PyBioNetFit

PyBioNetFit performs parameter estimation, uncertainty quantification, model checking, and Bayesian inference for systems biology models specified in BNGL and SBML.

Key Features:

• Parameterization: Employs parallelized metaheuristic optimization algorithms to fit parameters of BNGL and SBML models to quantitative and qualitative data, demonstrated on a 153-parameter yeast cell cycle control model.
• Model Checking and Design: Implements the Biological Property Specification Language (BPSL) to formally declare system properties and incorporate qualitative data for model checking and design applications.
• Bayesian Inference: Provides a Markov chain Monte Carlo method ("am") with an adaptive move proposal distribution for sampling multidimensional and non-Gaussian posterior distributions.
• Parallel Processing: Supports parallel generation of multiple MCMC chains on computer clusters to accelerate Bayesian inference and uncertainty quantification.
• Integration and Simulation Support: Integrates with libRoadRunner and BioNetGen for simulation, including numerical integration of ordinary differential equations defined in SBML or BNGL files, and uses Dask.Distributed for task scheduling on Linux clusters.

Scientific Applications:

• Cell cycle control parameterization: Demonstrated by parameterization of a 153-parameter yeast cell cycle control model using quantitative and qualitative data.
• Drug intervention analysis in autophagy signaling: Applied BPSL-based model checking and design to analyze targeted drug interventions in autophagy signaling models.
• Epidemiological forecasting: Used Bayesian inference with parallel MCMC to forecast new Coronavirus Disease 2019 case detection with quantified forecast uncertainty.

Methodology:

Uses parallelized metaheuristic optimization, the Biological Property Specification Language (BPSL), the "am" MCMC method with adaptive move proposals, parallel MCMC chains on clusters, libRoadRunner and BioNetGen simulations with numerical integration of ODEs from SBML/BNGL, and Dask.Distributed for task scheduling.