Effect of Synaptic Heterogeneity on Neuronal Coordination

Moritz Layer, Moritz Helias, David Dahmen

Abstract

Recent advancements in measurement techniques have resulted in an increasing amount of data on neural activities recorded in parallel, revealing largely heterogeneous correlation patterns across neurons. Yet, the mechanistic origin of this heterogeneity is largely unknown because existing theoretical approaches linking structure and dynamics in neural circuits are restricted to population-averaged connectivity and activity. Here we present a systematic inclusion of heterogeneity in network connectivity to derive quantitative predictions for neuron-resolved covariances and their statistics in spiking neural networks. Our study shows that the heterogeneity in covariances is not a result of variability in single-neuron firing statistics but stems from the ubiquitously observed sparsity and variability of connections in brain networks. Linear-response theory maps these features to the effective connectivity between neurons, which in turn determines neuronal covariances. Beyond-mean-field tools reveal that synaptic heterogeneity modulates the variability of covariances and thus the complexity of neuronal coordination across many orders of magnitude.

Introduction: The Heterogeneity Problem

Modern neuroscience can record from thousands of neurons simultaneously, revealing a significant "heterogeneity problem." Even within a single population, pairs of neurons exhibit widely varying correlation patterns. Traditional "mean-field" theories often simplify this by focusing on population averages, which fails to explain the broad distribution of activity seen in real data. This study addresses whether the statistics of individual neuron-to-neuron coordination (covariances) can be predicted solely from the statistics of the network's connectivity.

Spiking activity and distribution of statistics in the simulated E-I network.

Methods and Modeling

The researchers utilized a network of 10,000 Leaky Integrate-and-Fire (LIF) neurons (8,000 excitatory, 2,000 inhibitory) in a balanced state with 10% random connectivity. To bridge the gap between structure and dynamics, they mapped nonlinear spiking dynamics onto a linear rate model. By applying linear-response theory and advanced statistical physics techniques—specifically the replica method used in the study of disordered systems—they derived the second moment (variance) of covariances. This "beyond-mean-field" approach allows for a description of how coordination patterns are distributed across the network.

Key Results: Predicting Coordination from Connectivity

The theoretical framework demonstrates high predictive accuracy when compared against high-fidelity synthetic data from the NEST simulator. The diversity in neuronal co-fluctuations is primarily driven by the randomness and sparsity of synaptic connections rather than intrinsic differences in the neurons themselves.

Comparison of theoretical predictions versus simulation results for firing rates, covariances, and correlations.

As shown in the validation plots, the theory's predictions for firing rates and covariances closely match simulations, with correlation coefficients of approximately $\rho \approx 0.98$ for rates and $\rho \approx 0.69$ for covariances and correlations. A critical finding is that the spectral radius of the connectivity matrix acts as a "complexity dial": as it approaches unity, the variance of covariances increases by several orders of magnitude, leading to highly specific coordination patterns.

Discussion and Future Directions

The link between activity statistics and connectivity statistics suggests a powerful potential for inference: by measuring the variance of neural correlations, researchers may be able to infer the underlying heterogeneity of synaptic wiring.

However, the current framework has limitations. It relies on a linear approximation that may deviate in highly nonlinear regimes (e.g., near instability). Additionally, it assumes independent synaptic weights, whereas biological brains often exhibit specific connectivity motifs. Future work will focus on incorporating these non-random structures and exploring regimes where nonlinearities dominate.

Summary updated

Abstract

Recent advancements in measurement techniques have resulted in an increasing amount of data on neural activities recorded in parallel, revealing largely heterogeneous correlation patterns across neurons. Yet, the mechanistic origin of this heterogeneity is largely unknown because existing theoretical approaches linking structure and dynamics in neural circuits are restricted to population-averaged connectivity and activity. Here we present a systematic inclusion of heterogeneity in network connectivity to derive quantitative predictions for neuron-resolved covariances and their statistics in spiking neural networks. Our study shows that the heterogeneity in covariances is not a result of variability in single-neuron firing statistics but stems from the ubiquitously observed sparsity and variability of connections in brain networks. Linear-response theory maps these features to the effective connectivity between neurons, which in turn determines neuronal covariances. Beyond-mean-field tools reveal that synaptic heterogeneity modulates the variability of covariances and thus the complexity of neuronal coordination across many orders of magnitude.
BibTeX
@inproceedings{layer2024effect,
  author = {Moritz Layer and Moritz Helias and David Dahmen},
  title = {{Effect of Synaptic Heterogeneity on Neuronal Coordination}},
  booktitle = {PRX},
  year = {2024},
  doi = {10.1103/prxlife.2.013013},
}
Figure