Approximate Bayesian inference in a model for self-generated gradient collective cell movement.

Area of Science:

• Computational Biology
• Mathematical Modeling
• Biophysics

Background:

• Parameter inference for complex biological models, like cell movement in chemotactic gradients, is challenging.
• Stochastic processes in cell movement models often make traditional likelihood-based inference computationally intractable.
• Approximate Bayesian computation (ABC) offers a viable alternative for models with computationally expensive or intractable likelihoods.

Purpose of the Study:

• To evaluate and compare the performance of various approximate Bayesian computation (ABC) algorithms.
• To identify the most suitable ABC methods for parameter inference in a novel hybrid discrete-continuum cell movement model.
• To assess the accuracy and bias of ABC methods against a benchmark stochastic differential equation (SDE) model.

Main Methods:

• Utilized a hybrid discrete-continuum model for cell population movement in response to chemotaxis.
• Employed approximate Bayesian computation (ABC) techniques for parameter inference due to model complexity.
• Used a drift-diffusion stochastic differential equation (SDE) as a benchmark for algorithm evaluation.
• Assessed the accuracy and bias of ABC-posteriors compared to the exact posterior of the SDE model.

Main Results:

• Identified and ranked the top-performing ABC algorithms for parameter inference in the benchmark SDE model.
• Demonstrated the applicability of selected high-performing ABC algorithms to the complex cell movement model.
• Quantified the bias between ABC-derived posteriors and the true posterior distribution for the SDE.

Conclusions:

• The study provides a comparative analysis of ABC algorithm efficiency in the context of biological modeling.
• Selected ABC methods are effective for parameter inference in sophisticated cell movement models.
• This work aids in selecting appropriate computational methods for analyzing complex biological systems.