tutorial

** https://github.com/MouseLand/rastermap/tree/master/tutorial **

Using rastermap to explore visual cortical activity

Setting up

First start downloading the data from here. It includes all the visual cortical responses, the behavioral responses aligned to the neural frames, and a short example video of the mouse's face. In this experiment we are showing sparse noise stimuli to the mouse (for retinotopic mapping) as it freely runs on an air-floating ball.

Next we will make an environment with all the packages that we need with the conda package manager using the environment.yml file:

1. Download the environment.yml from this folder OR
   1. Clone this repository git clone https://github.com/MouseLand/rastermap.git (or pull the latest version if you already have it with git pull)
   2. cd rastermap/tutorial to be in the same folder with the environment.yml folder.
2. Open an anaconda prompt (windows) / command prompt (linux/Mac) with conda for python 3 in the path. In linux/Mac you can check which conda you have with which conda, it should be in a subfolder below anaconda3.
3. Run conda env create -f environment.yml.
4. To activate this new environment, run conda activate mouseland.
5. You should see (mouseland) on the left side of the terminal line. Now check that you can python -m suite2p or python -m facemap or python -m rastermap.

View the data in suite2p

^ 18,795 neurons in V1 ^

You will need 16GB of RAM to load this data into your computer. Start suite2p with

python -m suite2p

From the file menu you can load the data (or with CTRL+L), choose the stat.npy file.

You can view the correlations among neurons with the correlations color. If you want to bin the responses a certain before computing the correlations, you can use the bin= text edit box. The data is collected with an imaging rate of 3Hz, so 3 bins = 1 second. Some neurons are correlated with their neighbors, others aren't. Why might that be?

Retinotopy

We will now compute the receptive fields of single neurons, using the tutorial.ipynb notebook.

Dimensionality reduction

What are the overall patterns of activity in visual cortex? Are they well-defined by the principal components? We will look at the PCs in the notebook and their receptive fields.

Maybe there is a better way to visualize this activity. We can run rastermap from inside suite2p in the Visualizations menu (or with CTRL+V). You can then look at different groups of neurons in the main GUI by circling them with the RED ROI box in the top plot, and then clicking "show selected cells in GUI".

We can also run rastermap inside the notebook and look at the receptive fields of groups of neurons placed together in the embedding. These receptive field estimates will be less noisy. But what are these neurons doing which don't have clear receptive fields?

Behavioral analysis with facemap

Run

python -m facemap

Then open the video "cam1_TX39_20Hz.avi" in facemap (this is a subset of the video). You can process this small subset of data in the GUI.

I've run facemap on the whole movie and aligned them to the neural frames for you. So now let's see how the behavior relates to the neural activity in the tutorial.ipynb notebook.

BONUS: Explore data in rastermap GUI

These neural responses are high-dimensional, is a one-dimensional embedding enough to view the structure? We can also embed the neurons in a 2D space using rastermap. Run

python -m rastermap

Load in the data (spks.npy) and then run 2D rastermap.