Modeling cell signaling pathways through universal differential equations and joint inference of first-principle parameters and neural network weights

• Cristiano G. S. Campos UNICAMP
• Ronaldo N. Sousa USP / Instituto Butantan
• Hugo A. Armelin USP / Instituto Butantan
• Marcelo S. Reis USP / Instituto Butantan / UNICAMP

DOI: https://doi.org/10.5753/bsb.2024.245611

Resumo

The regulation of cellular processes is governed by chains of chemical reactions, known as cell signaling pathways. A key challenge in modeling these pathways is the “lack of isolation problem”, where reactions within the model fail to interact with those in the broader cellular context, reducing prediction accuracy in first-principle models. Moreover, often some first-principle parameters are missing and must be inferred from data. To address this, we propose a hybrid modeling approach combining ordinary differential equation (ODE)-based first-principle models with neural network-based data-driven models, which jointly infers both neural network weights and missing firstprinciple parameters. Computational experiments using an iron metabolism model and a model implementation based on universal differential equations (UDEs) demonstrated significant improvements in prediction accuracy compared to first-principle models. These results support UDE-based hybrid models as effective tools for studying the complex dynamics of biological systems.

Referências

Bangi, F. M. S., Kao, K., and Kwon, J. S. (2022). Physics-informed neural networks for hybrid modeling of lab-scale batch fermentation for beta-carotene production using saccharomyces cerevisiae. Chemical Engineering Research and Design, 179:415–423. DOI: 10.1016/j.cherd.2022.01.041.

Dong, S., Zhang, Y., and Zhou, X. (2023). Intelligent hybrid modeling of complex leaching system based on lstm neural network. Systems, 11(2). DOI: 10.3390/systems11020078.

Li, K., Duan, H., Liu, L., Qiu, R., van den Akker, B., Ni, B.-J., Chen, T., Yin, H., Yuan, Z., and Ye, L. (2022). An integrated first principal and deep learning approach for modeling nitrous oxide emissions from wastewater treatment plants. Environmental Science and Technology, 56(4):2816–2826. DOI: 10.1021/acs.est.1c05020.

Lima, F. R., Rebello, C. M., Costa, E. A., Santana, V. V., de Moares, M. G., Barreto, A. G., Secchi, A. R., de Souza, M. B., and Nogueira, I. B. (2023). Improved modeling of crystallization processes by universal differential equations. Chemical Engineering Research and Design, 200:538–549. DOI: 10.1016/j.cherd.2023.11.032.

Lopes, T. J., Luganskaja, T., Vujić, M. S., Hentze, M. W., Muckenthaler, M. U., Schümann, K., and Reich, J. G. (2010). Systems analysis of iron metabolism: the network of iron pools and fluxes. BMC Systems Biology. DOI: 10.1186/1752-0509-4-112.

Lüders, C., Sturm, T., and Radulescu, O. (2022). ODEbase: A repository of ODE systems for systems biology. Bioinformatics Advances, 2(1). DOI: 10.1093/bioadv/vbac027.

Narayanan, H., C. Bournazou, M. N., G. Gosálbez, G., and Butté, A. (2022). Functional-hybrid modeling through automated adaptive symbolic regression for interpretable mathematical expressions. Chemical Engineering Journal, 430:133032. DOI: 10.1016/j.cej.2021.133032.

Rackauckas, C., Ma, Y., Martensen, J., Warner, C., Zubov, K., Supekar, R., Skinner, D., Ramadhan, A., and Edelman, A. (2021). Universal differential equations for scientific machine learning. DOI: 10.48550/arXiv.2001.04385.

Reis, M. S., Noël, V., Dias, M. H., Albuquerque, L. L., Guimarães, A. S., Wu, L., Barrera, J., and Armelin, H. A. (2017). An interdisciplinary approach for designing kinetic models of the Ras/MAPK signaling pathway. In Methods in Molecular Biology Special Edition on Kinase Signaling Networks, pages 455–474. Humana Press, New York. DOI: 10.1007/978-1-4939-7154-1_28.

Santana, V. V., Costa, E., Rebello, C. M., Ribeiro, A. M., Rackauckas, C., and Nogueira, I. B. (2023). Efficient hybrid modeling and sorption model discovery for non-linear advection-diffusion-sorption systems: A systematic scientific machine learning approach. Chemical Engineering Science, 282:119223. DOI: 10.1016/j.ces.2023.119223.

Sousa, R. N., Campos, C. G. S., Wang, W., Hashimoto, R. F., Armelin, H. A., and Reis, M. S. (2023). Exploring identifiability in hybrid models of cell signaling pathways. In Reis, M. S. and de Melo-Minardi, R. C., editors, Advances in Bioinformatics and Computational Biology, pages 148–159, Cham. Springer Nature Switzerland. DOI: 10.1007/978-3-031-42715-2_14.

V. Wouwer, A., Renotte, C., and Bogaerts, P. (2004). Biological reaction modeling using radial basis function networks. Computers and Chemical Engineering, 28(11):2157–2164. DOI: 10.1016/j.compchemeng.2004.03.003.

Wittig, U., Kania, R., Golebiewski, M., Rey, M., Shi, L., Jong, L., Algaa, E., Weidemann, A., Sauer-Danzwith, H., Mir, S., Krebs, O., Bittkowski, M., Wetsch, E., Rojas, I., and Müller, W. (2011). SABIO-RK—database for biochemical reaction kinetics. Nucleic Acids Research, 40(D1):D790–D796. DOI: 10.1093/nar/gkr1046.

Zander, H.-J., Dittmeyer, R., and Wagenhuber, J. (1999). Dynamic modeling of chemical reaction systems with neural networks and hybrid models. Chemical Engineering Technology. DOI: 10.1002/(SICI)1521-4125(199907)22:7<571::AID-CEAT571>3.0.CO;2-5.