Generalized leaky integrate-and-fire models classify multiple neuron types.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• The mammalian neocortex exhibits significant neuronal diversity.
• Accurate system models require accurate representation of multiple cell types.
• Existing models may not fully capture the complexity of neuronal firing patterns.

Purpose of the Study:

• To generate a database of simplified point models for diverse neocortical neurons.
• To develop generalized leaky integrate-and-fire (GLIF) models for classifying cell types.
• To facilitate the construction of network models using cell type-specific neuronal models.

Main Methods:

• Constructed GLIF models of varying complexity for 645 neurons from 16 transgenic lines.
• Utilized unsupervised methods for cell type classification based on model parameters.
• Assessed model predictive capacity using hold-out stimuli and compared differentiation ability with electrophysiological features.

Main Results:

• Increasing GLIF model complexity enhanced prediction of spiking behavior.
• High-level GLIF model parameters effectively differentiated transgenic lines, comparable to electrophysiological features.
• Complex GLIF model parameters showed improved ability to distinguish between transgenic lines.

Conclusions:

• Simplified GLIF models serve as an effective dimensionality reduction technique for neuronal electrophysiology.
• Cell type differentiation is achievable from electrophysiological responses without pre-defined features.
• The generated database provides valuable, simplified neuronal models for community use in network simulations.