Computational analysis of cortical networks: The role of network structure, cellular and synaptic mechanisms in initiation, maintenance and suppression of network activity

We summarize our theoretical and computational work that addresses mechanisms
responsible for initiation, maintenance and suppressing of spontaneous population activity in cortical networks. The models are primarily motivated by in vitro studies, however the examined mechanisms have more general implications. Three aspects of this work are emphasized: 1) a framework connecting properties of neuronal morphology with graph theoretic concepts of connectivity, 2) a study examining effects of structured connectivity to spontaneous activity in neuronal population, 3) the most recent work integrating these previous results with more detailed models and experimental data. Part 1) describes our work on structural organization of connectivity, deriving the relation between morphometric properties of individual neurons and the graph theoretic description of network structure both in approximative models with mean field-type
characterization of neuron morphology (Aćimović et. al, 2015) and models with
statistically accurate neuronal morphology (Aćimović et. al, 2011). Starting from a homogeneous population, we show how neurite morphology alone constrains the connectivity. More recent analysis also considers non-homogeneities that lead to structures like hub networks and networks containing clusters of strongly coupled neurons. Part 2) incorporates these models of connectivity into spiking neuronal networks to examine the effect of structured connectivity on network activity. The activity consists of spontaneous network bursts (e.g. like those recorded in dissociated in vitro cultures), the periods of intensive synchronized activity spreading across neuronal population, separated by low-activity intervals. Network bursts are quantified by their
frequency and internal structure. Our previous work (see Mäki-Marttunen et. al; 2013) examines how connectivity and neuronal model complexity affect internal burst structure. Our recent unpublished work includes the properties of connectivity that support bursting initiation. Part 3), our most recent study, combines these previous results with more precise description of neuronal dynamics and suitable model fitting protocols to quantitatively reproduce statistics of experimental data. The role of cellular model complexity, short-term presynaptic activity, the contribution of most common glutamatergic and GABAergic synaptic receptors, as well as the details of network connectivity are included in the model. The obtained results are discussed and
compared to the relevant models from the literature summarized in our recent publication (Manninen et. al, 2018).