An efficient analytical reduction of detailed nonlinear neuron models.

Area of Science:

• Computational Neuroscience
• Computational Biology
• Neuroscience

Background:

• Detailed neuron models are crucial for understanding neural computation.
• Simulating these complex models is computationally intensive, limiting their application.
• Existing methods struggle with the scale required for modern neuroscience research.

Purpose of the Study:

• To introduce Neuron_Reduce, an analytical method for simplifying nonlinear neuron models.
• To reduce the morphological complexity and computational time of neuron simulations.
• To enable large-scale, accurate neural network simulations.

Main Methods:

• Neuron_Reduce maps synapses and active membrane channels to a reduced model, preserving transfer impedance.
• Synapses with identical transfer impedance are merged while retaining activation times.
• The approach focuses on simplifying morphological complexity for faster computation.

Main Results:

• Neuron_Reduce accelerates simulations by 40-250 fold for various cell types.
• The reduced models accurately replicate voltage dynamics and dendritic computations.
• Simulations used realistic numbers of synapses (10,000-100,000).

Conclusions:

• Neuron_Reduce significantly enhances simulation speed for complex neuron models.
• The method allows for realistic neural network simulations at unprecedented scales.
• This tool facilitates research in micro-connectomics and biologically-inspired deep networks.