Editorial: Deciphering population neuronal dynamics: from theories to experiments

What are the network mechanisms that may explain such dynamical activities observed at the levels of individual neurons and neuronal populations? It is known that balanced excitation (E) and inhibition (I) within a network is crucial for efficient information processing. For a single neuron, balanced E/I means that excitatory and inhibitory synaptic inputs closely track each other, leading to irregular firing. At the population level, the E/I balance is usually manifested as stable activity propagation among different neurons, avoiding either run-away or diminishing activities and leading to rich collective behaviors such as avalanche dynamics and oscillations. However, it is unclear how asynchronous firings of individual neurons can coexist with synchronous population activities in an E/I balanced state. Liang et al., 2020 built a neuronal network model that exhibits irregular firing, oscillations, and critical avalanche dynamics through changes in only the inhibitory synaptic decay timescale. Importantly, through a mean-field analysis, the authors showed that such an E/I balance is characterized by a stable state associated with a Hopf bifurcation process. These results provide mechanistic insight into how the interaction between excitation and inhibition in a network will determine microscopic and macroscopic dynamics, thereby affecting information processing in the brain.Liang et al. examined how networks of neurons can exhibit oscillations and other coordinated dynamics. However, properties of individual neurons may also contribute to such dynamics. In particular, Li et al., 2022 explored an unusual mechanism that could regulate gamma oscillations, namely the potential role of autapse. The autapse, or auto-synapse, refers to a synaptic connection of a neuron onto itself. Autapses have been found in a number of brain regions, including the cerebral cortex, hippocampus and striatum (e.g., (Bekkers, 2003;Tamás et al., 1997), however, their functional roles remain a mystery. By modeling excitatory and inhibitory autapses as excitatory and inhibitory self-feedback loops, the authors conducted simulation and theoretical analyses to explore their influence on gamma oscillations. The authors found that excitatory self-feedback connections promote the generation of gamma oscillation, while excitatory and inhibitory self-feedback connections regulate oscillation frequency in a complementary manner. This work provides new insights into the neuronal basis of gamma oscillation and suggests a functional role for autapses.The modeling analyses of (Li et al., 2022) focused primarily on neural activity in the absence of sensory stimuli. However, sensory stimuli also shape neural network dynamics. The work of Hashemnia et al., 2021 explored the similarities and differences between the temporal dynamics of human EEG and a recurrent neural network (RNN) model named Deep Speech performing the same speech recognition task. It examined whether the success of deep learning methods in recognizing natural human speech could be attributed to their ability to use time-dependent features, similar to the mechanisms observed in humans during speech perception. By presenting identical speech stimuli to both human listeners and the Deep Speech, the authors found that units, particularly in the recurrent layer, tracked envelope fluctuations of the speech signal in a similar way to the EEG signals measured at frontocentral electrodes. Furthermore, a Representational Dissimilarity Analysis suggested that both human EEG and Deep Speech clustered similarities between sentence representations in the same way. Uncovering shared informational architectures in biological brains and artificial neural networks is important because it helps us understand the computational principles underlying brain dynamics (Luczak et al., 2022;Richards et al., 2019) All these studies integrated experimental findings with theoretical modeling insights to provide new knowledge on the structure of neural dynamics. We hope that these results inspire further research of neural dynamics, including new theories and experiments to test them.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.