A Brain-wide Neuronal Spiking and Behavioral Dataset for Working Memory-Specific Activation and Reactivation in Mice

These files are named with a `_supp` suffix corresponding to the associated main NWB file. The candidate sparse-firing units are not included in the main curated unit count unless users explicitly combine them with the main release using their own post hoc filtering criteria.Task

Summary

Mice performed an olfactory delayed paired-association task. In each trial, one of two sample odors was presented, followed by a delay period of either 3 or 6 seconds. A test odor was then presented, followed by a response window.

Mice were trained to lick during the response window on paired trials and to withhold licking on unpaired trials. Trials were organized so that paired and unpaired trials occurred at a 1:1 ratio. File FormatAll data files are provided in Neurodata Without Borders (`.nwb`) format.

Each main NWB file corresponds to one recording session and follows a session-based naming convention:sub-m<mouse_id>_ses-<date>_task-DPA.nwbSupplementary files containing candidate sparse-firing units use the corresponding `_supp` suffix:sub-m<mouse_id>_ses-<date>_task-DPA_supp.nwbThe main data files occupy approximately 17,353 MB in total. The supplementary candidate sparse-firing unit files occupy approximately 8,107 MB in total.Main NWB StructureEach NWB file is organized into several major groups.`general`Contains session, subject, laboratory, experimenter, and recording-device metadata.

Subject metadata include species, strain, sex, age, genotype, and subject ID. Device metadata describe the Neuropixels probes used in each recording session.`intervals/trials`Contains a trial-by-trial behavioral table. Each row corresponds to one trial.

Key fields include:`start_time`: trial/sample onset time in seconds`stop_time`: delay offset / test-cue onset time in seconds`start_timestamp` and `stop_timestamp`: corresponding hardware clock ticks`sample_cue`: sample odor identity`test_cue`: test odor identity`is_correct`: trial outcome correctness`lick_response`: lick or no-lick response`delay_duration`: 3-second or 6-second delay condition`well_trained`: whether the trial falls within the predefined high-performance analysis windowIn this dataset, `start_time` and `stop_time` define the analysis interval of the working memory delay, from trial/sample onset to delay offset/test-cue onset, rather than the full behavioral trial including the response window.

Primary time fields are stored in seconds, and original hardware timestamps are retained as auxiliary columns. `processing/ecephys`Contains precomputed firing-rate matrices aligned to task events. These matrices are provided at two temporal resolutions for rapid visualization and analysis without recomputing firing rates from spike times.The firing-rate matrices use a flattened layout. The first row records the trial index, the second row records the within-trial time-bin index, and subsequent rows contain unit firing-rate traces.

Users should treat the first two rows as indexing metadata rather than neuronal firing-rate traces. `units`Contains spike times and metadata for sorted single units. Each row corresponds to one unit. The table includes:Unit and cluster identifiersSpike times in seconds relative to session onsetProbe channel and depth informationKilosort2-derived sorting labels and contamination metricsAdditional unit-level quality-control metricsCCFv3 coordinates and anatomical labels for anatomically registered unitsFor sessions with multiple probes, cluster identifiers from later probes were offset to maintain uniqueness within a session.Unit Quality MetricsThe main curated unit set was selected using automated Kilosort2-based criteria, including contamination-related measures, refractory-period structure, and a minimum mean firing-rate threshold.

Additional quality metrics, derived from a development of the Allen Brain Observatory ecephys spike sorting pipeline, are provided in the NWB `units` table to support user-defined post hoc filtering.Available metrics include, among others:`firing_rate``n_spikes``presence_ratio``amplitude_cutoff``isi_violations``snr``isolation_distance``d_prime``nn_hit_rate`These metrics allow users to apply custom filtering criteria according to their analysis goals, such as requiring stable session-long activity, reducing potential missed-spike effects, or applying more conservative cluster-separation thresholds.

No single universal post hoc cutoff is imposed beyond the automated inclusion criteria used for the main curated release.Anatomical InformationAnatomical annotations are provided using Allen Mouse Brain CCFv3 labels at multiple hierarchical resolutions. These labels allow users to filter units by broad anatomical divisions or by more specific brain regions, depending on the desired level of anatomical precision.For anatomically registered units, CCFv3 coordinates along the anterior-posterior, dorsal-ventral, and medial-lateral axes are included in the NWB `units` table.

Units without high-confidence anatomical registration do not have reliable anatomical assignments and should not be used for region-specific or continuous-space anatomical analyses. These units are retained for analyses that do not require brain-region assignment, such as global temporal analyses.Behavioral Inclusion and Trial SelectionRecordings were performed after mice reached proficient task performance, defined by a moving-window behavioral criterion (at least 75% correct over 40 consecutive trials).

Within each recording session, the `well_trained` trial-level field indicates whether a trial falls within the predefined high-performance analysis window.Recommended trial-selection practices:For analyses of correct task performance, use hit and correct-rejection trials within the well-trained window.For error-trial analyses, users may retain error trials from the full recording session depending on the analysis goal.Trials outside the well-trained analysis window are retained for completeness and may be useful for trial-by-trial variability analyses, but they should not be interpreted as learning-stage data.The sample cue onset serves as the primary alignment reference for peri-event analyses.Supplementary Candidate Sparse-firing Unit FilesIn addition to the main curated NWB files, supplementary NWB files provide candidate sparse-firing units with lower mean firing rates.

These files preserve the session-matched structure of the main files and include metadata, quality metrics, and matched behavioral records.These candidate sparse-firing units were not further processed into the main curated dataset. Users may apply their own inclusion criteria and, where appropriate, combine them with the main curated units for downstream analyses. Code and Demonstration ScriptsExample MATLAB scripts for accessing and visualizing the data are available at:gitee.com/XiaoxingZhang/hierarchical_replay_NSB_2025The repository includes demonstration scripts for:Loading NWB files and plotting single-unit rasters/PSTHsIdentifying memory-content-selective neurons and summarizing their regional distributionDetecting spike coupling and plotting cross-correlogramsVisualizing anatomical coverage in CCFv3 spacePlotting population activity dynamicsThe code was developed and tested in a MATLAB environment.

The tested versions and setup instructions are documented in the repository README.Python users can also access the NWB files using PyNWB and standard scientific Python packages such as NumPy, pandas, and SciPy.Version V2 Update

Notes

The current data files include several file-level improvements over the previous version, intended to support broader reuse:Trial-table timestamp fields are included in NWB column metadata for improved compatibility with MatNWB validation.Additional unit-level quality-control metrics are included in the `units` table.CCFv3 coordinates are included for anatomically registered units.Supplementary NWB files provide candidate sparse-firing units for users who need more flexible firing-rate thresholds.

Citation

If you use this dataset or the associated code in your work, please cite the original research paper:HUANG E, XU D, ZHU H, et al. Hierarchical Replay of Multi-regional Sequential Spiking Associated with Working Memory[J/OL]. Neuroscience Bulletin, 2025: 1–20. DOI:10.1007/s12264-025-01524-y.