MOAProfiler_IN_MOA_JUMP and CellPaintSSL_IN_MOA_JUMP: Preprocessed Cell Painting MOA_JUMP plate (images and metadata)

Summary

This record provides preprocessed Cell Painting data for the MOA_JUMP plate, used in our study as an external control. Images and metadata for this plate are sourced from the C_BIN2_3PLANES (z-projected) PLATE 1 subset of the JUMP Cell Painting dataset cpg0002-jump-scope (Tromans-Coia et al., 2023), distributed via the Cell Painting Gallery (Weisbart et al., 2024) on the Registry of Open Data on AWS. The original cpg0002 images were prepared specifically for running the MOAProfiler_Feature_Extractor and CellPaintSSL models in a strictly zero-shot regime (no fine-tuning). Code for running the embedding pipelines on these data is available in the MOAProfiler_Feature_Extractor repository (https://github.com/XeniaKuznetsova/MOAProfiler_Feature_Extractor) and the CellPaintSSL repository (https://github.com/Bayer-Group/CellPaintSSL).

Two archives are provided:

MOAProfiler_IN_MOA_JUMP – input images and metadata for MOAProfiler_Feature_Extractor.

CellPaintSSL_IN_MOA_JUMP – input images and metadata for CellPaintSSL.

These data can be used as a ready-to-use MOA_JUMP plate for benchmarking Cell Painting encoders.

Contents

The Zenodo record contains two compressed archives:

MOAProfiler_IN_MOA_JUMP.tar.gz

Expands into:

MOAProfiler_IN/

images/ – preprocessed single-plane 16-bit TIFF images (five fluorescence channels, 1080×1080 px), organised by plate, well, and site in the format expected by MOAProfiler_Feature_Extractor (one TIFF per channel and field of view, e.g. ch1.tif … ch5.tif).

manifest.csv – manifest linking each image to plate, well, site, channel and relative file paths (e.g. images/...), used by the MOAProfiler_Feature_Extractor code to locate input files and assemble per-field five-channel stacks.

CellPaintSSL_IN_MOA_JUMP.tar.gz

Expands into:

CellPaintSSL_IN/

images/ – preprocessed five-channel TIFF stacks (shape 5×H×W, 16-bit, 1080×1080 px), one multi-channel file per field of view, in the channel order expected by our CellPaintSSL pipeline.

dataset.csv – manifest linking each image to plate, well, site, perturbation_id (compound), MOA and other metadata, in the format expected by the official CellPaintSSL codebase.

How the data were generated

Source data were taken from the C_BIN2_3PLANES (z-projected) PLATE 1 subset of the JUMP Cell Painting dataset cpg0002-jump-scope, distributed via the Cell Painting Gallery on the Registry of Open Data on AWS under the CC0 1.0 Universal public domain dedication. In this acquisition condition, cells were imaged confocally with 2×2 binning and three z-planes, and the released images are already provided as 2D z-projections of these three planes.

 Images were illumination-corrected but not resized, as the released C_BIN2_3PLANES images are already 1080×1080 pixels, which we used directly for both MOAProfiler and CellPaintSSL preprocessing

For the MOAProfiler_IN_MOA_JUMP archive, the z-projected, corrected images were saved as separate 16-bit TIFF files per channel with a consistent naming scheme (ch1.tif … ch5.tif) and organised into the images/ directory by plate, well, and site. The accompanying manifest.csv was generated from the original plate map and image metadata and contains, for every channel and site, the relative image path together with plate, well, site and perturbation annotations.

For the CellPaintSSL_IN_MOA_JUMP archive, the same z-projected, corrected images were assembled into 5×H×W TIFF stacks (one file per field of view), with channels ordered as required by our CellPaintSSL preprocessing pipeline. The dataset.csv file was constructed to mirror the format of the official CellPaintSSL repository, providing per-image metadata (plate, well, site, perturbation_id, MOA label, etc.) for direct use with the CellPaintSSL inference script.

MOAProfiler_IN_MOA dataset (preprocessed data for our own MOA plate): 10.5281/zenodo.17707972

CellPaintSSL_IN_MOA dataset (preprocessed data for our own MOA plate): 10.5281/zenodo.17703102