A flexible likelihood approach for predicting neural spiking activity from oscillatory phase

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural oscillations, driven by synchronous ionic currents, exert complex influences on neuronal spiking activity.
• Characterizing these spike-phase relationships is crucial for understanding neural circuit dynamics but remains challenging.

Purpose of the Study:

• To present a novel method for estimating probabilistic relationships between neural spiking activity and field oscillation phases.
• To overcome limitations of existing methods that often only capture unimodal relationships or impose specific shapes.

Main Methods:

• Utilized a generalized linear model (GLM) with an overcomplete basis of circular functions.
• Employed L1-regularized maximum likelihood for regressor selection and standard maximum likelihood estimation for model fitting.
• Applied information-theoretic model selection to identify optimal regressors and coefficients, minimizing overfitting.
• Assessed goodness of fit using the time-rescaling theorem and quantile-quantile plots.

Main Results:

• Successfully characterized spike-phase relationships in synthetic data with robustness.
• Applied the method to in vivo hippocampal data from an awake behaving rat.
• Captured a multimodal relationship between CA1 interneuron spiking activity and nested theta (5-10 Hz) and high gamma (65-135 Hz) rhythms.

Conclusions:

• The developed method advances the visualization and quantification of spike-phase relationships.
• It effectively captures multimodal and multi-rhythm spike-phase interactions, unlike previous approaches.
• This powerful tool enhances the understanding of diverse neural circuit interactions.