Synapses learn to utilize stochastic pre-synaptic release for the prediction of postsynaptic dynamics

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Synaptic transmission in the brain exhibits significant trial-by-trial variability, a form of noise.
• High energy consumption of synapses makes this noise surprising from an efficiency standpoint.

Purpose of the Study:

• To propose and investigate the hypothesis that synaptic noise is utilized to encode uncertainty about postsynaptic neuron activity.
• To develop a mathematical framework modeling synapse-soma interaction as an agent in an uncertain environment.

Main Methods:

• Developed a mathematical framework to model synapse-soma interactions.
• Proposed that synapses employ an internal model of somatic membrane dynamics.
• Incorporated a synaptic learning rule analogous to long-term potentiation/depression (LTP/LTD).

Main Results:

• The framework suggests synapses use noisy release to encode uncertainty about somatic potential.
• Individual synapses aim to predict postsynaptic neuron dynamics.
• Emergent network dynamics of multiple synapses solve learning problems like pattern classification and control.

Conclusions:

• Synaptic noise is not merely variability but a mechanism for representing uncertainty.
• Synapses coordinate network-level uncertainty representation for behavior in ambiguous situations.
• This framework offers a novel perspective on synaptic function and neural computation.