Learning the dynamics of realistic models of C. elegans nervous system with recurrent neural networks DOI Creative Commons
Ruxandra Bărbulescu,

Gonçalo Mestre,

Arlindo L. Oliveira

и другие.

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Янв. 10, 2023

Abstract Given the inherent complexity of human nervous system, insight into dynamics brain activity can be gained from studying smaller and simpler organisms. While some potential target organisms are simple enough that their behavioural structural biology might well-known understood, others still lead to computationally intractable models require extensive resources simulate. Since such frequently only acting as proxies further our understanding underlying phenomena or functionality, often one is not interested in detailed evolution every single neuron system. Instead, it sufficient observe subset neurons capture effect profound nonlinearities neuronal system have response different stimuli. In this paper, we consider nematode Caenorhabditis elegans seek investigate possibility generating lower system’s with low error using measured simulated input-output information. Such termed black-box models. We show how C. modelled data-driven neural network architectures. Specifically, use state-of-the-art recurrent architectures Long Short-Term Memory Gated Recurrent Units compare these terms properties accuracy (Root Mean Square Error), well resulting Unit a hidden layer size 4 able accurately reproduce very furthermore explore relative importance inputs scalability more scenarios.

Язык: Английский

BioModels—15 years of sharing computational models in life science DOI Creative Commons
Rahuman S. Malik‐Sheriff, Mihai Glont, Tung V N Nguyen

и другие.

Nucleic Acids Research, Год журнала: 2019, Номер unknown

Опубликована: Ноя. 6, 2019

Computational modelling has become increasingly common in life science research. To provide a platform to support universal sharing, easy accessibility and model reproducibility, BioModels (https://www.ebi.ac.uk/biomodels/), repository for mathematical models, was established 2005. The current allows submission of models encoded diverse formats, including SBML, CellML, PharmML, COMBINE archive, MATLAB, Mathematica, R, Python or C++. submitted are curated verify the computational representation biological process reproducibility simulation results reference publication. curation also involves encoding standard formats annotation with controlled vocabularies following MIRIAM (minimal information required biochemical models) guidelines. now accepts large-scale auto-generated models. With gradual growth content over 15 years, currently hosts about 2000 from published literature. 800 world's largest emerged as third most used data resource after PubMed Google Scholar among scientists who use their Thus, benefits modellers by providing access reliable semantically enriched that share, reproduce reuse.

Язык: Английский

Процитировано

317
Inhibitory stabilization and cortical computation DOI
Sadra Sadeh, Claudia Clopath

Nature reviews. Neuroscience, Год журнала: 2020, Номер 22(1), С. 21 - 37

Опубликована: Ноя. 11, 2020

Язык: Английский

Процитировано

162
NetPyNE, a tool for data-driven multiscale modeling of brain circuits DOI Creative Commons
Salvador Durá-Bernal, Benjamin A. Suter, Padraig Gleeson

и другие.

eLife, Год журнала: 2019, Номер 8

Опубликована: Апрель 26, 2019

Biophysical modeling of neuronal networks helps to integrate and interpret rapidly growing disparate experimental datasets at multiple scales. The NetPyNE tool (www.netpyne.org) provides both programmatic graphical interfaces develop data-driven multiscale network models in NEURON. clearly separates model parameters from implementation code. Users provide specifications a high level via standardized declarative language, for example connectivity rules, create millions cell-to-cell connections. then enables users generate the NEURON network, run efficiently parallelized simulations, optimize explore through automated batch runs, use built-in functions visualization analysis - matrices, voltage traces, spike raster plots, local field potentials, information theoretic measures. also facilitates sharing by exporting importing formats (NeuroML SONATA). is already being used teach computational neuroscience students modelers investigate brain regions phenomena.

Язык: Английский

Процитировано

160
The Neurodata Without Borders ecosystem for neurophysiological data science DOI Creative Commons
Oliver Rübel, Andrew Tritt, Ryan Ly

и другие.

eLife, Год журнала: 2022, Номер 11

Опубликована: Окт. 4, 2022

The neurophysiology of cells and tissues are monitored electrophysiologically optically in diverse experiments species, ranging from flies to humans. Understanding the brain requires integration data across this diversity, thus these must be findable, accessible, interoperable, reusable (FAIR). This a standard language for metadata that can coevolve with neuroscience. We describe design implementation principles data. Our open-source software (Neurodata Without Borders, NWB) defines modularizes interdependent, yet separable, components language. demonstrate NWB's impact through unified description modalities species. NWB exists an ecosystem, which includes management, analysis, visualization, archive tools. Thus, enables reproduction, interchange, reuse More broadly, generally applicable enhance discovery biology FAIRness.

Язык: Английский

Процитировано

97
Neuromorphic computing at scale DOI
Dhireesha Kudithipudi, Catherine D. Schuman, Craig M. Vineyard

и другие.

Nature, Год журнала: 2025, Номер 637(8047), С. 801 - 812

Опубликована: Янв. 22, 2025

Язык: Английский

Процитировано

3
Modernizing the NEURON Simulator for Sustainability, Portability, and Performance DOI Creative Commons
Omar Awile, Pramod Kumbhar,

Nicolas Cornu

и другие.

