Probabilistic inference of short-term synaptic plasticity in neocortical microcircuits

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Short-term synaptic plasticity (STP) exhibits significant diversity across various biological contexts, including brain regions, cortical layers, cell types, and developmental stages.
• This diversity suggests a specialized role for STP in neural information processing, making its accurate characterization critical for understanding and modeling neural systems.
• Existing phenomenological models often rely on least-mean-square fitting, which can produce unreliable results for typical synaptic dynamics.

Purpose of the Study:

• To address the limitations of current methods in characterizing synaptic dynamics.
• To introduce and validate a Bayesian formulation for inferring model parameters from experimental data.
• To develop a method for classifying synaptic dynamics based on connection-specific distributions.

Main Methods:

• Developed a Bayesian framework to determine the posterior distribution of model parameters given experimental data.
• Proposed a novel experimental protocol designed to enhance the accuracy of synaptic dynamics parameter determination.
• Applied the Bayesian approach to analyze experimental data from three distinct neocortical excitatory connection types.

Main Results:

• Demonstrated that conventional short-term plasticity (STP) protocols can lead to broad posterior distributions for certain model parameters.
• The proposed experimental protocol facilitates more precise estimation of synaptic dynamics parameters.
• Inferred connection-specific parameter distributions for neocortical excitatory connections, enabling classification of synaptic dynamics.

Conclusions:

• The Bayesian formulation provides a more robust method for characterizing short-term synaptic plasticity compared to traditional least-mean-square fitting.
• The developed approach successfully classifies synaptic dynamics based on inferred connection-specific distributions.
• This work offers an improved method for analyzing existing neural data, potentially uncovering novel insights into synaptic function.