COSTA, R. S., STRIEDNER, G. & OLIVEIRA, R. 2022. A general deep hybrid model for bioreactor systems: Combining first principles with deep neural networks. Computers & Chemical Engineering, 165.
  • PINTO, J., RAMOS, J., COSTA, R. S. & OLIVEIRA, R. 2023. A General Hybrid Modeling Framework for Systems Biology Applications: Combining Mechanistic Knowledge with Deep Neural Networks under the SBML Standard. AI, 4, 303-318.
  • TEIXEIRA, A. P., ALVES, C., ALVES, P. M., CARRONDO, M. J. T. & OLIVEIRA, R. 2007. Hybrid elementary flux analysis/nonparametric modeling: application for bioprocess control. Bmc Bioinformatics, 8.
  • VON STOSCH, M., PERES, J., DE AZEVEDO, S. F. & OLIVEIRA, R. 2010. Modelling  biochemical networks with intrinsic time delays: a hybrid semi-parametric approach. Bmc Systems Biology, 4.