Data from: Exponential history integration with diverse temporal scales in retrosplenial cortex supports hyperbolic behavior
Cite this dataset
Danskin, Bethanny et al. (2024). Data from: Exponential history integration with diverse temporal scales in retrosplenial cortex supports hyperbolic behavior [Dataset]. Dryad. https://doi.org/10.5061/dryad.jq2bvq8fp
Abstract
Animals rely on their experience to guide their next choice. In foraging-type tasks guided by history-dependent value, these experiences are typically integrated such that the weights of past events initially decay quickly over time but show a longer tail than expected by exponential decay. Rather, such integration is better described by a hyperbolic function. Hyperbolic integration affords sensitivity to both recent environmental dynamics and long-term trends, however the mechanism by which the brain implements this hyperbolic integration is unknown. We trained mice on a history-dependent, value-based decision task and found that the mice indeed showed hyperbolic decay on their weighting of past experience. However, the activity of history-encoding cortical neurons showed weighting with exponential decay. In resolving this apparent mismatch, we observed that cortical neurons encode history information heterogeneously across a wide variety of exponential time-constants, with the retrosplenial cortex (RSC) overrepresenting longer time-constants compared to other areas. A model that combines these diverse timescales of exponential history integration can recreate the heavy-tailed, hyperbolic history integration observed in behavior. In particular, time-constants of RSC neurons best matched the behavior, and optogenetic inactivation of RSC uniquely reduced the use of history information. These results indicate that behavior-relevant history information is maintained in neurons across multiple timescales in parallel, and suggest that the neural population in RSC is a critical reservoir of this information guiding decision-making.
README: Exponential history integration with diverse temporal scales in retrosplenial cortex supports hyperbolic behavior.
This repository contains the code and processed data necessary to regenerate figures from:
Danskin B.P. et al. Exponential history integration with diverse temporal scales in retrosplenial cortex supports hyperbolic behavior (2023). Science Advances. DOI: 10.1126/sciadv.adj4897 .
Data deposited to Dryad, doi:10.5061/dryad.jq2bvq8fp .
Code deposited to Zenodo via Github, doi:10.5281/zenodo.8152662 .
All necessary analysis functions are contained in the associated notebook, or in bdanskin.py .
Repository layout
. BDanskin_SciAdvances_2023 ### source folder\
|- py_code/ ### expected location of bdanskin.py\
|- chr2_optogenetics/ ### zipped directory containing pkl files for analysis and plotting\
|- hattori_datasets_behavior/ ### nested zipped directory containing pkl files for analysis and plotting\
|- hattori_datasets_xarray/ ### nested zipped directory containing netcdf files in the xarray format; imaging and behavior data for analysis\
|- hattori_datasets_xarray_cellfits/ ### zipped directory containing netcdf files in the xarray format; the outputs of analysis and used for plotting\
| bdanskin_analysis_all.ipynb ### notebook file to run all analysis\
| bdanskin_plot_figure_1.ipynb ### notebook file to load analysis outputs and plot Fig 1\
| bdanskin_plot_figure_2.ipynb ### notebook file to load analysis outputs and plot Fig 2\
| bdanskin_plot_figure_3.ipynb ### notebook file to load analysis outputs and plot Fig 3\
| bdanskin_plot_figure_4.ipynb ### notebook file to load analysis outputs and plot Fig 4
Data files
Behavior data
|--- hattori_datasets_behavior//imaging_processed_data_behonly//(mouse_id)_(date)_(brain_area).pkl
Data was recorded during mouse in vivo imaging session. Behavior data is saved in Python pickle files in the pandas dataframe format.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
|-- hattori_datasets_behavior//all_behavior_df.pkl
The concatenation of the behavior data from all imaging sessions. Saved as a Python pickle file in the pandas dataframe format.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
Area: the area targeted for inactivation in that session
Date: the date of the session in YYMMDD
Mouse: the id of the mouse, string
|-- hattori_datasets_behavior//all_behavior_contingency_df.pkl
The concatenation of the behavior data from all imaging sessions, with metadata on the reward contingency. Saved as a Python pickle file in the pandas dataframe format.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
6:1 : boolean for the reward contingency with a 6:1 probability
3:1 : boolean for the reward contingency with a 3:1 probability
Area: the area targeted for inactivation in that session
Date: the date of the session in YYMMDD
Mouse: the id of the mouse, string
|-- hattori_datasets_behavior//early_behavior_df.pkl
The concatenation of the behavior data from early imaging sessions, before expert criterion. Saved as a Python pickle file in the pandas dataframe format.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
Area: the area targeted for inactivation in that session
Date: the date of the session in YYMMDD
Mouse: the id of the mouse, string
Imaging data
|-- hattori_datasets_xarray//(mouse_id)_(date)_(brain_area)_readyaligned.nc
Two-photon calcium imaging data, aligned to trial ‘ready’ period period.
