Aggression encoding

Analysis code accompanying Precision editing of an aggression-encoding network relay suppresses violent action in mice, by Yael S. Grossman, Austin Talbot, Neil M. Gallagher, Kathryn K. Walder-Christensen, Gwenaëlle E. Thomas, Alexandra Fink Skular, Scott J. Russo, David E. Carlson^†, and Kafui Dzirasa^†.

^* Joint first authors; these authors contributed equally.

^† Joint senior authors; these authors jointly supervised this work.

Start here

Analysis	Location
NMF implementations and historical training/projection analyses	Network analysis
Component selection and single-/two-region comparisons	Regional analysis
Specific source locations and verification limits	Analysis index
Closed-loop network mediation and regional theta controls, Figure 5e–g	closed_loop_mediation.ipynb
ESR1/CNO mediation, Supplementary Figure S8	esr1_mediation.ipynb
Inputs, methods, and run instructions for mediation	Mediation README

Organization and naming

analysis/
  network/
    models/
    experiments/
    thresholds/
    projections/
    joint_models/
    feature_experiments/
    archive_sources/
  regional/
    component_selection/
    single_region/
    two_region/
  mediation/
    closed_loop_mediation.ipynb
    esr1_mediation.ipynb
docs/
  analysis_index.md
  source_manifest.json
tools/
  check_release.py
  clear_notebook_outputs.py

LFP preprocessing

The existing MATLAB preprocessing and feature-generation pipeline is maintained in carlson-lab/lpne-data-analysis. Its entry points include runFeaturePipeline.m, preprocessData.m, and saveFeatures.m.

The learning framework is also available in carlson-lab/encodedSupervision.

Running analyses

Start with the network and regional analyses listed in the analysis index, followed by the mediation instructions. Inputs must be obtained separately. The manuscript directs data requests to Kafui Dzirasa ([email protected]) or David Carlson ([email protected]).

The network-modeling code retains its original environment and path assumptions. The manuscript describes Python 3.7 and TensorFlow 2.4 for model learning; this is historical context, not a tested modern installation recipe. Some scripts refer to local modules or fitted-model files that are not in this release. See the analysis index before choosing an entry point.

Mediation p-values are simulation-based and can vary across random seeds. The manuscript reports values from the original analysis run.

License

GPL-3.0.