How heterogeneity shapes dynamics and computation in the brain

David Dahmen, Axel Hutt, Giacomo Indiveri, Ann Kennedy, Jérémie Lefebvre, Luca Mazzucato, Adilson E. Motter, Rishikesh Narayanan, Melika Payvand, Henrike Planert, Richard GastNeuron 2025 · DOI

TL;DR

This paper argues that "within-type" heterogeneity – the variation in properties among neurons of the same class – is not just biological noise but a crucial feature that shapes brain dynamics and computation. By reviewing theoretical frameworks and experimental evidence, the authors propose that this heterogeneity is fundamental to how the brain processes information, adapts, and maintains stability.

What Problem Does This Paper Address?

The field of neuroscience has made significant strides in classifying neurons into distinct types based on their molecular, morphological, and physiological characteristics. However, this focus on categories often overlooks a critical aspect: even within a single cell type, individual neurons exhibit substantial variations in their properties. This "within-type" heterogeneity, akin to "disorder" in statistical physics, has been largely treated as noise or an epiphenomenon. This paper addresses the gap in understanding the functional consequences of this pervasive within-type heterogeneity, arguing that it plays a vital role in neural network dynamics, computation, and the overall organization of the brain.

Methods

The paper is a perspective article, meaning it synthesizes existing research rather than presenting new experimental data or computational models. It draws upon theoretical frameworks from computational neuroscience, statistical physics, and network science. The authors discuss how heterogeneity can be incorporated into computational models, ranging from detailed multi-compartment single-cell models to simpler point neuron models and mean-field models. They also review experimental findings that highlight the effects of heterogeneity in biological systems.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Within-type heterogeneity is more than noise: The paper strongly advocates for viewing within-type variation not as a nuisance but as a fundamental computational resource.
  • Heterogeneity shapes network dynamics: It can promote or suppress synchronization, influence the formation of complex spatiotemporal patterns (like chimera states), and stabilize network dynamics.
  • Linearization of transfer functions: Heterogeneity can make neural populations more sensitive to input fluctuations and linearize their overall response, improving signal detection.
  • Trivialization and stability: Heterogeneity can constrain neural dynamics to fewer, more stable states, making the system less prone to sudden, qualitative shifts.
  • Flexibility and multi-tasking: Intermediate levels of heterogeneity can enhance a neural population's ability to respond flexibly to a broader range of inputs and engage in multiple tasks.
  • Neuromorphic computing as a model: The authors suggest that neuromorphic systems, with their inherent device variability, can serve as excellent physical models for studying the role of heterogeneity.
  • Disease and heterogeneity loss: A loss of neuronal heterogeneity is linked to neurological disorders like epilepsy and Parkinson's disease, suggesting heterogeneity is crucial for resilience.

Key Results

The paper doesn't present quantitative results in the traditional sense, as it's a theoretical perspective. However, it synthesizes findings that demonstrate:

  • Improved signal detection: Heterogeneity can enhance signal detection in feedforward networks by linearizing responses and decorrelating population activity.
  • Stable odor and stimulus representation: Heterogeneity contributes to reliable odor representations in the olfactory bulb and stimulus orientation encoding in the visual cortex.
  • Enhanced sequence learning: Heterogeneous neural networks can achieve better performance in sequence prediction tasks.
  • Resilience to perturbations: Loss of heterogeneity is associated with reduced functional resilience and increased susceptibility to pathological dynamics in neurological disorders.

Limitations and Future Work

The authors acknowledge that experimental characterization of heterogeneity's impact in vivo is sparse, partly due to the difficulty in manipulating it. They highlight several promising future directions:

  • Investigating neuromodulation: Exploring how neuromodulatory systems dynamically control neural heterogeneity.
  • Studying self-organized pattern formation: Examining the interplay between heterogeneity and synaptic plasticity in adaptive neural networks.
  • Developing theoretical frameworks: Creating a robust theoretical basis for empirical research on heterogeneity's role in brain function.
  • Experimental approaches: Utilizing in vitro, in vivo, and neuromorphic systems to test predictions about heterogeneity's computational role.
  • Rethinking cell types: Shifting from viewing cell types as averages to considering them as distributions of properties, where variance and other statistical moments are functionally significant.

Summary updated

Abstract

Much effort has been spent clustering neurons into transcriptomic or functional cell types and characterizing the differences between them. Beyond subdividing neurons into categories, we must recognize that no two neurons are identical and that graded physiological or transcriptomic properties exist within cells of a given type. This often overlooked "within-type" heterogeneity is a specific neuronal implementation of what statistical physics refers to as "disorder" and exhibits rich computational properties, the identification of which may shed crucial insights into theories of brain function. In this perspective article, we address this gap by highlighting theoretical frameworks for the study of neural tissue heterogeneity and discussing the benefits and implications of within-type heterogeneity for neural network dynamics, computation, and self-organization.
BibTeX
@inproceedings{dahmen2025how,
  author = {David Dahmen and Axel Hutt and Giacomo Indiveri and Ann Kennedy and Jérémie Lefebvre and Luca Mazzucato and Adilson E. Motter and Rishikesh Narayanan and Melika Payvand and Henrike Planert and Richard Gast},
  title = {{How heterogeneity shapes dynamics and computation in the brain}},
  booktitle = {Neuron},
  year = {2025},
  doi = {10.1016/j.neuron.2025.11.023},
}
Figure