Saved as a netcdf file in the xarray dataset format, bundling neural activity for all cells recorded in a session with that session’s behavior metrics.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
spks: deconvolved estimate spike rate, aligned to ready cue onset. First data point corresponds to 2sec before ready cue onset. Last data point corresponds to 10sec after ready cue onset
z_spks: zscore normalized spks, each cell zscored over the entire session
|--- hattori_datasets_xarray//preprocess_data_2s//(mouse_id)_(date)_(brain_area)_(temporal_alignment)_2s.nc
Two-photon calcium imaging data, with the 2-second average of neural activity aligned to various trial epochs, across all trials and cells for the session
Saved as a netcdf file in the xarray dataset format, bundling neural activity for all cells recorded in a session with that session’s behavior metrics
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
spks: deconvolved estimate spike rate, aligned to ready cue onset. Mean over 2 sec
z_spks: zscore normalized spks, each cell zscored over the entire session. Mean over 2 sec
Inactivation data
|-- chr2_optogenetics//ChR2_session_summary.pkl
Behavior data from the inactivation sessions includes variable ‘a’ and ‘R’ as well as the inactivation condition ‘Opto’ and metadata associated with the inactivation parameters.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
Opto: inactivation status for the trial. (cortical inactivation, control inactivation) = (1,0)
Inac Type: trial-aligned timing of inactivation, either aligned to the ready-answer periods, the intertrial interval, or the both. String = (ra, iti, 0i1ra)
Area: the area targeted for inactivation in that session
Date: the date of the session in YYMMDD
Mouse: the id of the mouse, string
Analysis outputs
All of the outputs associated with the analysis are located in the respective directories. Reference the Jupyter notebook files for the specific analysis performed and the generation of the figures from the output data.
Intermediate processing files derived from and used by the notebook files are as follows:
Behavior data
|-- hattori_datasets_behavior//behlogit_exp_hyp_history.nc
Model performance and parameter estimates for the exponential and hyperbolic behavior models, compared across history terms. Saved as a netcdf file in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_imaging_15_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the imaging data. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_imaging_15_hist_constrained.pkl
Model performance and parameter estimates for the temporal behavior models on the imaging data with different optimization parameters. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_imaging_compare_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the imaging data, compared across history terms. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_rl_base_15_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the simulated behavior. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_rl_base_compare_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the simulated behavior, compared across history terms. Saved in Python pickle files in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_rl_full_15_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the simulated behavior. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//behlogit_rl_full_compare_hist.pkl
Model performance and parameter estimates for the temporal behavior models on the simulated behavior, compared across history terms. Saved in Python pickle files in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//logistic_base_rl_10_hist.pkl
Model performance and parameter estimates for the logistic regression on the simulated behavior. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//logistic_full_rl_10_hist.pkl
Model performance and parameter estimates for the logistic regression on the simulated behavior. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//logistic_imaging_10_hist.pkl
Model performance and parameter estimates for the logistic regression on the imaging behavior. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
|-- hattori_datasets_behavior//quasihyp_subsample_cv.nc
Model performance for the quasihyperbolic behavior model, with the outputs from 1000 iterations. Saved as a netcdf file in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#QuasiHyp.
Used for in Jupyter notebook file bdanskin_plot_figure_3.ipynb.
|-- hattori_datasets_behavior//rl_simulation_full_base_all_imaging_(n_trials).pkl
Output of the behavior simulation with either 600 or 10000 trials. Saved in Python pickle files in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#BehaviorAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_1.ipynb.
