LARGE SCALE DISCRIMINATION BETWEEN NEURAL MODELS AND EXPERIMENTAL RECORDINGS

Professor Rick Gerkin1, Professor Sharon Crook2, Russell Jarvis3


1 Department of Mathematics and Statistics, Arizona State University, USA
2 Department of Mathematics and Statistics, Arizona State University, USA
3 Department of Life Sciences, Arizona State University, USA

nocite: | @[Brette, Romain and Gerstner, Wulfram], @[Pack, Christopher C. and Berezovskii, Vladimir K. and Born, Richard T.] ---

Abstract

NeuroML-DB [2] catalogues over 1,500 published models obtained in NeuroML format from Open Source Brain [5]. Complementing OSB, NeuroML-DB provides systematic characterizations of model complexity, electrophysiology, and morphology, making it easy to find, evaluate, and reuse models and their components

Example Features

We unified three different feature extraction protocols to create a very high dimensional feature space.

Modeling Methods

Publications Associated with Model Sources:

Large Scale Model Blue Brain Project (1035) Gouwens et al (2018) [6]
Publication Markram et al (2015) [8] Gouwens et al (2018) [6]

Extracted Feature Sources:

Feature1 Feature2 Feature3
Ephys Feature Extraction Library allensdk.ephys.ephys_features Druckmann et al. (2012) [3]

Virtual Experiment Three Step Protocol Stimulate for 2s

Injection 1 Injection 2 Injection 3
at 1.0×Rheobase at 1.5×Rheobase at 3.0×Rheobase

Results

Histogram to query Model/Data agreement

PCA Embedding

EigenValue Component Loadings

Next Steps

Optimize biophysically accurate models on the features that most disagree between models and experiments.

Interactive Plots:

Can be included or externally linked. dimension_reduced_TSNE

Building the Poster

This poster was created with rmarkdown utilizing a python kernel

The code for this poster lives here To compile this poster do:

BASH terminal Run:
R -e "install.packages('posterdown')"
R -e "rmarkdown::render('template_two.Rmd', \
                         output_file=\
                         'CNS_poster_2020.html')"

To use a python kernel in this poster:

r, include=FALSE,echo=FALSE}
library(reticulate)
use_python("/anaconda3/bin/python")

References:

bibliography: Russell_Prospectus_.bib