A Model of Rodent Neocortical Micro- and Mesocircuitry

[Version 3; 04/05/24]: Fixes: The afferent_section_type and efferent_section_type synapse properties were previously off by 1. This did not affect simulation results, as these properties are not used by the simulator. It is fixed in this new version. The layer property has been turned into a categorical property instead of a raw string. This should make certain analyses easier, but may require minor updates to analysis code. The strenghts of thalamic input synapses onto excitatory neurons have been re-calibrated to match biological data better.

[Update 12/06/23]: We were made aware that the node_sets.json file included in the first version of this release included invalid node set specifications. We have removed them in this release. If you have already downloaded the first release, you do not have to re-download the entire archive: we have also included the fixed version of the node_sets.json in this v2 - simply replace the existing one. Otherwise the contents of O1_data_physiology.xz is unchanged.

We also included in a separate file the .mod files describing ion channel and synaptic mechanisms used in the model. Finally, we included exemplary simulation configurations that were used in the accompanying publication.

For instructions how to use the .mod files and simulation configuration files, see below.

---

A full description of the model is available in the two companion manuscripts: 

Modeling and Simulation of Neocortical Micro- and Mesocircuitry. Part I: Anatomy

Modeling and Simulation of Neocortical Micro- and Mesocircuitry. Part II: Physiology and Experimentation

We kindly ask that you cite these papers, as well as the Zenodo repository, in any articles or presentations using the model or any of its constituent components.

---

We present a data-driven computational model of the anatomy and physiology of non-barrel primary somatosensory cortex of juvenile rat. The modeling process is based on a previously established workflow for a single cortical column, but is extended here to build a much larger circuit in an atlas-based geometry. Neurons in the model belong to 60 different morphological types and are connected by synapses placed by two established algorithms, one modeling local connectivity determined by axo-dendritic overlap, and one for long-range connectivity between sub-regions. Long-range connectivity is defined with topographic mapping and laminar connectivity profiles, providing intrinsic feed-forward and feedback pathways. Additionally, we incorporate core- and matrix-type thalamocortical projection systems, associated with VPM and POm thalamic nuclei respectively, that enable extrinsic input.

The full model contains over 4 million modeled neurons. Due to the immense size of that data, it is hard to publish it in its entirety. Here we release a 1.5 mm diameter subset of the full volume comprising 211712 neurons in the front limb and jaw subregions and the dysgranular zone of the Paxinos & Watson rat brain atlas.

To get access to the full model (or other parts of the model) contact one of the lead authors of the accompanying publication.

The model is formatted in the open SONATA standard and contains neuron locations and their properties (such as morphological types, cortical layer, etc.), their detailed morphologies, and synaptic connectivity associated with all systems described above. Modeled synapses are associated with their exact location in the dendritic tree, and additional anatomical parameters, such as spine length (where biologically plausible). Extrinsic synaptic connections from neurons in the remainder of non-barrel somatosensory cortex and thalamic inputs are also contained.

Analyzing the model

The model can be analyzed in terms of its anatomy, physiology and connectivity using the packages NeuroM, BlueBrain SNAP and ConnectomeUtilities. A small number of exemplary jupyter notebooks of instructive analyses are included, many more are included in the ConnectomeUtilities repository!

Simulating the model

This model can be simulated. We provide exemplary simulation configurations as part of this release.

The easiest way is to use our open-source simulator Neurodamus. The reference version is the branch nbS1-2023, which is archived under the following DOI: 10.5281/zenodo.8075202. Instructions on how to use the simulator are provided on the github pages linked above.

Briefly, you will first have to install Neurodamus. Next, build a "special" executable that include compiled versions of ion channel and  synapse models. To do that, first extract the contents of O1_mods.xz from this repository, then follow these instructions, where mod-files-from-released-circuit is replaced by the location of the extracted files on your system. Finally, run a simulation. The specific simulation conditions and stimuli are specified in simulation configuration files. A number of exemplary simulation configurations are included in this release (spontaneous_activity_simulations.zip). They configure a spectrum of different spontaneous states, that can serve as a basis of your in-silico campaigns. For details on the spontaneous activity spectrum, see our accompanying publication.

For details how to add stimuli or manipulations to a simulation, refer to the Sonata documentation, or contact one of the lead authors of the accompanying publication.

---

Creation of this model was supported by funding to the Blue Brain Project, a research center of the Ecole polytechnique federale de Lausanne (EPFL), from the Swiss government’s ETH Board of the Swiss Federal Institutes of Technology.