Information-theoretic analysis of spike trains from the retino-geniculate synapse

Introduction & References

The scripts in this repository perform information-theoretic analyses of spike trains recorded from the retinogeniculate synapse of the cat. Scripts produce results on the correlation of local information storage and information transfer at the synapse, which are presented in

• Wollstadt, P., Rathbun, D. L., Usrey, W. M., Moraes Bastos, A., Lindner, M., Priesemann, V., Wibral, M. (2022). Information-theoretic analyses of neural data to minimize the effect of researchers' assumptions in predictive coding studies. ArXiv preprint arXiv:2203.10810 [q-bio.NC]

The spike data used by the scripts is described in detail in

• Rathbun, D. L., Warland, D. K., Usrey, W. M. (2010). Spike Timing and Information Transmission at Retinogeniculate Synapses. Journal of Neuroscience, 30(41) 13558-13566. https://doi.org/10.1523/JNEUROSCI.0909-10.2010

and can be obtained from

https://github.com/scottiealexander/PairsDB.jl/raw/main/data/

using the script lstc_analysis/download_spike_data.py.

The code uses the following python packages to perform information-theoretic analyses:

• IDTxl: Wollstadt, P., Lizier, J. T., Vicente, R., Finn, C., Martinez-Zarzuela, M., Mediano, P., Novelli, L., Wibral, M. (2019). IDTxl: The Information Dynamics Toolkit xl: a Python package for the efficient analysis of multivariate information dynamics in networks. Journal of Open Source Software, 4(34), 1081. https://github.com/pwollstadt/IDTxl
• pyentropy: Ince, R. A. A., Petersen, R. S., Swan, D. C. and Panzeri, S. (2009). Python for Information Theoretic Analysis of Neural Data", Frontiers in Neuroinformatics 3(4) https://code.google.com/archive/p/pyentropy/

Requirements

Code in the folder lstc_analysis requires Python 3 and packages described in the lstc_analysis/requirements.txt. Additionally, download and install IDTxl v1.4 (not yet on pypi).

Code in the folder local_bias_correction requires Python 2 and packages described in the lstc_analysis/requirements.txt. Additionally, the folder contains a slightly modified version of the pyentropy toolbox, which is called by the respective analysis scripts (see https://code.google.com/archive/p/pyentropy/ for details and original source code).

Data

Raw data are obtained from the online repository

https://github.com/scottiealexander/PairsDB.jl/raw/main/data/

by calling python3 lstc_analysis/download_spike_data.py. The script downloads raw spike timings from the repository, generates spike train data, and saves it into the data folder specified in paths.json. For details on the data see Rathbun et al. (2010).

Running the analysis

Data and output paths are defined in paths.json. All analysis scripts read in this file. To change input- or output paths, simply modify the paths provided in paths.json.

Prior to starting the analysis, data have to be downloaded via python3 lstc_analysis/download_spike_data.py.

To estimate local, bias-corrected active information storage (AIS) and transfer entropy, run the following analysis in the described order (note that step 3 calls an older version of pyentropy and requires Python2):

1. python3 lstc_analysis/estimate_te_ais.py $cell_pair: Run active information storage (AIS) and transfer entropy (TE) algorithms on input data to optimize non-uniform past-state embeddings. Estimate AIS in the RGC spike train and TE from RGC to LGN spike train.
2. python3 lstc_analysis/estimate_local_values.py $cell_pair: Calculate local TE/AIS, encode and save past states for local PT correction in the next step
3. python2 local_bias_correction/estimate_local_bias_correction.py $cell_pair: Calculate bias correction for local TE/AIS using pyentropy

To estimate the partial information decomposition of the TE source and target past states, and the current value run

python3 lstc_analysis/estimate_pid.py $cell_pair

Each call to any of the scripts performs the respective analysis step for the cell pair ID provided. To run all analyses over all cell pairs call the scripts

./lstc_analysis/run_estimation.sh ./local_bias_correction/run_bias_correction.sh

To generate figures and results from estimated information-theoretic quantities for individual pairs, run:

• python3 generate_results_correlation.py: Calculate and plot local storage-transfer correlations (LSTC)
• python3 generate_results_isi.py: Calculate inter-spike intervals (ISI) and plot corresponding local AIS and TE estimates
• python3 generate_results_sta.py: Calculate spike-triggered averages (STA) for estimated local AIS and TE estimates
• python3 generate_results_tuples.py: Generate results for spike tuples

When all results have been generated, run

• python plots_for_paper.py: Generate plots and outputs from paper
• python plot_classification_relayed.py: Run classification of RGC spikes into relayed and non-relayed from AIS and spike stats

For convenience, the script

./lstc_analysis/generate_all_results.sh

runs all scripts to generate the results and Figures shown in the paper.

For details on the analysis refer to Wollstadt et al. (2022).