This repository contains code and visualizations needed to reproduce the results presented in "Decoding locomotion from population neural activity in moving C. elegans". All scripts are written for Python 2.7.
The associated data can be found here.
To reproduce the figures from the paper, first download the data from the OSF repository. Each dataset is indentified by a strain name and condition, e.g. AML32_moving, and is wrapped in a .tar.gz archive. After extracting an archive, you will find subfolders corresponding to individual recordings. Additionally, a text file, e.g. AML32_moving_datasets.txt, indicates which datasets should be used for analysis. For recordings of the AML310 strain, these files also contain the volume number at which the BFP laser was turned on to identify AVA. The part of the recording after this volume number is excluded for behavior analysis.
Within each dataset folder, there are five files:
- centerline.mat
- heatData.mat
- heatDataMS.mat
- pointStatsNew.mat
- positionDataMS.mat These are the output of the neuron registration pipeline described here.
Next, edit utility/user_tracker.py to add your hostname to the dataPaths and codePaths dictionaries. To find your hostname, run hostname in a terminal.
The figure code is contained in figures/fig1, figures/fig2, and so on. Many of these scripts rely on precomputed data files, so you should start by running the following scripts:
> python utility/get_all_recordings.py
> python prediction/train_linear_models.pyAfter these scripts run, you should be able to run any of the scripts in figures (except for figures/fig3/model_comparison.py, for which you should first run prediction/train_all_models.py).