2025 AIChE Annual Meeting

(595f) Validating a Cell-Scale Calcium Signaling Digital Twin Using Science-Based Design of Experiments

Authors

Jeremiah Zartman, University of Notre Dame
Alexander Dowling, University of Notre Dame
A digital twin, a sophisticated, predictive, computational representation of a physical system, rapidly transforms fields from healthcare to biological research. Recent advances in computational modeling and high-throughput data acquisition empower researchers to construct digital twins at increasingly granular scales, including the single-cell level. Studies leveraging Drosophila melanogaster as a powerful model organism demonstrate the potential to integrate complex biochemical and mechanical properties into comprehensive cellular models (Mirzoyan et al., 2019). By deploying these models, researchers extract insights into cellular dynamics, catalyzing innovative systems biology and disease modeling approaches. Constructing these high-fidelity single-cell digital twins necessitates the acquisition of massive, multi-modal datasets. Researchers must engineer efficient experimental protocols and leverage automated data pipelines to address this. Engineering effective protocols include optimizing sensor placement, designing high-throughput imaging systems, and implementing robust data pre-processing algorithms to minimize noise and maximize information content.

As demand for biological and medical digital twins increases, so does the imperative for efficient experimental design strategies. Science-based design of experiments (SBDoE) is a pivotal methodology to maximize information gain while minimizing resource expenditure (Chakrabarty et al., 2013). Synergizing SBDoE with mechanistic and data-driven models substantially amplifies the predictive accuracy of digital twins. Open-source tools like Pyomo.DoE provide a structured framework to optimize experimental conditions and refine model parameter estimation (Wang & Dowling, 2022). These approaches propel the evolution of digital twins from conceptual frameworks to practical tools for dissecting complex biological phenomena. Despite these advancements, critical gaps persist in developing and applying biological digital twins. A primary challenge lies in the robust validation of models across diverse experimental conditions and biological systems. Moreover, combining information from diverse data streams, such as calcium signaling pathways and morphogenetic processes, into a unified digital twin framework remains an active area of research (Berridge, 2016; Lobato-Rios et al., 2022). Furthermore, while applying digital twins to human disease modeling holds immense promise, translating findings from model organisms to human systems demands rigorous refinement and validation (Yamaguchi, 2018). Addressing these gaps is essential to unlock the full potential of digital twins in biomedical research and personalized medicine. As the field accelerates, sustained interdisciplinary collaboration and methodological innovation will be pivotal in bridging these gaps and pioneering new frontiers in life sciences (Committee on Digital Twins, 2024).

This work validates a cellular digital twin using calcium signaling data from Drosophila melanogaster. A key challenge is navigating the dynamical system models' high-dimensional, non-convex parameter space, often containing multiple local optima. By initiating the algorithm from diverse starting points, this approach broadens the search for potential solutions. It reduces the risk of getting trapped in suboptimal local minima, resulting in parameter estimates that better reflect biological reality and improve the digital twin's predictive capabilities (Martí, 2003). The calcium signaling cell scale digital twin will enable scientists to differentiate laboratory data through parameter estimation against selected mathematical models. The resulting parameter estimates will allow a deeper understanding of the mechanistic differences present with a quantified amount of uncertainty and facilitate discovery in the life sciences community.

References:

  1. Mirzoyan, Z.; Sollazzo, M.; Allocca, M.; Valenza, A.; Grifoni, D.; Bellosta, P. Drosophila melanogaster: A Model Organism to Study Cancer. Front. Genet. 2019, 10, 51.
  2. Chakrabarty, S.; Wang, H. Y.; Stephanopoulos, G. Model‐Based Design of Experiments for Cellular Processes. WIREs Mech. Dis. 2013, 1 (3), 277–292.
  3. Wang, H.; Dowling, A. Pyomo.DoE: An Open-Source Package for Model-Based Design of Experiments in Python. AIChE J. 2022, 68 (4), e17620.
  4. Berridge, M. J. The Inositol Trisphosphate/Calcium Signaling Pathway in Health and Disease. Physiol. Rev. 2016, 96 (4), 1261–1296.
  5. Lobato-Rios, V.; Mendes, A.; de Vries, J.; Boyen, S.; Karamouzas, I.; Devaine, M. NeuroMechFly: A Neuromechanical Model of Adult Drosophila melanogaster. Nat. Methods 2022, 19 (8), 1015–1025.
  6. Yamaguchi, M. Drosophila as a Model Organism for Neurodegenerative Diseases Associated with Genetic Variation in Humans. Exp. Anim. 2018, 67 (4), 421–430.
  7. Committee on Foundational Research Gaps and Future Directions for Digital Twins, National Academies of Sciences, Engineering, and Medicine. Foundational Research Gaps and Future Directions for Digital Twins; The National Academies Press: Washington, DC, 2024.
  8. Martí, R. Multi-Start Methods. In Handbook of Metaheuristics; Glover, F., Kochenberger, G. A., Eds.; Springer US: Boston, MA, 2003; pp 355–368. https://doi.org/10.1007/0-306-48056-5_12.