Efficient parameterization of large-scale dynamic models based on relative measurements.

Area of Science:

• Systems Biology
• Computational Biology
• Biochemical Reaction Networks

Background:

• Mechanistic models are crucial for understanding biological processes and integrating data.
• Large-scale models with comprehensive reaction networks present parameter estimation challenges, especially with relative data.

Purpose of the Study:

• To propose a novel hierarchical approach for efficient parameter estimation in large-scale biochemical models.
• To address challenges in parameterizing models using heterogeneous, relative-scale measurements.

Main Methods:

• A hierarchical approach combining analytic evaluation of scaling/offset parameters with adjoint sensitivity analysis for objective function gradients.
• Parameterization of a large-scale pan-cancer ordinary differential equation model (>1000 state variables, >4000 parameters).
• Utilized relative protein, phosphoprotein, and viability measurements.

Main Results:

• The hierarchical formulation significantly improved optimizer performance.
• Enabled estimation of error model parameters with negligible computational overhead.
• Provided an unbiased method for weighting heterogeneous data.

Conclusions:

• The proposed hierarchical approach is broadly applicable to various models.
• Facilitates efficient parameterization of large-scale models using heterogeneous relative measurements.