NetPyNE: a high-level interface to NEURON to facilitate the development, parallel simulation and analysis of data-driven multiscale network models

TitleNetPyNE: a high-level interface to NEURON to facilitate the development, parallel simulation and analysis of data-driven multiscale network models
Publication TypeConference Paper
Year of Publication2018
AuthorsDura-Bernal, S., Gleeson P., Neymotin S., Suter B. A., Quintana A., Cantarelli M., Hines M., Shepherd G., & Lytton W. W.
Conference NameComputational Neuroscience Meeting (CNS 18')
Keywords2018, BMC, BMC Neuroscience 2018, CNS
Abstract

Experimental data is accumulating at an unprecedented–-and accelerating–-rate. However, as the BRAIN Initiative 2025 report points out: (1) even excellent quality data will not yield solid conclusions unless it is adequately integrated and interpreted, and (2) turning experimental knowledge into understanding inevitably requires rigorous theory and modeling. Biophysically realistic modeling provides a tool to integrate, organize and bridge data at multiple scales and develop hypothesis about the biological mechanisms underlying physiological and pathological brain function. NetPyNE (www.netpyne.org) is a high-level Python interface to the widely used NEURON simulator [1]. It provides high-level declarative language designed to facilitate the definition of data-driven multiscale models, e.g., a concise set of connectivity rules vs. millions of explicit cell-to-cell connections. The user can then easily generate NEURON network instances from these specifications, run efficient parallel simulations (with predefined setup for supercomputers), and exploit the wide array of built-in analysis functions (e.g. connectivity matrix, voltage traces, raster plot, information transfer measures). A recent feature provides the ability to place extracellular LFP recording electrodes at arbitrary 3D locations and plot the LFP signal, power spectra or spectrogram. All this functionality is also accessible via a graphical user interface (GUI) based on the state-of-the-art Geppetto technology. The GUI provides an intuitive way to define the model, including an interactive Python console and full synchronization with the underlying Python-based model (Fig. 1). The user can visualize the 3D network, run simulations and choose from the available analysis plots. NetPyNE's standardized format clearly separates model parameters from implementation and can be exported/imported to NeuroML, thus making it easier to understand, reproduce, reuse and share models. This has motivated the conversion of several published models to NetPyNE specifications, including the Potjans&Diesmann cortical network, the Traub thalamocortical network, the Cutsuridis CA1 microcircuit and the Tejada dentate gyrus network. The tool is also being used to develop a variety of new models exploring mouse M1 microcircuits [2], the claustrum network, cerebellum circuits, transcranial magnetic stimulation (TMS) in cortex, or the underlying biophysics of EEG recordings. We expect the NetPyNE tool to make data-driven biophysically-detailed network modeling accessible to a wider range of researchers and students, including those with limited programming experience, and encourage further collaboration between experimentalists and modelers.Fig. 1The NetPyNE GUI provides an intuitive way to generate, simulate and analyze data-driven biophysically detailed network models. NetPyNE GUI showing 3D representation of (simplified) M1 model and simulation results (raster plot, statistics, voltage traces and power spectra)