Estimating and Assessing Differential Equation Models with Time-Course Data.

Area of Science:

• Computational Biology
• Systems Biology
• Mathematical Modeling

Background:

• Ordinary differential equation (ODE) models are crucial for understanding chemical and biological systems.
• Analyzing time-course data from these systems is challenging due to noise, unobserved components, and computational demands of numerical integration.

Purpose of the Study:

• To evaluate the MAnifold-constrained Gaussian process Inference (MAGI) method for ODE model inference using time-course data.
• To demonstrate MAGI's capability in parameter estimation, trajectory inference, and model assessment.

Main Methods:

• Application of the MAGI method to various ODE models using time-course datasets.
• MAGI utilizes Gaussian processes with manifold constraints for inference.
• The method bypasses the need for traditional numerical integration.

Main Results:

• MAGI successfully infers model parameters and system trajectories, including unobserved components, with accurate uncertainty quantification.
• The method efficiently computes model predictions, enabling effective model assessment and selection.
• Demonstrated efficacy across a range of example ODE models.

Conclusions:

• MAGI offers a powerful and computationally efficient alternative for ODE model analysis with time-course data.
• The method addresses key challenges like data noise and unobserved states.
• MAGI facilitates robust model inference and selection without relying on numerical integration.