A Synthetic Likelihood Solution to the Silent Synapse Estimation Problem.

Area of Science:

• Neuroscience
• Computational Biology
• Synaptic Plasticity

Background:

• Inferring synaptic population function typically relies on electrophysiological sampling, raising questions about sample bias.
• Quantifying silent (AMPAR-lacking) synapses often uses failure-rate analysis, a method whose accuracy is under scrutiny.

Purpose of the Study:

• To develop and validate a biophysically constrained statistical framework to assess sampling bias in synaptic populations.
• To evaluate the performance of failure-rate analysis for quantifying silent synapses.
• To introduce a novel simulation-based approach for unbiased estimation of synaptic properties.

Main Methods:

• In silico simulation of a failure-rate analysis method to identify biases.
• Validation of simulation findings using whole-cell recordings from hippocampal neurons.
• Development of a simulator for experimental protocols to compute synthetic likelihood.

Main Results:

• In silico simulations demonstrated that failure-rate analysis exhibits significant biases, low reliability, and weak statistical power.
• Experimental validation confirmed the limitations of the failure-rate analysis method.
• The novel synthetic likelihood approach, derived from a simulator, provided unbiased estimation with low variance and high statistical power.

Conclusions:

• Standard methods for analyzing synaptic populations, like failure-rate analysis, are subject to systematic biases.
• A simulator-based approach significantly enhances the accuracy and statistical power of estimating silent synapse fractions.
• This generalizable framework offers a powerful tool for improving the estimation of physiological properties from experimental data.