ephys_analysis

This repository holds code to conduct ephys analysis on data files that come from trodes and phy files. It also has code to extract behavior data from Boris and ECU data and some tools to manipulate behavioral epochs.

Prerequisites

• conda or miniconda installed

Setup

1. Clone this repository:

   git clone https://github.com/padillacoreanolab/ephys_analysis.git
   

   or use github desktop to clone it locally.

   Open the repository in a terminal that can access conda

   cd users/yourname/github/ephys_analysis
   

2. Create and activate conda environment

conda env create -f ephys_env.yml
conda activate ephys_env

3. Install the packages

bash 
pip install -e . 

4. Clone your own project repository and import functions from ephys_analysis at top like this:

from lfp.lfp_collection import LFPCollection

What do you need

• LFP: Current version of this code assumes you have recorded your data through Trodes and have kept the *.rec/*merged.rec data folder structure.
• Spike: current version of this code assumes you have done manual spike sorting through Phy and have saved everything under a *merged.rec/phy/.. data structure.
• Behavior: both Spike and LFP analysis require the same format of behavior input for analysis, currently helper functions exist under Behavior for BORIS and ECU derived behavior data.

Setting up your Trodes folder using helper_fxns folder

Extracting relevant files: LFP + ECU + Analog

trodes_export.py has functions that will allow you batch export various files through trodes. You will need trodes installed onto your local device.

• time: export needed to find the first timestamp of a recording for all LFP, ECU and analog analysis
• DIO: if you used the ECU to record operant chamber inputs, you will need to extract DIO files
• analog: if you used any Ain inputs to record, you will need to extract analog files

Converting h264 videos into sleap friendly mp4s

Under helper functions is also a suite of conversion functions to convert h264 files (or any video format) into a sleap friendly mp4. This also helps if you want to use BORIS or AnyMaze or any other software to do behavior classification.

Contents of this repository

It is highly recommended that users understand what a class is in python to most effectively use the following scripts. The lfp and spike folders each have the own README for more detailed overviews of the contents of those folders.

Spike

This folder contains scripts to analyze the output of phy manually curated spikes. It contains scripts to do single cell analyses (Wilcoxon tests, raster plots, and bootstrap firing rate comparisons), PCA analyses and trajectory visualizations, classifiers to predict behavioral events based on population activity, and functions to normalize firing rates.

LFP

This folder contains scripts to analyze the output of a trodes recording. It contains scripts to calculate power, coherence and Granger causality using the package Multitaper Spectral Connectivity. It is highly recommended to use the Hipergator (or other high computing cluster) to do so. GPU greatly increases speed of computation for Granger causality specifically and allows you to more readily load all your recordings into working memory to perform data analysis and plotting.

Below are a few instructions on how to use rclone to get your large datasets from dropbox to the remote computing cluster:

dropbox + rclone

1. install rclone on your local computer: https://rclone.org/downloads/
2. run on personal computer

rclone.exe authorize "dropbox"

crtl + click on link that terminal spits out and press agree, terminal will now spit out a token key, finish step 3 before copying and pasting it into the hipergator terminal

3. open terminal in hipergator and run the following lines:

module load rclone
rclone config

• n) New remote
• name your dropbox
• Storage > 13 (type 13 (which is dropobx) and hit enter)
• leave client_id and client_secret blank (aka press enter)
• type n and click enter (No) for edit advanced config
• type n and click enter (No) for web broswer question
• copy and paste config token from local terminal into hipergator terminal
• click y and hit enter for default settings to finish

if authentication breaks press choose:

• e) Edit existing remote
• 4. find data on dropbox

rclone ls pc-dropbox:"Padilla-Coreano Lab/path/to/data"

5. copy data from dropdox source = pc-dropbox path ="path/to/folder" run dry run first to confirm path and size of download

example of destination path dest:path = ./data

rclone copy source:path dest:path --progress --dry-run

then run it for reals

rclone copy source:path dest:path --progress

example real command:

rclone copy pc-dropbox:"Padilla-Coreano Lab/2024/Cum_SocialMemEphys_pilot2/Only_Subjects (phase 5)/sorted" ./only_subjects_spike_data --progress  --dry-run

or to upload data to dropbox:

rclone copy ./r4_predictions pc-dropbox:"Padilla-Coreano Lab/2024/Cum_SocialMemEphys_pilot2/SLEAP/only_subject_mp4s/r4_predictions" --progress --dry-run

Developer installation

How to run tests

Download test data

We're going to use this small recording from trodes: https://docs.spikegadgets.com/en/latest/basic/BasicUsage.html

There is a helper script that downloads it and unzips it into tests/test_data/.

python -m tests.utils download_test_data

please open terminal and run in this directory

python -m unittest discover -s *folder_you_want_to_test* -p '*test*.py

How to run single files as modules

For example, to turn a dataset into a test:

python -m tests.utils create /Volumes/SheHulk/cups/data/11_cups_p4.rec/11_cups_p4_merged.rec

This is nice because it allows all imports to be relative to the root directory.