Imaging data
|--- hattori_datasets_xarray_cellfits//ready_cellfits_RCp1//RCp1_2s_(mouse_id)_(date)_(brain_area).nc
Model performance and parameter estimates for the null cell model with one past history term. Saved as netcdf files in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#ImagingAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_2.ipynb.
|--- hattori_datasets_xarray_cellfits//ready_shuffle//partition_300_2s_(mouse_id)_(date)_(brain_area)_15hist_shuffle.nc
Model performance and parameter estimates for the cell temporal history model on trial-shuffled data. Saved as netcdf files in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#ImagingAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_2.ipynb and bdanskin_plot_figure_3.ipynb
|---- hattori_datasets_xarray_cellfits//ready_cellfits//cv_compare_mdl//sse_(mouse_id)_(date)_(brain_area)_15hist.nc
Detailed model performance, compared across multiple versions of the cell temporal history model. Saved as netcdf files in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#ImagingAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_2.ipynb and bdanskin_plot_figure_3.ipynb
|---- hattori_datasets_xarray_cellfits//ready_cellfits//halves_compare_mdl//sse_(mouse_id)_(date)_(brain_area).nc
Model performance and parameter estimates of the cell history model across temporal windows within a session. Saved as netcdf files in the xarray dataset format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#ImagingAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_2.ipynb and bdanskin_plot_figure_3.ipynb
Inactivation data
|-- chr2_optogenetics//ChR2_wsls.pkl
Win-Stay and Lose-Switch probabilities under optogenetic inactivation. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#OptoAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_4.ipynb.
|-- chr2_optogenetics//logistic_ChR2_l1l2_subsample1000.pkl
Model performance and parameter estimates for the logistic regression on inactivation data. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#OptoAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_4.ipynb.
|-- chr2_optogenetics//opto_behlogit_RA_10hist.pkl
Model performance and parameter estimates for the temporal behavior models on inactivation data. Saved as a Python pickle file in the pandas dataframe format.
Generated in Jupyter notebook file bdanskin_analysis_all.ipynb#OptoAnalysis.
Used for in Jupyter notebook file bdanskin_plot_figure_4.ipynb.
Methods
This dataset includes the processed data from in vivo two-photon calcium imaging, optogenetic stimulation, and behavioral choices from mice performing a decision-making task. The imaging data is trial-aligned fluorescence and deconvolved activity traces, as well as further analysis products of that data. The optogenetics data includes the stimulus parameters and the trial-by-trial choices made by the mice. Further details for the processing and acquisition can be found in the associated publication.
Usage notes
Analysis Code
Code written in Python available in the linked Zenodo repository.
Behavior data
Data was recorded during mouse in vivo imaging or inactivation sessions. Behavior data is saved in Python pickle files in the pandas dataframe format.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
Imaging data
Processed two-photon calcium imaging data, aligned to trial ‘ready’ period or ‘choice’ period. Saved as a netcdf file in the xarray dataset format, bundling neural activity for all cells recorded in a session with that session’s behavior metrics.
spks: deconvolved estimate spike rate, aligned to ready cue onset. First data point corresponds to 2sec before ready cue onset. Last data point corresponds to 10sec after ready cue onset.
z_spks: zscore normalized spks, each cell zscored over the entire session.
Inactivation data
Behavior data from the inactivation sessions includes variable ‘a’ and ‘R’ as well as the inactivation condition ‘Opto’ and metadata associated with the inactivation parameters.
a: choice type for each trial (Right choice, Left choice, Alarm, Miss) = (1,2,3,4)
R: reward outcome type for each trial. (Reward delivered, Reward not delivered) = (1,0)
Opto: inactivation status for the trial. (cortical inactivation, control inactivation) = (1,0)
Inac Type: trial-aligned timing of inactivation, either aligned to the 'ready-answer' periods, the 'intertrial interval', or the both. String = ('ra', 'iti', '0i1ra')
Area: the area targeted for inactivation in that session
Analysis outputs
All of the outputs associated with the analysis are located in the respective directories. Reference the Jupyter notebook files for the specific analysis performed and the generation of the figures from the output data.
Funding
David and Lucile Packard Foundation
National Institute of Mental Health, Award: R03 MH120426
National Institute of Mental Health, Award: F31 MH116613
National Eye Institute, Award: R01 EY025349
National Eye Institute, Award: P30 EY02259
National Institute on Deafness and Other Communication Disorders, Award: R01 DC018545
National Institute of Neurological Disorders and Stroke, Award: R01 NS091010
National Institute of Neurological Disorders and Stroke, Award: R01 NS125298
Simons Foundation
Uehara Memorial Foundation
Achievement Rewards for College Scientists Foundation