Experiment design through dynamical characterisation of non-linear systems biology models utilising sparse grids

Area of Science:

• Systems Biology
• Computational Biology
• Experimental Design

Background:

• Biological systems exhibit complex dynamics with inherent uncertainties in parameters and structure.
• Traditional experiment design often focuses on parameter uncertainty, neglecting structural ambiguity.
• Ill-posed problems in systems biology are common, especially when data is limited relative to uncertain parameters.

Purpose of the Study:

• To develop a novel experiment design algorithm using sparse grids to discriminate between competing hypotheses.
• To simultaneously address both parameter and structural uncertainty in complex models.
• To efficiently explore the global uncertain parameter space for model structure and parameter value resolution.

Main Methods:

• Sequential selection of experimental design points using a sparse grid-based algorithm.
• Screening of the global uncertain parameter space to identify acceptable parameter subspaces.
• Clustering of parameter vectors based on simulated model trajectories to characterize data-compatible dynamics.
• Leveraging the diversity of system output dynamics to select design points that distinguish between hypotheses.

Main Results:

• The algorithm successfully identified key design points to differentiate between model structures and parameter values.
• Demonstrated on a mitogen-activated protein kinase cascade model, it resolved significant uncertainty.
• Required only three additional experimental data points to achieve discrimination.

Conclusions:

• The sparse grid-based experiment design offers a systematic and computationally efficient approach for exploring model uncertainty.
• This method effectively resolves uncertainty in non-linear systems biology models, even with limited data.
• It provides a powerful tool for distinguishing between competing models and their parameters in complex biological systems.