cheerios

Neural population decoding pipeline for the DMS temporal waiting task. Uses Ridge regression to decode elapsed time from MSN (medium spiny neuron) population firing rates, and characterizes how the DMS encodes a behavioral clock — including clock-speed extraction, reward history effects, and confound controls.

Setup

git clone https://github.com/rebz444/cheerios.git
cd cheerios
conda env create -f environment.yml
conda activate cheerios

Data

Data is not in the repo (gitignored). Copy your data directory to the expected location before running:

~/data/neural_data/
    logs/               # CSVs including RZ_msn_waveform.csv
    session_pickles/    # Per-session spike/trial pickle files
    figures/            # Output figures written here

If your data lives somewhere else, edit paths.py to update DATA_DIR.

Pipeline

Scripts are numbered in run order:

Step	Script	Description
0a	0a_datajoint_processing_check.ipynb	DataJoint validation
0b	0b_neural_data_processing.ipynb	Neural data ingestion
0c	0c_neural_data_examining.ipynb	Data examination
0d	0d_neural_data_quality_metrics.ipynb/py	Spike sorting QC
0e–0j	0e_ … 0j_	Neuron location matching, waveform diagnostics, cell-type relabeling, region labeling, track deviation
1	1a_plot_raster_histo_by_quantile.ipynb	Raster + PSTH plots
2	2_neuron_clustering.ipynb	Neuron clustering & firing rate analysis
3	3a_bg_predict_time_waited.ipynb	Behavioral prediction from background activity
4	4a/4b/4c_encoding_GLM*.ipynb	GLM encoding models
5	5a_timing_analysis.ipynb	Timing trajectory analysis
6	6_trial_firing_rate.ipynb	Trial-level firing rates
—	population_decoder_v2.py	Full population decoding pipeline (see below)

Population Decoder

population_decoder_v2.py is the main analysis script. Run it directly:

python population_decoder_v2.py

It runs on all qualifying sessions (>15 Tier 2 MSN units, >150 trials) and produces per-session and cross-session figures in ~/data/neural_data/population_decoding/results/.

To regenerate summary plots from existing results without re-running the decoder:

# In population_decoder_v2.py, set:
PLOT_ONLY = True

Key Files

File	Purpose
population_decoder_v2.py	Full decoding pipeline with clock speed, history effects, confound controls
population_decoder.py	Base decoder utilities and anchor comparison
constants.py	Animal group assignments, region colors, QC thresholds
paths.py	Data directory paths — edit this for your machine
utils.py	Shared helper functions