SMoRe GloS: An efficient and flexible framework for inferring global sensitivity of agent-based model parameters

Area of Science:

• Computational Biology
• Systems Biology
• Computational Science

Background:

• Agent-based models (ABMs) simulate complex systems but require robust uncertainty quantification.
• Global sensitivity analysis (GSA) assesses ABM reliability but is computationally expensive.
• Existing GSA methods are often impractical for complex, resource-intensive ABMs.

Purpose of the Study:

• Introduce SMoRe GloS (Surrogate Modeling for Recapitulating Global Sensitivity), a computationally efficient GSA method for ABMs.
• Enable accurate uncertainty quantification and parameter space exploration for complex ABMs.
• Demonstrate the method's compatibility and performance against established GSA techniques.

Main Methods:

• Developed SMoRe GloS utilizing explicitly formulated surrogate models for ABM analysis.
• Applied SMoRe GloS to a 2D cell proliferation assay and a 3D vascular tumor growth model.
• Compared SMoRe GloS performance with Morris one-at-a-time and eFAST methods.

Main Results:

• SMoRe GloS accurately recovered global sensitivity indices for both biological ABMs.
• Achieved substantial computational speedups (minutes vs. days) compared to eFAST for complex ABMs.
• Successfully estimated sensitivities for parameters not explicitly in the surrogate model.

Conclusions:

• SMoRe GloS provides a computationally feasible solution for GSA in complex ABMs.
• Enhances uncertainty quantification and confidence in predictions from computationally expensive models.
• Facilitates deeper exploration of model behavior and parameter influence in biological systems.