|Title||NetPyNE: A GUI-based tool to build, simulate and analyze large-scale, data-driven network models in parallel NEURON|
|Publication Type||Conference Paper|
|Year of Publication||2018|
|Authors||Dura-Bernal, S., Suter B. A., Quintana A., Cantarelli M., Gleeson P., Rodriguez F., Neymotin S. A., Hines M. L., Shepherd G. M., & Lytton WW.|
|Conference Name||Society for Neuroscience 2018 (SFN '18)|
|Keywords||SFN, Society for Neuroscience|
Transforming experimental data into solid conclusions and theory requires integrating and interpreting disparate datasets at multiple scales. The BRAIN Initiative 2025 report highlights this requires rigorous theory and modeling. The widely used NEURON simulator allows researchers to develop biophysically realistic models of neurons and networks. However, building and running parallel simulations of complex brain networks usually requires years of highly technical training. Here we present NetPyNE (www.netpyne.org), a tool that extends NEURON's capabilities and makes it accessible to the wider scientific community. NetPyNE provides both a programmatic and graphical interface that facilitates the definition, parallel simulation and analysis of data-driven multiscale models. Users can provide specifications at a high level via its standardized declarative language, e.g. a probability of connection, instead of millions of explicit cell-to-cell connections. With a single command, NetPyNE can then generate the NEURON network model and run an efficiently parallelized simulation. The user can then select from a variety of built-in functions to visualize and analyse the results, including connectivity matrices, voltage traces, raster plot, local field potential spectra or information transfer measures. The graphical user interface was developed using state-of-the-art technology (www.geppetto.org) and allows users to more intuitively access all NetPyNE functionalities: specifying model parameters using drop-down lists or autocomplete forms, interactively visualizing the 3D network, running parallel simulations or plotting results. NetPyNE models can be imported/exported to NeuroML specifications, facilitating model sharing and simulator interoperability. As a case-study we present the development of a multiscale model of mouse primary motor cortex (M1) microcircuits. The model simulates a cylindrical volume of cortical tissue with over 10,000 biophysically detailed neurons and 30 million synaptic connections. Neuron densities, classes, morphology and biophysics, and connectivity at the long-range, local and dendritic scale were derived from published experimental data. NetPyNE enabled the integration of these complex experimental datasets and in silico exploration of the microcircuit neural dynamics, information flow and underlying biophysical mechanisms.