visual_behavior_glm

Fits a kernel regression model to df/f traces during visual behavior.

Installing and setting up the package

Set up an environment

Before installing, it's recommended to set up a new Python environment with Python 3.7:

For example, using Conda:

conda create -n mFish_glm python=3.8

Then activate the environment:

conda activate mFish_glm

Installation

Currently, we are working with a forked version of AllenSDK that lives on Matt's git page:

git clone 'https://github.com/mattjdavis/AllenSDK.git't
cd AllenSDK
pip install -e .

An additional dependency of the package is brain_observatory_analysis (BOA)
Assuming that most users of this package will also be contributing to BOA, it should also be installed in 'editable' mode:

git clone https://github.com/AllenInstitute/brain_observatory_analysis.git
cd brain_observatory_analysis
pip install -e .

Alternatively, the current master branch could be installed with:

pip install git+https://github.com/AllenInstitute/brain_observatory_analysis.git

Test that the package was installed properly by importing the GLM class from outside of the visual_behavior_glm directory:

cd ~
python
>>> from visual_behavior_glm.glm import GLM

Please report issues at https://github.com/yavorska-iryna/visual_behavior_glm_mFish/issues

Use

Defining new kernels/regressors

• Adding a new kernel requires changes in two places.
• Add the kernel parameters to make_run_json() as a dictionary with keys length and offset
• Define the event times in add_kernel_by_label()

Fit the model

• Make the run json using delete_rebuild_run_json.py in scripts with python delete_rebuild_run_json.py --version <version> --label <a descriptive label> --src-path <path_to_source_code>
• Start the run for a single session with python scripts/fit_glm.py --oied <ophys_experiment_id> --version <version>
• Start the run for all sessions at the command line on hpc-login with python scripts/deploy_fits.py --glm-version <version> --env <name_of_conda_environment>
• Collect the results across sessions using retrieve_results(glm_version=<version>) from src/GLM_analysis_tools.py

Model Iteration System

• delete_rebuild_run_json.py saves a copy of current files, as well as a JSON file with run parameters to ../nc-ophys/visual_behavior/ophys_glm/v_<version>/