A doubly stochastic renewal framework for partitioning spiking variability.

Area of Science:

• Computational Neuroscience
• Neural Coding
• Statistical Neuroscience

Background:

• Estimating neural firing rates is crucial for understanding brain function but is challenging with irregular spike trains.
• The standard inhomogeneous Poisson process model fails to capture the full range of neuronal spiking irregularity.
• Diverse spike statistics across neurons necessitate more flexible models for partitioning neural variability.

Purpose of the Study:

• To introduce a novel mathematical framework for partitioning spiking variability in neurons.
• To develop a method for accurately estimating spiking irregularity from neural data.
• To investigate the factors influencing spiking irregularity in cortical neurons and neural networks.

Main Methods:

• Introduced a doubly stochastic renewal point process, a flexible framework for modeling point processes.
• Validated the framework using intracellular voltage recordings from cortical neurons.
• Developed and applied a data-driven method for estimating spiking irregularity.
• Utilized spiking network models to explore the relationship between connectivity, input, and spiking irregularity.

Main Results:

• The new model captures a broad spectrum of spiking irregularity, from periodic to super-Poisson.
• Spiking irregularity in cortical neurons decreases from sensory to association areas.
• Spiking irregularity is generally stable for individual neurons but can vary with task epochs.
• Network models demonstrate that spiking irregularity is influenced by neuronal connectivity and external input.

Conclusions:

• The doubly stochastic renewal point process provides a powerful tool for analyzing neural spike trains.
• The findings offer insights into the spatial and dynamic changes of spiking irregularity in the cortex.
• This work enhances the precision of single-trial firing rate estimation and constrains mechanistic models of neural circuits.