TopoFilter

TopoFilter performs topological-filtering Bayesian model selection to identify minimal or alternative mechanistic models consistent with experimental data in systems biology.

Key Features:

• Topological filtering for Bayesian model selection: Implements a topological filtering methodology as the core approach for Bayesian model selection.
• Heuristic automated submodel search: Employs a heuristic, automated search to explore submodels within a parametrized model framework.
• Sampling-based parameter-space analysis: Uses sampling techniques to explore parameter space for robust model selection.
• Efficient parameter resampling: Supports efficient re-sampling of parameters to iteratively refine candidate models.
• Balanced exhaustiveness and speed: Implements strategies to balance thoroughness of model-space exploration with computational speed.
• Parallelization: Supports parallel processing to accelerate searches across complex network ensembles.
• Custom scoring functions: Allows user-defined scoring functions to tailor model-selection criteria.

Scientific Applications:

• Mechanistic model identification: Identification and reduction of mechanistic models consistent with experimental data in systems biology.
• Uncertainty management in network models: Addressing uncertainty in biological knowledge and experimental measurements during model selection and reduction.
• Yeast signaling network analysis: Demonstrated on a yeast signaling network with more than 250,000 possible model structures to evaluate scalable model selection.

Methodology:

Topological filtering–based Bayesian model selection combined with heuristic automated submodel search, sampling-based parameter-space analysis, parameter resampling, parallel computation, and support for custom scoring functions.