Examples

We provide a collection of example notebooks to get a better idea of how to use pyABC, and illustrate core features.

The notebooks can be run locally with an installation of jupyter (pip install jupyter), or online on Google Colab or nbviewer, following the links at the top of each notebook. To run the notebooks online, at least an installation of pyABC is required, which can be performed by

# install if not done yet
!pip install pyabc --quiet

Potentially, further dependencies may be required. Unfortunately, at the moment (2022-06), Google Colab is using Python 3.7, while pyABC and many other packages have proceeded to require Python >= 3.8. Thus, not everything may work properly.

Getting started

• Parameter inference
  • Problem definition
  • ABC inference
  • Visualization and analysis of results
• Model selection
  • Defining the model
  • Configuring the ABCSMC run
  • Setting the observed data
  • Running the ABC
  • Visualization and analysis of results

Algorithms and features

• Early stopping of model simulations
• Resuming stored ABC runs
• Custom priors
• Adaptive distances
  • Scale normalization
  • Robustify against model error
  • Conclusion
  • Problems
• Informative distances and summary statistics
  • Simple illustration example
  • Challenging problem features
• Aggregating and weighting diverse data
  • Aggregating diverse distance functions
  • Weighted aggregation
• Wasserstein distances
• Data plots
  • Define summary statistics
  • Plot
• Measurement noise and exact inference
  • True posterior
  • Ignoring noise
  • Add noise to the model output
  • Modify the acceptance step
  • Estimate noise parameters
• Optimal acceptance thresholds
  • Issues of quantile-based acceptance thresholds
  • Optimal threshold based on predicting the acceptance rate
• Discrete parameters
• Look-ahead sampling
  • Sampling
  • Results

External interfaces

• Using R via rpy2
  • Dumping the results to a file format R can read
• Using Julia via pyjulia
• Script based external simulators
• PEtab import and yaml2sbml
  • Generate PEtab problem from YAML
  • Import PEtab problem to AMICI and pyABC
  • Run analysis with exact inference
  • Generate data
• Using COPASI via BasiCO

Application examples

• Ordinary differential equations: Conversion reaction
  • Data
  • ODE model
• Markov jump process: Reaction network
• Multi-scale model: Tumor spheroid growth
  • Installation prerequisites
  • Model description
  • Exemplary pyABC inference run with the model
  • Results of a stored, distributedly executed pyABC inference run
  • Adaptive population sizes
  • The complete stored inference
• Stochastic differential equation: Ion channel noise in Hodgkin-Huxley neurons
  • Download and compilation of the Fortran model
  • Generating the observed data
  • Defining distance and prior
  • Performing parameter inference with pyABC
  • Visualization of the estimated parameters
  • Evaluation of the fit
• PEtab application example

Warning

Upgrade to the latest pyABC version before running the examples. If you installed pyABC some weeks (or days) a ago, some new features might have been added in the meantime. Refer to the Upgrading section on how to upgrade pyABC.