BPS2026 – A modular framework for tunable hybrid modeling of calcium signaling dynamics

Biological systems like calcium signaling involve stochastic and complex spatiotemporal dynamics that are difficult and often expensive to model. Most current methods focus on a single scale and/or modeling technique, but we see a growing trend toward hybrid and multiscale models to describe these systems with increasing complexity. However, building these systems is non-trivial and often requires an immense amount of custom code. To address these limitations, we propose a novel architecture that covers how we integrate algorithms and data structures, biological phenomena, multiscale/multi-discipline orchestration, and how we attach these to biological systems of interest. By focusing on clean abstractions, well-defined interfaces, and modular design, our architecture enforces a structured and principled approach to the development of building blocks at each of the mentioned layers, creating a flexible system without the loss of performance through zero-cost abstractions. We showcase this using a simple reaction-diffusion model of calcium signaling cascade and demonstrate the potential for larger and longer simulations using a combination of hybrid (Agents, PDEs, and ODEs) techniques and dynamically tunable resolution in which the model can shift the underlying algorithms and solvers over time. Together, these lay the foundation for the development of increasingly complex models with rapid iteration and improved performance as our understanding of the model evolves. This approach will enable more accurate and scalable simulations of signaling pathways, and can be broadly applied to the design of mechanistic modeling frameworks aimed at uncovering biological behavior.