We developed a model of primary motor cortex (M1) microcircuits  with over 10,000 biophysically detailed neurons and 30 million synaptic connections. It simulates a cylindric cortical volume with a depth of 1350 μm and a diameter of 300 μm. Neuron densities, classes, morphology and biophysics, and connectivity at the long-range, local and dendritic scale were derived from experimental data published in over 30 studies. The model was developed using the NetPyNE tool, which facilitated the integration of this complex experimental data at multiple scales. Our model exhibited spontaneous neural activity patterns and oscillations consistent with M1 data. Neural activity depended on cell class, cortical layer and sublaminar location. Different output dynamics were seen when the network was driven by brief activation of particular long-range inputs, or in the setting of different neuromodulatory conditions. Results yielded insights into circuit information pathways, oscillatory coding mechanisms and the role of HCN in modulating corticospinal output.
LFP revealed physiological oscillations in delta (0.5-4 Hz) and high beta to low gamma (25-40 Hz) ranges across layers and populations. Oscillations occurred in the absence of rhythmic external inputs, emergent from neuronal biophysical properties and circuit connectivity. Filtering the LFP signal from the electrode located in upper L5B revealed phase-amplitude coupling of fast oscillations on delta wave phase. LFP spectrogram demonstrated that the fast oscillations occurred robustly during the time course of simulations. Strong LFP beta and gamma oscillations are characteristic of motor cortex activity, and have been found to enhance signal transmission in mouse neocortex. Phase-amplitude coupling may help integrate information across temporal scales and across networks.
Analysis of firing dynamics and information flow in our model confirmed and extended our understanding of information flow in cortical microcircuits. Consistent with existing models, sensory-related long-range inputs targeted superficial layers which in turn projected to deeper layers. Our simulations, however, provided further details: information flow was cell-class specific, going unidirectionally from IT to PT cells; sublaminar-specific, with superficial ITs targeting primarily the upper portion of L5B PT cells; and oscillation frequency-specific, with Granger causality peaks occurring at shifted beta/gamma range frequencies for different internal connections.
The model parameters were constrained by literature using data from over 40 publications. The model results have been compared against in vitro and in vivo experimental data. See publication: https://www.biorxiv.org/content/10.1101/201707v4
The tool (NetPyNE) to develop the model has been validated with a methods publication: https://elifesciences.org/articles/44494
NetPyNE tool for multiscale modeling
Dura-Bernal S, Neymotin SA, Suter BA, Shepherd GMG, Lytton WW. Multiscale dynamics and information flow in a data-driven model of the primary motor cortex microcircuit. bioRxiv. 2019. p. 201707. doi:10.1101/201707
Dura-Bernal S, Neymotin SA, Kerr CC, Sivagnanam S, Majumdar A, Francis JT, et al. Evolutionary algorithm optimization of biological learning parameters in a biomimetic neuroprosthesis. IBM J Res Dev. 2017;61: 6.1–6.14.
Gleeson P, Cantarelli M, Marin B, Quintana A, Earnshaw M, Sadeh S, et al. Open Source Brain: A Collaborative Resource for Visualizing, Analyzing, Simulating, and Developing Standardized Models of Neurons and Circuits. Neuron. 2019. doi:10.1016/j.neuron.2019.05.019
Alber M, Buganza Tepole A, Cannon WR, De S, Dura-Bernal S, Garikipati K, et al. Integrating machine learning and multiscale modeling-perspectives, challenges, and opportunities in the biological, biomedical, and behavioral sciences. NPJ Digit Med. 2019;2: 115.
Peng GCY, Alber M, Buganza Tepole A, Cannon WR, De S, Dura-Bernal S, et al. Multiscale Modeling Meets Machine Learning: What Can We Learn? Arch Comput Methods Eng. 2020. doi:10.1007/s11831-020-09405-5
Sivagnanam S, Gorman W, Doherty D, Neymotin SA, Fang S, Hovhannisyan H, et al. Simulating large-scale models of brain neuronal circuits using Google Cloud Platform. 2020. doi:10.31219/osf.io/m8vza. PEARC'20
Cecilia Romaro Fernando Araujo Najman William W Lytton Antonio C Roque Salvador Dura-Bernal. NetPyNE implementation and rescaling of the Potjans-Diesmann cortical microcircuit model. Submitted to Neural Computation. 2020.