Bayesian Inference of Synaptic Quantal Parameters from Correlated Vesicle Release

Area of Science:

• Neuroscience
• Computational Biology
• Biophysics

Background:

• Synaptic transmission is inherently variable due to history-dependence and stochasticity.
• Accurate inference of synaptic parameters is challenging with traditional methods.
• Existing Bayesian approaches improve parameter inference but often neglect response correlations.

Purpose of the Study:

• To develop a more complete Bayesian framework for synaptic parameter inference.
• To incorporate correlations between serial synaptic events into parameter estimation.
• To provide a tractable method for calculating likelihoods in correlated post-synaptic potential trains.

Main Methods:

• Developed a compact mathematical form for likelihood calculation using matrix products.
• Applied Bayesian inference to models of synaptic dynamics, considering correlations.
• Utilized recent theoretical advances to make likelihood calculations feasible.

Main Results:

• The new Bayesian approach significantly improves parameter inference, especially for synapses with high release probability or limited data.
• Marginals of the posterior distribution reveal covariance information for parameter distributions.
• Demonstrated improved inference compared to mean-variance fitting and simpler Bayesian methods.

Conclusions:

• The developed method offers a powerful tool for inferring quantal and dynamical synaptic parameters from experimental data.
• Accounting for correlations in synaptic responses enhances the precision of parameter estimation.
• The provided computational code facilitates the application of these advanced Bayesian techniques.