neurocorr-dash

A dashboard for neurocorr, the DataJoint pipeline that reproduces the analysis of:

Dimitri Yatsenko, Krešimir Josić, Alexander S. Ecker, Emmanouil Froudarakis, R. James Cotton, and Andreas S. Tolias (2015). Improved estimation and interpretation of correlations in neural circuits. PLOS Computational Biology 11(3): e1004083. doi:10.1371/journal.pcbi.1004083

The recordings and the original analysis code are in atlab/cov-est (MIT License, © 2014 Tolias Lab). The article is published under the Creative Commons Attribution 4.0 License.

Pages

Page	Shows
Overview	the sites and active cells, and how many estimates each method has computed
Sites	each correlation set: directions, cells, active cells, and its mean sample correlation
Site	a site's correlation matrix (sample or diag) and the positions of its active cells
Validation	each estimator's cross-validated loss relative to a reference, per site, as in the paper's Fig 4
Figure 3	E–F: the sample correlations at the paper's example site (session 8828, site 2031) and their distribution
Figure 4	each simpler estimator's 10-fold validation loss relative to sparse + latent, across the 27 analyzed sites
Figure 5	the sparse + latent estimate at the example site: correlations, partial correlations, and its sparse and low-rank components
Figure 6	across sites: latent units and connectivity against the number of cells; partial against sample correlations; negative interactions
Figure 7	correlations and connectivity against differences in preferred orientation, lateral distance, and depth, at the five most connected sites
Figure S1	panel A: the cross-validated loss over the two hyperparameters at the example site, and the chosen values

The dashboard reads the pipeline's schema from the database with dj.VirtualModule, so it does not need the pipeline's code installed.

Run it

cp .env.example .env          # database host, user, password, and the schema prefix if any
uv venv && uv pip install -e .
uv run python app.py          # http://localhost:8052

Or in a container:

docker compose up --build     # http://localhost:8052

From inside a container, a database on the host is at host.docker.internal.

Variable	Meaning
DJ_HOST, DJ_PORT, DJ_BACKEND	the database (postgresql or mysql)
DJ_USER, DJ_PASS	credentials with read access to the pipeline's schema
DATABASE_PREFIX	the prefix of the pipeline's schema names, if it was declared with one
DJ_STORES	the pipeline's object store, as JSON, where its <npy@> arrays are
DASH_REQUESTS_PATHNAME_PREFIX, DASH_ROUTES_PATHNAME_PREFIX	serve the app under a path, behind a proxy

Development

uv pip install -e . pytest ruff
pytest                        # figures from synthetic data; no database needed
ruff check . && ruff format .

License

MIT; see LICENSE.