GpABC: a Julia package for approximate Bayesian computation with Gaussian process emulation.

Area of Science:

• Systems biology
• Computational biology
• Statistical inference

Background:

• Approximate Bayesian computation (ABC) is crucial for parameter inference in systems biology models, particularly for those with intractable likelihood functions.
• Stochastic and nonlinear dynamics in biological systems often necessitate ABC methods.
• High computational costs associated with standard ABC limit its application to models with rapid simulation times.

Purpose of the Study:

• To present GpABC, a Julia package for parameter inference and model selection.
• To implement efficient ABC methods, including Gaussian process emulation.
• To reduce the computational burden of ABC in systems biology.

Main Methods:

• Implementation of standard rejection ABC and sequential Monte Carlo ABC.
• Integration of Gaussian process emulation within the ABC framework.
• Development of a Julia package (GpABC) for flexible model analysis.

Main Results:

• GpABC offers parameter inference and model selection for deterministic and stochastic models.
• Gaussian process emulation significantly reduces the computational cost of ABC.
• The package supports both standard and advanced ABC techniques.

Conclusions:

• GpABC provides an efficient computational framework for systems biology modeling.
• Gaussian process emulation is a powerful strategy for accelerating ABC.
• The package facilitates broader application of ABC in analyzing complex biological systems.