Cell Painting Compound Profiling (BBBC022)

Cell Painting images of human U2OS osteosarcoma cells treated with bioactive compounds, from the Broad Bioimage Benchmark Collection (BBBC022v1, Gustafsdottir et al. 2013). The original release ships every field of view as five separate single-channel 16-bit TIFFs; here each field is one five-channel OME-TIFF, so you can show, hide and window-level each organelle stain in the browser and filter fields by compound, concentration or well role.

This is the first complete plate (plate 20585) of the 20-plate pilot experiment. The other 19 plates share the same layout and will be added over time.

Contents

Images	3,456 fields (384 wells x 9 sites) from plate 20585
Annotations	none; each file carries compound, concentration, SMILES, well, site and well-role metadata
Classes	n/a (phenotypic profiling, not labelled objects)
Dataset	cell-painting-bbbc022-v1
Split	320 compound-treated wells (320 distinct compounds, one well each) and 64 mock (vehicle control) wells
Imaging	Five channels per field: Hoechst 33342 (nuclei), concanavalin A (endoplasmic reticulum), SYTO 14 (nucleoli and cytoplasmic RNA), WGA + phalloidin (Golgi, plasma membrane, F-actin), MitoTracker Deep Red (mitochondria). 696 x 520 px, 16-bit (12-bit range), 20x, 0.656 um/px

The compounds on this plate come from the Prestwick Chemical and Biomol International bioactive libraries; the concentration recorded in the source metadata (CPD_MMOL_CONC) is carried through unchanged. The full experiment covers 1,600 compounds across five plate maps with four replicate plates each; plate 20585 is the first replicate of plate map H-BIOA-002-1.

Using it

Open any field and toggle the five channels in the viewer. To download everything (free account required):

• ZIP: use the Download button on this project.
• Python SDK: pip install datatorch, then:

datatorch login
datatorch pull datatorch/cell-painting-bbbc022

Every file's metadata (visible in the file panel and in the SDK) includes broad_id, compound_name, compound_source, smiles, conc_mmol, well, site, well_role (compound or mock), plate_map, acquired_ms and the original source filename of each channel, so you can group fields by compound and compare them with the mock wells.

Conversion notes

Each OME-TIFF was built by reading the five original single-channel TIFFs of a field (696 x 520, uint16, uncompressed) and writing them as five uncompressed pages in the order Hoechst, ConA, SYTO14, WGA_Phalloidin, MitoTracker, with channel names, display colours and the 0.656 um pixel size in the OME-XML header. Pixel values are unchanged; every packaged field was verified pixel-identical to its source planes. File names are bbbc022-<plate>-<well>-s<site>.ome.tif.

Field identity (well and site) was decoded from the source filenames (IXMtest_<well>_s<site>_w<channel><GUID>.tif) and checked against the plate metadata CSV for all 3,456 fields. The official BBBC022_v1_image.csv has defects that were handled as follows:

• 45 rows (wells F22, H08, H24, L14, N06) have the Site and Time columns replaced by the literal values IXMtest and 5; site was taken from the filename instead.
• In wells F22 and H08 the compound-name field is an unterminated quote that swallows the site, time and source columns. The acquisition time was recovered from the swallowed text, the source name is missing, and the compound names (7-CHLORO-3-METHYL-3,4-DIHYDRO-2H-1,2,4-BENZOTHIADIAZINE-S and 2,3-DIOXO-6-NITRO-1,2,3) are truncated in the source and left as-is.
• A further 126 well-formed rows (14 wells) record an acquisition time of 5; acquired_ms is null for those 153 fields in total.
• Rows flagged this way carry source_csv_row_malformed: true in their metadata.

Well A01 site 1 was uploaded twice during an upload test; the copy whose name ends in a random suffix is a duplicate and will be removed.

License and citation

CC0 1.0 (public domain). Anne Carpenter waived all copyright and related rights to BBBC022v1; attribution is still good practice. If you use this data, please cite:

Gustafsdottir, S.M., Ljosa, V., Sokolnicki, K.L. et al. Multiplex cytological profiling assay to measure diverse cellular states. PLOS ONE 8(12): e80999 (2013). https://doi.org/10.1371/journal.pone.0080999

and the collection:

Ljosa, V., Sokolnicki, K.L. & Carpenter, A.E. Annotated high-throughput microscopy image sets for validation. Nature Methods 9, 637 (2012).

Source: BBBC022v1, Broad Bioimage Benchmark Collection.

Related

• Nuclei Segmentation (BBBC039): 200 Hoechst images from this experiment with hand-annotated nuclei
• More microscopy datasets on DataTorch
• All public datasets