Systematic parameter estimation in data-rich environments for cell signalling dynamics.

Area of Science:

• Systems Biology
• Computational Biology
• Biochemistry

Background:

• Ordinary differential equation (ODE) models are crucial for understanding cellular dynamics.
• Traditional parameter estimation relies on matching simulated and observed concentrations.
• Advancements in proteomics enable comprehensive molecular concentration measurements, offering new modeling opportunities.

Purpose of the Study:

• To develop a novel computational method for parameter estimation in biological signaling networks.
• To leverage proteomic data for a more efficient and decomposed optimization of rate parameters.
• To introduce the Spline-based Parameter Estimation via Derivative Matching (SPEDRE) method.

Main Methods:

• SPEDRE fits spline curves to observed concentration data to estimate derivatives.
• It matches these derivatives to species production and consumption rates within ODE models.
• The method decomposes high-dimensional optimization into low-dimensional factors solved via loopy belief propagation and local optimization.

Main Results:

• SPEDRE reformulates parameter estimation, enabling extreme decomposition of optimization problems.
• The method exhibits polynomial runtime with respect to molecules and timepoints.
• Performance was evaluated on a novel Akt activation dynamics model, including PTEN inactivation.

Conclusions:

• SPEDRE offers a novel and efficient approach to parameter estimation in complex biological networks.
• The method's decomposition strategy addresses challenges in high-dimensional optimization.
• SPEDRE facilitates more accurate computational modeling of cellular signaling pathways.