Modernizing the NEURON Simulator for Sustainability, Portability, and Performance

Abstract

The need for reproducible, credible, multiscale biological modeling has led to the development of standardized simulation platforms, such as the widely-used NEURON environment for computational neuroscience. Developing and maintaining NEURON over several decades has required attention to the competing needs of backwards compatibility, evolving computer architectures, the addition of new scales and physical processes, accessibility to new users, and efficiency and flexibility for specialists. In order to meet these challenges, we have now substantially modernized NEURON, providing continuous integration, an improved build system and release workflow, and better documentation. With the help of a new source-to-source compiler of the NMODL domain-specific language we have enhanced NEURON's ability to run efficiently, via the CoreNEURON simulation engine, on a variety of hardware platforms, including GPUs. Through the implementation of an optimized in-memory transfer mechanism this performance optimized backend is made easily accessible to users, providing training and model-development paths from laptop to workstation to supercomputer and cloud platform. Similarly, we have been able to accelerate NEURON's reaction-diffusion simulation performance through the use of just-in-time compilation. We show that these efforts have led to a growing developer base, a simpler and more robust software distribution, a wider range of supported computer architectures, a better integration of NEURON with other scientific workflows, and substantially improved performance for the simulation of biophysical and biochemical models.

Large-scale biophysically detailed model of somatosensory thalamocortical circuits in NetPyNE

“…We note that using CPU cycles/timestep would provide a more direct measure than the total simulation time, which may be affected by other factors such as background processes ( Girardi-Schappo et al, 2017 ). Nonetheless, the speedup obtained is consistent with the 2–7x speedups recently reported when using CoreNEURON on CPUs to simulate large-scale models ( Kumbhar et al, 2019 ; Awile et al, 2022 ). For example, the NetPyNE-based motor cortex model exhibited a speedup of 3.5x on Google Cloud.…”

Large-scale biophysically detailed model of somatosensory thalamocortical circuits in NetPyNE

“…The S1 model now joins other NetPyNE cortical simulations: generic cortical circuits (Romaro et al 2021), auditory and motor thalamocortical circuits (Sivagnanam et al 2020; Dura-Bernal, Neymotin, et al 2022; Dura-Bernal, Griffith, et al 2022), as well as simulations of thalamus (Moreira et al 2021), dorsal horn of spinal cord (Sekiguchi et al 2021), Parkinson’s disease (Ranieri et al 2021) and schizophrenia (Metzner et al 2020). These large cortical simulations can be extremely computer-intensive, which is a major motivation for NetPyNE’s facilities that allow one to readily simplify the network by swapping in integrate-and-fire or small-compartmental cell models, or by down-scaling to more manageable sizes.…”

“…These contrast with previous models of S1 (Huang, Zeldenrust, and Celikel 2022) or of generic sensory cortex (Potjans and Diesmann 2014) that employ simpler neuron models (leaky integrate and fire point neurons). Models with detailed conductance-based and morphologically-detailed neurons have been developed for other cortical regions, including V1 (Billeh et al 2020; Arkhipov et al 2018), M1 (Dura-Bernal, Neymotin, et al 2022), A1 (Dura-Bernal, Griffith, et al 2022), and CA1 (Bezaire et al 2016; Ecker et al 2020). Our model is also unique in incorporating thalamic neurons and thalamocortical bidirectional topological connectivity.…”

“…We also describe the methodology for future model parameter explorations, and provide a basic code set up example to explore the effects of inhibitory GABAergic connections on network dynamics. Examples of parameter explorations in NetPyNE-based biophysically detailed circuit models can be found in our related publications on motor and auditory cortex models (Sivagnanam et al 2020; Dura-Bernal, Neymotin, et al 2022; Dura-Bernal, Griffith, et al 2022), including an exploration of the effects of long-range and neuromodulatory inputs.…”

“…Conversion to NetPyNE also makes it easier to connect to previous models developed within the platform, such as our primary motor cortex model (Sivagnanam et al 2020; Dura-Bernal, Neymotin, et al 2022), and models implemented in other tools (e.g. NEST) by exporting to the NeuroML or SONATA standard formats.…”