Data from: Slow cortical dynamics generate context processing and novelty detection

Published Jun 25, 2025 on Dryad. https://doi.org/10.5061/dryad.xsj3tx9q6

Abstract

The cortex amplifies responses to novel stimuli, compared to those elicited by redundant stimuli—a function key to efficiently processing sensory information and building predictive models of the environment. Novelty detection is measured by the “Mismatch Negativity” (MMN) signal, the reduction of which represents the best functional biomarker of schizophrenia. To better understand the circuit mechanisms of novelty detection, we used an auditory “oddball” paradigm and two-photon calcium imaging to measure responses to simple and complex stimuli in neuronal populations across the mouse auditory cortex. Stimulus statistics and complexity generated differences in neural response profiles across contexts and auditory cortical subregions. At the population level, neuronal ensembles separately and reliably encoded basic auditory features, as well as temporal context. Interestingly, stimuli-evoked responses were particularly long-lasting, persisting after the stimuli ended and affecting responses to future stimuli. These slow network dynamics encoded stimulus history and temporal context, generating novelty detection. Recurrent neural network models trained on the oddball task exhibited slow network dynamics and recapitulated the biological data, including context selectivity, MMN, and stimulus-specific adaptation. We conclude that the slow dynamics of recurrent cortical networks underlies temporal processing of stimuli, a canonical computation that gives rise to context-specific encoding and novelty detection.

https://doi.org/10.5061/dryad.xsj3tx9q6

Description of the data and file structure

Each .zip file contains all datasets from a single mouse. Mice M1 - M10 (“missmatch”) are oddball data that were used for Figures 1, 2, 3A-3F. Mice M226, M4264, M4265, M4266, M4371, M4372 are variable ISI experiments (“echo”) used for Figures 3 H-3J.

Each “M” file contains all the datasets collected from the given mouse. Each dataset contains processed calcium imaging data within files ending with “_results_cnmf_sort.mat”, along with stimulus and imaging information within files ending with “_processed_data.mat”. Finally, ”AC_data_list.csv” file contains additional information about datasets and mice.

Extended details about the experimental methods and analysis are described in detail in the attached publication.

Extended descriptions for data files:

“*_results_cnmfsort.mat”: files containing processed calcium imaging data. Data was motion corrected with modified Suite2P motion correction, and single neuron traces were demixed using the CaImAn pipeline. After demixing, individual cells were selected using signal-to-noise (SNR) metrics with a custom code (caiman_sorter). Below listed are key variables needed for data analysis

est: raw CaImAn outputs (documentation https://caiman.readthedocs.io/en/latest/)
- A: neuronal spatial components corresponding to regions of interest (ROIs) in calcium imaging data
- C: neuronal temporal components
- YrA: temporal residual components. Raw calcium imaging traces are reconstructed with C + YrA
- b: spatial background components
- f: temporal background components
proc: additional parameters after sorting caiman outputs from est via the caiman_sorter GUI
- idx_components: list of indexes containing the selected components from est.A, est.C, est.YrA, etc.
- idx_components_bad: list of discarded components
- SNR2_vals: computed SNR values with a more stable algorithms that one in CaImAn, and used for selection
- deconv: contains deconvolved traces for all CaImAn extracted components using one of the three algorthms (smooth_dfdt (default), foopsi, MCMC)
  - smooth_dfdt: simple algorithm that uses smoothed, rectified first derivative to infer the neuronal firing rate from the raw caclium imaging traces
    - S: trace of deconvolved firing rate proxy data
ops: contains parameters used for CaImAn and postprocessing
- init_params_caiman: contains CaImAm parameters
- eval_params2**: parameters used for cell selection in caiman_sorter GUI
- deconv: contains deconvolution parameters
  - smooth_dfdt.params.gauss_kernel_sigma: standard deviation of gaussian smoothing kernel during deconvolution

“*_processed_data.mat”: files containing input stimulus information, imaging parameters, and recorded behavior (locomotion).

data: contains stimulus information data
- frame_data: imaging parameters
  - volume_period: volume imaging period for each dataset (number of planes*frame period)
  - frame_period: frame imaging period
  - frame_times_mpl: times of each frame, for both single plane or multiplane data
- stim_params: contains data on oddball parameters
  - MMN_freq: index of frequencies used for oddball
  - stim_duration: stimulus duration
  - isi: interstimulus interval duration
  - trials: number of trials in oddball
  - start_freq: in Hz, the first and lowest frequency
  - increase_factor: octave difference between frequencies
  - num_freqs: number of total frequencies
- stim_times_frame**: the times of each stimulus in terms of frame index
- stim_times_volt: the times of each stimulus in terms of ms
- volt_data_binned: voltage data recorded with microscope containing raw stimulus times, locomotion, etc.
ops: parameters used for processing stimulus information data

“*_registration_cmnf.mat”: files containing spatial registration information to map the same cells recorded across datasets from the same field of view (FOV). These were not used in any of the analysis published, but are included here nevertheless. The FOV number for each dataset is listed in the ”AC_data_list.csv” file, in the “FOV” column.

A_list: cell containing all input spatial footprints “A” from corresponding datasets.
fname_list: list of dataset names for the corresponding FOV
reg_out: cell containing the outputs of registration
- idx1: combined “A” file across datasets where all unique spatial footprints have unique indexes.
- idx2: array containing registration information, where each row is a unique footprint, with its contents containing the indexes of those same footprints from the input datasets.
  *

“AC_data_list.csv”: file contains additional information about datasets and mice. “n/a” values in the ”AC_data_list.csv” file correspond to “data not available” because it was not recorded.

Extended details on experiments and analysis are provided in the attached manuscript, and accompanying code in the GitHub libraries linked below.