[Paper Review] Two-compartment neuronal spiking model expressing brain-state specific apical-amplification, -isolation and -drive regimes

Elena Pastorelli, Alper Yegenoglu+6 more|arXiv (Cornell University)|Nov 10, 2023

Neural dynamics and brain function4 citations

TL;DR

This paper introduces a biologically plausible two-compartment spiking neuron model that implements brain-state-specific apical mechanisms—amplification during wakefulness, isolation during deep NREM sleep, and drive during REM sleep—using a piecewise linear transfer function (ThetaPlanes) optimized via a machine learning-driven evolutionary algorithm on high-performance computing infrastructure. The model enables large-scale simulations of state-dependent learning and is natively compatible with major simulation platforms like NEST, Neuron, and Brian.

ABSTRACT

Mounting experimental evidence suggests that brain-state-specific neural mechanisms, supported by connectomic architectures, play a crucial role in integrating past and contextual knowledge with the current, incoming flow of evidence (e.g., from sensory systems). These mechanisms operate across multiple spatial and temporal scales, necessitating dedicated support at the levels of individual neurons and synapses. A notable feature within the neocortex is the structure of large, deep pyramidal neurons, which exhibit a distinctive separation between an apical dendritic compartment and a basal dendritic/perisomatic compartment. This separation is characterized by distinct patterns of incoming connections and brain-state-specific activation mechanisms, namely, apical amplification, isolation, and drive, which are associated with wakefulness, deeper NREM sleep stages, and REM sleep, respectively. The cognitive roles of apical mechanisms have been demonstrated in behaving animals. In contrast, classical models of learning in spiking networks are based on single-compartment neurons, lacking the ability to describe the integration of apical and basal/somatic information. This work aims to provide the computational community with a two-compartment spiking neuron model that incorporates features essential for supporting brain-state-specific learning. This model includes a piece-wise linear transfer function (ThetaPlanes) at the highest abstraction level, making it suitable for use in large-scale bio-inspired artificial intelligence systems. A machine learning evolutionary algorithm, guided by a set of fitness functions, selected the parameters that define neurons expressing the desired apical mechanisms.

Motivation & Objective

• To develop a computationally efficient, two-compartment spiking neuron model that captures brain-state-specific apical dynamics (amplification, isolation, drive) observed in cortical pyramidal neurons.
• To bridge the gap between single-compartment models and biological realism by incorporating apical dendritic mechanisms critical for state-dependent learning and memory consolidation.
• To enable large-scale simulations of brain-state-dependent neural dynamics by integrating the model into standard simulation frameworks like NEST, Neuron, and Brian.
• To demonstrate that evolutionary optimization via a learning-to-learn (L2L) framework can efficiently identify parameter sets supporting desired neural dynamics across multiple brain states.

Proposed method

• The model uses a two-compartment structure with separate somatic and apical dendritic compartments, each governed by distinct integrate-and-fire dynamics with voltage-dependent ion channels.
• A piecewise linear transfer function, termed ThetaPlanes, is introduced to capture the nonlinear input-output transformation characteristic of apical dendritic Ca2+ spikes and burst firing.
• An evolutionary algorithm, driven by a set of fitness functions, explores the parameter space to identify neuron configurations that reproduce brain-state-specific behaviors.
• The optimization is performed on high-performance computing (HPC) infrastructure using the Learning-to-Learn (L2L) framework, enabling efficient, parallelized exploration of complex parameter dependencies.
• The model is implemented in the NEST simulation environment, ensuring compatibility with large-scale network simulations and extensibility to additional compartments (e.g., for NMDA spikes).
• Fitness functions are designed to enforce state-specific dynamics: high-frequency bursting during wakefulness (apical amplification), low apical activity during deep sleep (isolation), and sustained apical drive during REM-like states.

Experimental results

Research questions

• RQ1 Can a two-compartment spiking neuron model reproduce the three distinct apical dynamics—amplification, isolation, and drive—associated with wakefulness, deep NREM sleep, and REM sleep, respectively?
• RQ2 How can a machine learning-driven evolutionary algorithm efficiently identify biologically plausible parameter sets for such a model across multiple brain states?
• RQ3 To what extent can a piecewise linear transfer function (ThetaPlanes) accurately capture the nonlinear dynamics of apical dendritic spikes without sacrificing computational efficiency?
• RQ4 How does the inclusion of apical mechanisms in a two-compartment model improve learning and network dynamics compared to single-compartment models in state-dependent tasks?
• RQ5 Can this model be natively integrated into established large-scale neural simulation platforms like NEST, Neuron, and Brian for whole-brain modeling?

Key findings

• The L2L framework successfully identified parameter sets that reproduce the three distinct apical dynamics—amplification, isolation, and drive—across different brain states, demonstrating the feasibility of evolutionary optimization for complex neural models.
• The resulting two-compartment model with the ThetaPlanes transfer function accurately captures the interplay between somatic action potentials and dendritic Ca2+ spikes, enabling state-specific bursting and modulation.
• The model maintains compatibility with the Adaptive Exponential Integrate-and-Fire (AdEx) framework, ensuring consistency with existing large-scale simulations and mean-field models.
• The model is natively implementable in NEST, enabling direct integration into large-scale network simulations and future extension to additional compartments such as those supporting NMDA spikes.
• The use of HPC and L2L allowed for efficient, parallel exploration of a high-dimensional parameter space, identifying optimal configurations with minimal manual tuning.
• The results suggest that evolution could have easily discovered the cognitive advantages of apical mechanisms using a simple two-compartment geometry, supporting the hypothesis of incremental morphological complexity in cortical pyramidal neurons.