Hedyeh Rezaei, Ad Aertsen, Arvind Kumar, Alireza Valizadeh
A fundamental requirement for brain function is the efficient communication between different brain regions. This requirement turns out not to be trivial in the presence of various noise sources in the brain. Two possible strategies taken by the nervous system to face the abundance of noise could be to integrate signals either across time (firing rate) or across a population of neurons (synchrony) to retain the signal in spite of the uncorrelated noisy background. However, the reliable transmission of signals requires either strong and sparse or dense and weak connections for reliable transmission of rate codes or synchrony codes, respectively [1].
However, the typical connectivity between brain regions is neither strong nor dense. Recent work has highlighted the importance of feedback connections. Feedback connections can strengthen the signals through reverberations in the bi-directionally coupled modules [2]. This mechanism depends on the matching of the total effective delay along forward and backward intermodule connections and the period of the local oscillations in the modules, determined by the within and between module connections. This raises the question how the networks in the brain might tune their parameters in a range that favours such reliable and economic signal transmission. Here, we tested if the biological synaptic plasticity rules can self-organize an initially disorganized network to such a tuned regime for reliable signal transmission. Inspired by Hebb’s postulate, we hypothesized that in the presence of abundant synaptic connections between the modules in a developmental stage of the nervous system, only those with matching parameters for reliable transmission can potentiate. While potentiation of these synapses facilitates the reliable transmission of signals, depression of other “unfit” connections reduces the structural cost and gives rise to an efficient substrate for reliable signal transmission. We found that with STDP, the intermodule connections with delays matching the oscillation period of a single network module were potentiated, whereas other connections were ultimately eliminated. We also found how this mechanism facilitated reliable signal transmission to downstream areas in case the network consisted of several (up to 10) such modules. Our results suggest that STDP can lead to the emergence of networks with tuned parameters for reliable and efficient signal transmission out of an initially inefficient network with an extravagant structural cost.
References:
1. Kumar A et al. (2010) Spiking activity propagation in neuronal networks: reconciling different perspectives on neural coding. Nature Rev Neurosci. 11(9):615-27.
2. Rezaei H et al. (2019) Facilitating the propagation of spiking activity in feedforward networks by including feedback. BioRxiv. 712547.
