Brain Visualizations

I wrote some Python code using the library 'brainrender' according to their tutorial https://github.com/BrancoLab/BrainRender/blob/master/Examples/notebooks/Regions.ipynb The library is built to interoperate with Allen atlases. The visualizations can be found here https://github.com/mariakesa/NeuroMatchAcademy2020Project/blob/master/Visualizations.ipynb You can visualize the spatial location of brain regions in 3D according to their alias using this library.

I am really enjoying my time at the NeuroMatchAcademy https://github.com/NeuromatchAcademy NMA is a novel endeavor designed to bring VIP quality computational neuroscience to people over the world. It is entirely online!

Group projects on cool open datasets are designed to help apply the skills learned in NMA tutorials and lectures.

Here's a project proposal that we created with Jean Simmonnet and Xin Wei:-) We're cross-referencing openly available large-scale neuropixel probe recordings from the mouse brain during a visual discrimination behavior from Nick Steinmetz https://www.youtube.com/watch?v=WXn4-FpVaOo and the Allen Mouse Brain Connectivity atlas www.brain-map.org . Here's what we came up with.

PROPOSAL

 What factors control functional correlations between brain areas?

 In the Steinmetz paper, the activity patterns of identified brain areas have been correlated to distinct phases of a visual discriminatory task (choice, motor, reward). On the other hand, the brain’s global anatomical connectivity is well charted in endeavors such as Allen Mouse Brain Atlas. Our goal in this proposal is to look at the functional connectivity between anatomically connected brain regions. We will use the Steinmetz dataset, and suggest to primarily focus on a subset of the data, looking at the coupling between prefrontal areas and midbrain areas, because the original Steinmetz paper identified these regions as encoding the animal’s behavioral action.

 As a first step, we will explore the data using dimensionality reduction e.g. principal component analysis (PCA), non-negative matrix factorisation (NMF) and canonical correlation analysis (CCA) on population spike rates and behavior to find trends and clusters. To determine if brain areas are functionally coupled, we suggest to analyse the lag of neuronal activities and Granger causality between brain areas. Those analyses may indicate the direction of the information flow. We then suggest performing  supervised regression analyses (linear regression, Generalised Linear Model (GLM)) on activity patterns from different brain regions, to determine if the activity in a given area can be derived from another one. Additionally, we want to determine if behavior and its performance influence functional coupling between brain regions.

 Our predictions are that anatomically connected areas should be functionally coupled, that the direction of the coupling depends on the anatomical connectivity, and that this functional coupling is altered by behavior and performance.

This is so much fun!