Frontiers in Neuroinformatics, Год журнала: 2022, Номер 16

Опубликована: Июнь 27, 2022

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

Язык: Английский

Процитировано

42
Data-driven multiscale model of macaque auditory thalamocortical circuits reproduces in vivo dynamics DOI Creative Commons
Salvador Durá-Bernal, Erica Y. Griffith, Annamaria Barczak

и другие.

Cell Reports, Год журнала: 2023, Номер 42(11), С. 113378 - 113378

Опубликована: Ноя. 1, 2023

We developed a detailed model of macaque auditory thalamocortical circuits, including primary cortex (A1), medial geniculate body (MGB), and thalamic reticular nucleus, utilizing the NEURON simulator NetPyNE tool. The A1 simulates cortical column with over 12,000 neurons 25 million synapses, incorporating data on cell-type-specific neuron densities, morphology, connectivity across six layers. It is reciprocally connected to MGB thalamus, which includes interneurons core matrix-layer-specific projections A1. multiscale measures, physiological firing rates, local field potentials (LFPs), current source densities (CSDs), electroencephalography (EEG) signals. Laminar CSD patterns, during spontaneous activity in response broadband noise stimulus trains, mirror experimental findings. Physiological oscillations emerge spontaneously frequency bands comparable those recorded vivo. elucidate population-specific contributions observed oscillation events relate them presynaptic input patterns. offers quantitative theoretical framework integrate interpret predict its underlying cellular circuit mechanisms.

Язык: Английский

Процитировано

28
Multiscale model of primary motor cortex circuits predicts in vivo cell-type-specific, behavioral state-dependent dynamics DOI Creative Commons
Salvador Durá-Bernal, Samuel A. Neymotin, Benjamin A. Suter

и другие.

Cell Reports, Год журнала: 2023, Номер 42(6), С. 112574 - 112574

Опубликована: Июнь 1, 2023

Understanding cortical function requires studying multiple scales: molecular, cellular, circuit, and behavioral. We develop a multiscale, biophysically detailed model of mouse primary motor cortex (M1) with over 10,000 neurons 30 million synapses. Neuron types, densities, spatial distributions, morphologies, biophysics, connectivity, dendritic synapse locations are constrained by experimental data. The includes long-range inputs from seven thalamic regions noradrenergic inputs. Connectivity depends on cell class depth at sublaminar resolution. accurately predicts in vivo layer- cell-type-specific responses (firing rates LFP) associated behavioral states (quiet wakefulness movement) manipulations (noradrenaline receptor blockade thalamus inactivation). generate mechanistic hypotheses underlying the observed activity analyzed low-dimensional population latent dynamics. This quantitative theoretical framework can be used to integrate interpret M1 data sheds light multiscale dynamics several conditions behaviors.

Язык: Английский

Процитировано

26
Methods and considerations for estimating parameters in biophysically detailed neural models with simulation based inference DOI Creative Commons
Nicholas Tolley, Pedro Luiz Coelho Rodrigues, Alexandre Gramfort

и другие.

PLoS Computational Biology, Год журнала: 2024, Номер 20(2), С. e1011108 - e1011108

Опубликована: Фев. 26, 2024

Biophysically detailed neural models are a powerful technique to study dynamics in health and disease with growing number of established openly available models. A major challenge the use such is that parameter inference an inherently difficult unsolved problem. Identifying unique distributions can account for observed dynamics, differences across experimental conditions, essential their meaningful use. Recently, simulation based (SBI) has been proposed as approach perform Bayesian estimate parameters SBI overcomes not having access likelihood function, which severely limited methods models, by leveraging advances deep learning density estimation. While substantial methodological advancements offered promising, large scale biophysically challenging doing so have established, particularly when inferring time series waveforms. We provide guidelines considerations on how be applied waveforms starting simplified example extending specific applications common MEG/EEG using modeling framework Human Neocortical Neurosolver. Specifically, we describe compare results from oscillatory event related potential simulations. also diagnostics used assess quality uniqueness posterior estimates. The described principled foundation guide future wide variety dynamics.

Язык: Английский

Процитировано

8
The SONATA data format for efficient description of large-scale network models DOI Creative Commons
Kael Dai, Juan Hernando, Yazan N. Billeh

и другие.

PLoS Computational Biology, Год журнала: 2020, Номер 16(2), С. e1007696 - e1007696

Опубликована: Фев. 24, 2020

Increasing availability of comprehensive experimental datasets and high-performance computing resources are driving rapid growth in scale, complexity, biological realism computational models neuroscience. To support construction simulation, as well sharing such large-scale models, a broadly applicable, flexible, data format is necessary. address this need, we have developed the Scalable Open Network Architecture TemplAte (SONATA) format. It designed for memory efficiency works across multiple platforms. The represents neuronal circuits simulation inputs outputs via standardized files provides much flexibility adding new conventions or extensions. SONATA used modeling visualization tools, also provide reference Application Programming Interfaces model examples to catalyze further adoption. free open community use build upon with goal enabling efficient building, sharing, reproducibility.

Язык: Английский

Процитировано

64