Behavioral Analysis for the homecage data

Using the DeepLabCut pose data, running behavioral analysis with MotionMapper to extract behavior

1. https://github.com/DeepLabCut/DeepLabCut
2. https://github.com/bermanlabemory/motionmapperpy

Installing dependencies

Make sure you have a virtual environment setup before installing all dependencies. You can do this with either Anaconda(conda) or traditional Python(pip).

Pipeline

The overall analysis is broken up into 3 steps:

1. Query trial data from DataJoint
2. Run Behavioral Analysis (PostProcessing DLC and MotionMapper inference)
3. Store analyzed data into its respective trial .mat files

Configuration

All the parameters used throughout the process are contained in the config.yaml file. Use template_config.yaml as an example of what parameters to modify and understand what each does. Make a copy of the template_config.yaml file and modify this file instead of the template.

Running Analysis

Step 1: Query data from DataJoint

The initial step of the process is gathering all data from DataJoint and saving its contents into a CSV file. Note, this step can only be done on local Windows machines in the lab since we aren't able to open SSH tunnels on any SCC compute nodes.

Important parameters to set in your config file:

1. processing_folder
2. datajoint_credentials
3. only_run_datajoint
4. animal_list.

How to run:

1. Go to <repo_directory>
2. python behavior_inference.py --config_file_name <name of your config file>.yaml

The script will go through each animal in your <animal_list>, create an animal folder for the specific animal in the <processing_folder> and create a CSV file called FOUND_TRIALS.csv with a list of file paths to all trial .mat files that belong to that animal.

Step 2: PostProcessing DLC and MotionMapper Inference

The next step is running the behavioral analysis given the list of trials collected by the previous step. For improved speedup, we will use the SCC.

Important parameters to set in your config file:

1. processing_folder
2. post_processing_dlc_paths
3. post_processing_dlc_params
4. motion_mapper_version
5. motion_mapper_file_paths
6. motion_mapper_inference_params
7. animal_list

How to run on the SCC

1. Log in to SCC
2. Go to <repo_directory>/scc
3. qsub run_behavior_job_array.sh <name of your config file>.yaml

How to run on Windows

1. Go to <repo_directory>
2. python behavior_inference.py --config_file_name <name of your config file>.yaml

This script will run the behavioral analysis for each animal in <animal_list> and save intermediate files (WAVELETS.mat, HEATMAP.png, etc...) in their respective animal folder throughout the analysis.

Step 3: Storing data to trial mat files

The final step is to use all the intermediate files generated to save the contents into each trial's respective .mat file.

Important parameters to set in your config file:

1. processing_folder
2. post_processing_dlc_paths
3. post_processing_dlc_params
4. motion_mapper_version
5. motion_mapper_file_paths
6. motion_mapper_inference_params
7. animal_list

How to run on the SCC

1. Log in to SCC
2. Go to <repo_directory>/scc
3. qsub store_behavior_job_array.sh <name of your config file>.yaml

How to run on Windows

1. Go to <repo_directory>
2. python store_behavior_data.py --config_file_name <name of your config file>.yaml