Neuroscience Gateway - Software Dissemination, Large Scale
Modeling, Data Analysis on Supercomputers
Satellite
Workshop - 2023 Society for Neuroscience Annual Meeting
Date and Time: 11/11/2023, 08:30 AM - 12:30 PM
Location: George
Washington University, B1220, Science and Engineering Hall, 800 22nd St NW,
D.C.
https://www.seas.gwu.edu/science-and-engineering-hall
Please enter the Science and Engineering Hall via the entrance
from the 23rd Street
Registration (is free
but is required given limited space, please register): https://na.eventscloud.com/ereg/newreg.php?eventid=774279&
Organizers: Amit Majumdar, Subhashini Sivagnanam, Kenneth
Yoshimoto, San Diego Supercomputer Center, University of California San Diego
Ted Carnevale,
Neuroscience Department, Yale School of Medicine University
Workshop theme: The Neuroscience
Gateway (NSG), funded by both the National Institutes of Health (NIH) and the
National Science Foundation (NSF), serves the neuroscience community by
providing easy, open and free access to a large number of neuroscience software
and tools on supercomputing resources, academic cloud computing resources, and
associated storage resources, which are located at various national academic
supercomputer centers in the US. NSG can be used by neuroscientists from any
country. NSG eliminates administrative and technical barriers for all and
enables neuroscientists to carry out research that requires large scale
modeling and data processing. Another function of NSG is to disseminate neuroscience
modeling and data processing software on computing resources such that the
broader neuroscience community is easily able to access and use those software.
This workshop will bring together experts from the broader neuroscience
software developer and user community, irrespective of their direct involvement
with NSG or not. Researchers and experts from agencies such as the NIH, the
neuroscience software developer community and the neuroscience user community
will share their research and experiences within the context of large-scale
modeling and data processing in neuroscience.
AGENDA
8:30 AM - 8:45 AM
Welcome
8:45 AM - 9:15 AM
Title: Human Neocortical Neurosolver: A Software
Tool for Cell and Circuit Level Interpretation of MEG/EEG signals
Authors: Stephanie
Jones, Nicholas Tolley
Affiliation: Dept of
Neuroscience, Brown University
Abstract: MEG/EEG signals are correlated with nearly all healthy
and pathological brain functions. However, it is still extremely difficult to
infer the underlying cellular and circuit level origins. This limits the
translation of MEG/EEG signals into novel principles of information processing,
or into new treatment modalities for pathologies. To address this limitation,
we built the Human Neocortical Neurosolver (HNN): an open-source software tool
to help researchers and clinicians without formal computational modeling or
coding experience interpret the neural origin of their human MEG/EEG data. HNN
provides a graphical user interface (GUI) to an anatomically and biophysically
detailed model of a neocortical circuit, with layer specific thalamocortical
and cortical-cortical drives. Tutorials are provided to teach users how to
begin to study the cell and circuit level origin of sensory event related
potentials (ERPs) and low frequency rhythms. Once users have
an understanding of the basic workflows and tutorials in the HNN GUI,
those familiar with Python can work in the HNN-core Pythonic interface. We will
give a didactic overview of the background and development of HNN and describe
current and planned resources to use HNN through the Neuroscience Gateway
Portal.
9:15 AM - 10:00 AM
Title: NIH BRAIN Initiative (Part I): The
convergence of modeling and data science in the NIH BRAIN Initiative
Authors: Grace C.Y. Penga,
Susan N. Wrightb, Hermon Gebrehiwetc
Affiliations: NIH BRAIN Initiative (aNational Institute of Biomedical Imaging and
Bioengineering (NIBIB), bNational Institute on Drug
Abuse (NIDA), c National Institute of Neurological Disorders and Stroke (NINDS)
Abstract:
Since 2014, the NIH BRAIN Initiative has been promoting cutting-edge
technology that would accelerate the understanding of brain function as a
dynamic integrative system. These
technologies include quantitative and computing technologies that place
model-driven experiment design at its core.
All investigators performing research to uncover brain circuit
mechanisms are required to deliver a predictive model that is quantitative or
conceptual. This new community of
modelers are realizing the need to utilize rigorous data science methods and tools
to link neural processes across multiple scales of neural structure and
time. The NIH BRAIN initiative flagship
brain circuit program is striving to understand various behaviors from
perception to executive function. Each
of these projects incorporate modeling efforts covering scales from
cellular/molecular to inter-regional networks, and at many scales of modeling
approaches from biophysics to analytical and numerical, with many programs
spanning multiple scales of study. As
well the data science efforts required for these projects present many
challenges and opportunities for harmonizing the data management
practices.
Title: NIH BRAIN Initiative (Part II): Data
science across the NIH BRAIN initiative brain circuits projects
Authors: Manuel
Schottdorf1, Guoqiang Yu2,
Edgar Y. Walker3
Affiliations: NIH U19
Data Science Consortium (1Princeton, 2Virginia Tech, 3University of Washington)
Abstract:
The rise of large scientific collaborations in neuroscience requires
systematic, scalable and reliable data management. How this is best done in practice
remains an open question. To address this, we conducted a data science survey
among currently active U19 grants, funded through the NIH's
BRAIN Initiative.
The survey was answered by both data science liaisons and Principal
Investigators, speaking for ~500 researchers across 21 nation-wide
collaborations. We describe the tools, technologies and methods currently in
use, and identify several shortcomings of current data science practice.
Building on this survey, we develop plans and propose policies to improve data
collection, use, publication and re-use in the neuroscience community.
