Network-aware self-supervised learning enables high-content phenotypic screening for genetic modifiers of neuronal activity dynamics

This dataset contains the data archive for the paper titled "Network-aware self-supervised learning enables high-content phenotypic screening for genetic modifiers of neuronal activity dynamics".

The data archive is comprised of six folders:

The first folder is dataset_statistics, which contains JSON files that have the associated statistics for each plate, which is used in the inference for the self-supervised learning models.

The second folder is figure1_data, which contains the information necessary to recreate the panels in Figure 1 of the paper.

The third folder is metadata_files, which contains CSV files that provide a mapping for treatment to plate and well ID for all experiments. 

The fourth folder is called model_checkpoints. This folder contains all of the model checkpoints, which can be read for inference following the code found at the associated GitHub repository. Within the model checkpoint folder, the subfolders contain the checkpoints for models trained on each of the datasets, from either the neuroactive stimulation experiment, the simulated experiment, or the CRISPRi screen.

The fifth folder plexus_embeddings, contains h5ad files containing either the plexus model embeddings or manual features. The .obs DataFrame also contains all of the associated metadata for each of the experiments.

The sixth and final folder is processed_zarr_files, which contains all of the Zarr files (one per 384-well plate across the experiments). The Zarr files are hierarchical and have the form Plate / Well / Field-of-view. For each field of view, there are datasets containing the raw GCaMP6m traces, the baseline corrected traces, the cell segmentation masks, the inferred spike events, the nuclei masks in the case of the CRISPRi screening dataset, and the coefficients associated with the autoregressive model used to infer spike events.

This data archive is leveraged by the Plexus GitHub Repository to recreate the results presented in the paper.