Quantitative comparison of a mesocircuit model with resting state activity in the macaque motor cortex

We aim to derive a full-scale spiking network model [1,2] of part of the macaque motor cortex that statistically reproduces experimentally observed dynamics. For that matter we designed a validation workflow using the Python module NetworkUnit (github.com/INM-6/NetworkUnit) [3]. A validated model will allow to analyze the underlying structural connectivity and enable in-silico experiments, e.g. modification of structural connectivity to probe its influence on neural activity. Here, we describe the characteristics of the experimental resting state data from macaque motor cortex that is used for comparison and present preliminary results for a generic version of the network model.The network model is an upscaled version of the layered spiking network model by Potjans & Diesmann (2014) extended to a size of 4x4mm2 and a total of ~1.2 million leaky integrate-and-fire neurons [2]. In contrast to the original model this mesocircuit model uses lateral distance-dependent connection probabilities estimated from cortical neuroanatomical data. We subsample single unit activities from the corresponding layer in the simulated network with the same number of neurons and with the same spatial arrangement of the recording array as in the experimental data.The model describes a system in ground, idle or resting state with uncorrelated input. In order to perform a quantitative comparison with experimental data we therefore conducted a resting state experiment with macaque monkeys not given any specific task or stimulus. We recorded neuronal activity from premotor and motor cortex using a chronically implanted 4x4mm2 Utah array with 100 electrodes [4]. A video of the monkey was used to differentiate between periods of rest and spontaneous movements.The quantitative comparison with the model suggests that average quantities of firing rates, inter-spike intervals and local coefficients of variation are qualitatively comparable (effect sizes References[1] Potjans and Diesmann (2014) The cell-type specific cortical microcircuit: Relating structure and activity in a full-scale spiking network model. Cereb. Cort., 24(3), 785-806, 10.1093/cercor/bhs358[2] Senk et al. (2016) Reconciliation of weak pairwise spike-train correlations and highly coherent local field potentials across space. arXiv:1805.10235 [q-bio.NC][3] Gutzen et al. (2018) Reproducible neural network simulations: model validation on the level of network activity data. Front. Neuroinf., submitted[4] Brochier et al. (2018) Massively parallel recordings in macaque motor cortex during an instructed delayed reach-to-grasp task. Scientific Data 5:180055, 10.1038/sdata.2018.55[5] Dahmen et al. (2017) Two types of criticality in the brain, arXiv:1711.10930 [cond-mat.dis-nn