Dataset for "Meta-reinforcement learning via orbitofrontal cortex"

Neural and behavioral dataset for "Meta-reinforcement learning via orbitofrontal cortex" published in Nature Neuroscience.The zip file includes 3 folders: "Imaging", "Inactivation", "paAIP2"

"Imaging"Datasets are separated into different folders based on the mouse name (e.g. RH365) and imaging plane (deeper or shallower). Each data file in .npz format contains the following variables:zSR_readystart: OFC neural population activity. Deconvolved calcium signals are aligned to a ready cue onset and saved in 3d tensor format ([Neuron ID] x [Frames] x [Trials]). Neural activity of 10 sec before and 14 sec after the ready cue onset is included.SR_fps: Frame rate of neural activity data (frame per sec).pre_ready_sec: Pre-ready onset time (10 sec) included in zSR_readystart.post_ready_sec: Post-ready onset time (14 sec) included in zSR_readystart.pre_ready_frame: Pre-ready onset frames included in zSR_readystart. The frame number corresponds to pre_ready_sec.post_ready_frame: Post-ready onset frames included in zSR_readystart. The frame number corresponds to post_ready_sec.Ready2Ans_frameSR: Frame number from ready cue onset to go cue. This can be used to re-align activity to go cue onset.Ready2Choice_frameSR: Frame number from ready cue onset to choice. This can be used to re-align activity to choice onset.Trials4Use: Trial numbers (begins from 0) with good imaging quality. Use only these trials for neural activity analyses.a: Actions of mice during behavioral session. 1: right choice, 2: left choice, 3: alarm trial, 4: miss trialR: Reward outcomes during behavioral session. 1: reward, 0: no-rewardrprob: Reward assignment probability on right side in each trial.lprob: Reward assignment probability on left side in each trial.

"Inactivation"Datasets for optogenetic OFC inactivation experiments. Datasets are separated into different folders based on experiment type (Bilateral or Unilateral OFC inactivation), mouse type (e.g. ChrimsonR, tdTomato control), illumination type (e.g. iti_2sec, ready), and the mouse name (e.g. RH895). Each data file in .npz format contains the following variables:a: Actions of mice during behavioral session. 1: right choice, 2: left choice, 3: alarm trial, 4: miss trialR: Reward outcomes during behavioral session. 1: reward, 0: no-rewardOpto: 1: inactivation trials, 0: masking light onlyrprob: Reward assignment probability on right side in each trial.lprob: Reward assignment probability on left side in each trial.

"paAIP2"Datasets for OFC plasticity suppression experiments. Datasets are separated into different folders based on experiment type (Suppression at expert or during training), mouse type (e.g. paAIP2 or Control), and the mouse name (e.g. RH923). Each data file in .npz format contains the following variables:a: Actions of mice during behavioral session. 1: right choice, 2: left choice, 3: alarm trial, 4: miss trialR: Reward outcomes during behavioral session. 1: reward, 0: no-rewardrprob: Reward assignment probability on right side in each trial.lprob: Reward assignment probability on left side in each trial.

Details of the datasets and experimental conditions are described in the publication (Hattori et al., Meta-reinforcement learning via orbitofrontal cortex, Nature Neuroscience, 2023).

There was a report that Mac users may get "extra bytes at beginning" error when trying to unzip the file. A solution to this problem was to run "zip -FFv foo.zip --out fixed.zip" where foo.zip is the path of the file downloaded from Zenodo. Then, run "unzip fixed.zip".