Parameter Synthesis for Cardiac Cell Hybrid Models Using Delta-Decisions

Bing Liu, Soonho Kong, Sicun Gao, Paolo Zuliani, Edmund M. Clarke

Abstract

A central problem in systems biology is to identify parameter values such that a biological model satisfies some behavioral constraints (e.g., time series). In this paper, we focus on parameter synthesis for hybrid (continuous/discrete) models, as many biological systems can possess multiple operational modes with specific continuous dynamics in each mode. These biological systems are naturally modeled as hybrid automata, most often with nonlinear continuous dynamics. However, hybrid automata are notoriously hard to analyze—even simple reachability for hybrid systems with linear differential dynamics is an undecidable problem. In this paper, we present a parameter synthesis framework based on $\delta$-complete decision procedures that sidesteps undecidability. We demonstrate our method on two highly nonlinear hybrid models of the cardiac cell action potential. The results show that our parameter synthesis framework is convenient and efficient, and it enabled us to select a suitable model to study and identify crucial parameter ranges related to cardiac disorders.

Introduction: The Challenge of Parameter Synthesis

In systems biology, researchers often utilize hybrid models to capture complex biological phenomena. These models combine discrete "modes"—representing different stages of a process like the cell cycle—with continuous dynamics governed by Ordinary Differential Equations (ODEs). A fundamental challenge in this field is parameter synthesis: identifying the specific rates, thresholds, and initial conditions that allow a model to accurately reproduce experimental observations.

Mathematically, determining if a hybrid system can reach a specific state (the reachability problem) is undecidable for most realistic, nonlinear models. Traditional methods often rely on extensive sampling and simulation, which may fail to explore critical edge cases or provide definitive proof that a certain behavior is impossible. This research introduces a rigorous, automated framework to explore these parameter spaces despite their inherent mathematical complexity.

Methodology: $\delta$-Complete Decision Procedures

The authors leverage a logic-based approach called $\delta$-decisions, implemented through the dReal and dReach toolsets. This framework sidesteps undecidability by relaxing the problem: for a given precision $\delta$, the solver returns either unsat (the property definitely does not hold) or $\delta$-sat (the property holds within a tiny error bound $\delta$). Crucially, an "unsat" result provides a formal guarantee of impossibility.

The reachability problem is encoded as a first-order logic formula over real numbers. The system is "unrolled" for a fixed number of discrete steps ($k$) using Bounded Model Checking (BMC), while continuous flows are handled via verified numerical integration. By integrating a binary search algorithm with this solver, the authors can pinpoint exact parameter thresholds where biological behavior changes.

Case Studies: Fenton-Karma and Bueno-Cherry-Fenton Models

The framework was evaluated on two highly nonlinear hybrid models of the cardiac cell action potential: the established Fenton-Karma (FK) model, often used for vortex dynamics, and the more recent Bueno-Cherry-Fenton (BCF) model, designed to better capture human ventricular cell dynamics.

Hybrid automata representations for the BCF (a) and FK (b) models, illustrating the discrete modes and continuous flows.

As shown in Figure 3, these models are structured as hybrid automata where transitions between modes are triggered by state variables (like voltage $u$) crossing specific thresholds. This structure allows the models to represent the various phases of a cardiac action potential.

Key Results: Model Falsification and Disease Identification

A significant outcome of this formal approach is the ability to definitively "falsify" models. The authors proved that the FK model is incapable of reproducing the "spike-and-dome" morphology characteristic of epicardial cells, regardless of the parameter values chosen. This formal proof allows biologists to select the BCF model as a more physiologically accurate representation for certain cardiac studies.

Simulated time series for the BCF (a) and FK (b) models, showing the evolution of state variables over time.

Beyond model selection, the framework identified critical parameter ranges associated with cardiac disorders. For instance, in the BCF model, it was found that if the parameter $\tau_{o1} < 0.006$, the cell loses excitability and cannot trigger an action potential—a condition that can lead to arrhythmias such as tachycardia. The framework also verified that both models are robust to small, noisy stimulations, correctly filtering out signals below a specific amplitude threshold.

Conclusion and Future Directions

While the $\delta$-complete approach provides a powerful tool for formal verification, it is not without limitations. The $\delta$-relaxation can occasionally lead to "spurious" solutions that exist only within the error bound, and Bounded Model Checking only analyzes a fixed number of transitions. Future work aims to extend this tool to support the Systems Biology Markup Language (SBML) and explore parameter estimation from noisy experimental data, making these formal methods more accessible to the broader biological research community.

Summary updated

Abstract

A central problem in systems biology is to identify parameter values such that a biological model satisfies some behavioral constraints (\eg, time series). In this paper we focus on parameter synthesis for hybrid (continuous/discrete) models, as many biological systems can possess multiple operational modes with specific continuous dynamics in each mode. These biological systems are naturally modeled as hybrid automata, most often with nonlinear continuous dynamics. However, hybrid automata are notoriously hard to analyze --- even simple reachability for hybrid systems with linear differential dynamics is an undecidable problem. In this paper we present a parameter synthesis framework based on $\delta$-complete decision procedures that sidesteps undecidability. We demonstrate our method on two highly nonlinear hybrid models of the cardiac cell action potential. The results show that our parameter synthesis framework is convenient and efficient, and it enabled us to select a suitable model to study and identify crucial parameter ranges related to cardiac disorders.
BibTeX
@inproceedings{liu2014parameter,
  author = {Bing Liu and Soonho Kong and Sicun Gao and Paolo Zuliani and Edmund M. Clarke},
  title = {{Parameter Synthesis for Cardiac Cell Hybrid Models Using Delta-Decisions}},
  booktitle = {CMSB},
  year = {2014},
  url = {https://openreview.net/forum?id=fJBvWP-A1l},
}
Figure