10:00
AM - 10:30 AM
Title:
Modeling the development of the visual system: computational challenges vs
high-performance computing
Authors: Ruben A Tikidji-Hamburyan and Matthew T. Colonnese
Affiliation:
Department of Pharmacology and Physiology, The George Washington
University, Washington D.C., United States
Abstract: During the
early period of development, neurons and networks are gradually maturing toward
what will be a fully functional visual system in adulthood. At early ages,
neurons are slow, with prolonged dynamics in scales up to seconds. Networks are
connected imprecise and need refinement. To refine projections, the thalamus
and cortex extract positional information encoded in the correlated activity of
retinal ganglion cells in the course of the first two weeks of postnatal
development. Modeling this developmental process poses substantial
computational challenges. Here, we will discuss the usage of high-performance
computing and Neuroscience Gateway specifically to meet these challenges,
from fitting a single neuron model to computing the information lost in
different biological conditions. We will review the usage of various
techniques for model construction, simulation, and assessment of model quality
and biological relevance. We will show how to use principal component analysis
of single neuron parameters to assess fitting quality, distributed computation
for inferring the information timescales, and usage of biology of studied
networks for load distribution on parallel HPC cluster or multicore CPU.
10:30 AM - 11:00 AM BREAK
11:00 AM - 11:30 AM
Title: NIC:
An integrated neuroinformatics tool for studying
brain interaction dynamics in neurological disorders
Authors: Katrina Prantzalos1, Dipak Upadhaya1, Shafiabadi N1, Gurski N1, Fernandez-BacaVaca G1, Yoshimoto K2, Sivagnanam
S2, Majumdar A2, Sahoo SS1
Affiliations: 1Case Western Reserve University, University
Hospitals Cleveland Medical Center; 2San Diego Supercomputing Center/University
of California San Diego
Abstract: High fidelity brain recording from both
scalp and intracranial electrodes are widely used to study brain interaction
dynamics, including in neurological disorders such as epilepsy. However,
large-scale analysis of these neurophysiological signal data, such as
electroencephalogram (EEG), requires multiple data pre-processing steps and
subsequent analysis the use of specialized methods such as graph network
analysis or algebraic topology algorithms. In this talk, we describe the
Neuro-Integrative Connectivity (NIC) tool that features integrated support for
signal data processing, analysis, and use of machine learning models to
characterize brain interactions in epilepsy. The NIC tool features a modular
architecture that supports extensibility, integration of new functionalities
such as machine learning interpretability methods, and ease of maintenance. The
NIC tool is available through the NSG portal.
11:30 AM - 12:00 PM
Title: Whole brain
connectome models mapping brain structure to function to study lifespan ageing and neurological diseases
Authors: Arpan Banerjee1, Suman Saha1,
Samuel Berkins1, Anagh Pathak1,
Neeraj Kumar1, Priyanka Chakraborty1, Varun Madan Mohan1,
Anirudh Vattikonda1, Amit Naskar1, Dipanjan
Roy1,2
Affiliations:
1. National Brain Research Centre, NH 8 Manesar, Gurugram
122052, India
2. Indian Institute of Technology Jodhpur, NH 62, Jodhpur, India
Abstract: A significant
development in recent times has been the establishment of whole-brain
computational models (WBMs) to understand brain dynamics at multiple
spatiotemporal scales. In addition to understanding whole brain dynamics across
scales based on biologically realistic anatomical connectivity, it has been
pivotal in providing fundamental insights into structure-function relations in
brain mapping, an often-cherished goal of neuroscientific explorations. I will
present a pipeline developed by our group that can take individual subject-
specific tractography data as input to generate the functional dynamics
observed in electro-encephalogram (EEG)/ magneto-electrogram (MEG) and fMRI
recordings. Further, I will demonstrate how key factors in human lifespan aging
such as synaptic scaling and conduction speeds can explain the underlying neurocompensatory mechanisms that preserves functional
integration during healthy aging using a WBM, constrained on subject-specific
connectome. Using biophysically realistic extensions that capture the
multi-scale neurotransmitter-neuroelectric interactions we demonstrate how
GABA/Glutamate relationships vary over lifespan aging. Subsequently, we show
how such implementations of WBM can extract the conduction speed as the most
relevant patient specific clinical marker of cognitive impairment in multiple
sclerosis. Using controlled experimental paradigms such as auditory state
rhythms and language discrimination tasks we show how brain hemispheric
lateralization of speech/ tonal input processing can also be explained by WBMs.
Together, we outline the various ways WBM a.k.a
virtual brain, sculpted from idiosyncrasies of subject-specific connectome,
contribute to basic neuroscientific explorations involving resting state and
task-based experimental paradigms and also generate predictive markers for
clinical practice.
12:00 PM - 12:30 PM
Title: The role of computational models in experimental research:
predictive vs. explanatory power
Authors: Sherif M. Elbasiouny1,2 and
Mohamed H. Mousa1
Affiliations:1Department of Neuroscience, Cell Biology, and Physiology, Boonshoft School of Medicine and College of Science and
Mathematics
Wright State University, Dayton, OH, United States
2Department of Biomedical, Industrial, and Human Factors Engineering,
College of Engineering and Computer Science
Wright State University, Dayton, OH, United States
Abstract: Computational models provide a means to study biological
systems at multi levels. When combined with experiments, computational models
could be used to explain the experimental data based on an underlying
hypothesis or use the data to generate new testable hypotheses. In this talk,
we show examples from our work on the computational models' predictive vs.
explanatory power, illustrating the strength and limitations of each approach
and discussing some challenges involved in this simulation/experiment
co-development approach.