Cell Painting analysis

In recent years, morphological profiling using the Cell Painting assay has emerged as a powerful tool in early drug discovery research. By leveraging high-content imaging with a multi-stain approach, this assay simultaneously captures morphological changes across various cellular compartments, enabling the rapid identification of compound effects. We recently contributed to a comprehensive morphological profiling dataset generated using the EU-OPENSCREEN Bioactive Compound Set, a carefully curated and well-annotated collection of compounds targeting diverse biological processes [20]. Our dataset was collected across four imaging sites in Europe, primarily using the Hep G2 cell line, with one site also generating data using the U-2 OS cell line. We employed an extensive evaluation and assay optimization process to ensure high data quality across different imaging sites. Additionally, we established an analysis protocol that validated the robustness and reproducibility of the data, as well as facilitated comparisons between the two cell lines.

To maximize the utility of Cell Painting datasets for advanced drug development, we recently introduced a semi-supervised contrastive (SemiSupCon) deep learning approach [14]. This method integrates biological annotations in supervised contrastive learning with the advantages of leveraging large unannotated image datasets through self-supervised contrastive learning (Figure 5). SemiSupCon enhances downstream prediction performance for classifying MeSH pharmacological classifications from PubChem, as well as mode of action and biological target annotations from the Drug Repurposing Hub, across two publicly available Cell Painting datasets. Notably, our approach effectively predicted the biological activities of several unannotated compounds, and these findings could be validated through literature searches. These findings highlight the potential of our approach to accelerate the exploration of biological activity based on Cell Painting image data with minimal human intervention.

[20] Wolff, C.; Neuenschwander, M.; Beese, C. J.; Sitani, D.; Ramos, M. C.; Srovnalova, A.; Varela, M. J.; Polishchuk, P.; Skopelitou, K. E.; Škuta, C.; Stechmann, B.; Brea, J.; Clausen, M. H.; Dzubak, P.; Fernández-Godino, R.; Genilloud, O.; Hajduch, M.; Loza, M. I.; Lehmann, M.; von Kries, J. P.*; Sun, H.*; Schmied, C.*, Morphological Profiling Dataset of EU- OPENSCREEN Bioactive Compounds Over Multiple Imaging Sites and Cell Lines, iScience, 2025, accepted. bioRxiv 2024.08.27.609964. doi: 10.1101/2024.08.27.609964

[21] Bushiri Pwesombo, D.; Beese, C.; Schmied, C.; Sun, H.*, Semisupervised Contrastive Learning for Bioactivity Prediction Using Cell Painting Image Data. J. Chem. Inf. Model., 2025, 65(2), 528-543. doi: 10.1021/acs.jcim.4c00835