CellTuneDepot

CellTuneDepot is a curated collection of multiplexed imaging datasets with expert-validated labels, intended for benchmarking, model training, and reproducible analysis.

Download CellTuneDepot-v0.2 (~5 GB ZIP)

CellTuneDepot is organized into two main sections:

1. Manual Labels

Fully manually labeled crops/images.

2. High-Quality Labels

Semi-supervised labels generated with CellTune for entire datasets.

Each section is then further subdivided by dataset.

Dataset Composition

Each dataset includes:

1. Images Folder

Structure: Contains subfolders for each image, aka Field of View (FOV).
Contents: Each image subfolder includes:
- ~40+ TIFF images representing various biological markers.
- Segmentation labels image (‘segmentation_labels’) for delineating individual cells.

2. Tables Folder

Structure: Contains CSV files with key data annotations and features.
Contents:
- Channels.csv: List of channels in the dataset with their expected expression cell type or function (Expected_Expression)
- CellTypes.csv: List of cell types in the dataset with their expression definition (PrimaryMarker)
- Labels.csv: Table of manual cell labels. Columns= (image, cellID, class). [Manual Labels Datasets Only]
- Populations.csv: Table of cell classification labels for all cells. Columns= (image, cellID, class). [High-Quality Labels Datasets Only]
- CellTable.csv: Mean expression of each protein channel on all the cells in the dataset. [High-Quality Labels Datasets Only]

Cell Types

CellTypes_ALL.csv contains a list of all cell types across datasets and the expression used to define them (PrimaryMarker).

Datasets associated with unpublished studies will be released upon publication of their respective manuscripts.

See the Methods section of our manuscript for more details on all datasets.