Calibration of stochastic, agent-based neuron growth models with approximate Bayesian computation.

Area of Science:

• Computational Neuroscience
• Developmental Neuroscience
• Bayesian Inference

Background:

• Understanding neuronal growth is key to brain architecture.
• Agent-based models (ABMs) simulate neuronal growth but face calibration challenges.
• Accurate model calibration is crucial for reliable simulation results.

Purpose of the Study:

• To present a novel application of Approximate Bayesian Computation (ABC) for calibrating agent-based models (ABMs) of neuronal growth.
• To establish a robust Bayesian framework for neuronal growth model calibration.
• To enable future investigations using Bayesian techniques for model building and verification.

Main Methods:

• Utilized Approximate Bayesian Computation (ABC) within a Bayesian framework to solve the stochastic inverse problem of model calibration.
• Quantified neuronal morphology using morphometrics for data-simulation comparison.
• Employed Sequential Monte Carlo sampling and Wasserstein distance to measure discrepancies between simulated and experimental data.

Main Results:

• Demonstrated that ABC with Sequential Monte Carlo and Wasserstein distance accurately finds posterior parameter distributions for ABMs.
• Showcased that the calibrated ABMs capture key morphological features of hippocampal CA1 pyramidal cells.
• Validated the approach using both synthetic and experimental neuronal growth data.

Conclusions:

• Established a robust framework for calibrating agent-based neuronal growth models using Bayesian inference.
• The proposed ABC method provides accurate parameter estimation for complex neuronal growth simulations.
• This work facilitates advanced model building, verification, and adequacy assessment in neuroscience.