Sensitivity Analysis
Systematic study of how variation in model parameters or inputs affects model outputs, identifying the most influential factors.
Sensitivity Analysis is a method for quantifying how changes in model parameters propagate to changes in model outputs, revealing which parameters most strongly influence system behavior 1.
How It Works
Local sensitivity analysis computes partial derivatives of model outputs with respect to each parameter at a nominal operating point. These sensitivity coefficients indicate the direction and magnitude of output change for small parameter perturbations. While computationally cheap, local methods miss nonlinear interactions and are only valid near the chosen point.
Global sensitivity analysis explores the entire parameter space simultaneously. Variance-based methods such as Sobol indices decompose total output variance into contributions from individual parameters and their interactions. A high first-order Sobol index means that parameter alone explains much of the output variability; high total-order indices indicate important interactions 1.
In synthetic biology, sensitivity analysis guides circuit design by identifying which rate constants or part characteristics most impact performance metrics like fold-change, response time, or noise levels 2.
Computational Considerations
Global sensitivity analysis requires thousands to millions of model evaluations across the parameter space. Gaussian process emulators and polynomial chaos expansions serve as cheap surrogate models that approximate the full simulator, making Sobol index estimation tractable for computationally expensive models like whole-cell simulations 1.
Woolf Software specializes in computational modeling and simulation for biological systems. Get in touch.
Global sensitivity methods like Sobol indices require extensive sampling; surrogate models and polynomial chaos expansions reduce computational cost by orders of magnitude.