Parameter Estimation in Multiple Dynamic Synaptic Coupling Model Using Bayesian Point Process State-Space Modeling...

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Biophysics

Background:

• Neuronal spiking activity is driven by complex synaptic interactions.
• Previous models oversimplify synaptic dynamics and connection numbers.
• Accurate characterization of neural ensembles requires more sophisticated models.

Purpose of the Study:

• To develop a scalable system identification solution for estimating dynamic synaptic connections.
• To incorporate accurate synaptic response dynamics and sparsity into neural models.
• To improve the modeling of individual neuron firing properties within cell ensembles.

Main Methods:

• Proposed a Bayesian point-process state-space model to capture synaptic sparsity.
• Developed an extended expectation-maximization algorithm for parameter estimation.
• Applied the methodology to simulated data and intracellular recordings with 96 presynaptic connections.

Main Results:

• The proposed model accurately estimates parameters of dynamic synaptic connections.
• Demonstrated successful application to simulated data across various parameter ranges.
• Validated estimation accuracy using goodness-of-fit measures on real neural data.

Conclusions:

• The developed Bayesian framework effectively models sparse synaptic connections in neural networks.
• This approach offers a more accurate and scalable method for characterizing neuronal firing properties.
• The methodology advances computational neuroscience by improving the fidelity of neural network models.