Share
About Event
Cell Painting is a high-content imaging technique for the phenotypic characterization of cellular responses to drug treatments and genetic perturbations. The method is based on multiplex staining of major cellular structures and organelles, including the nucleus, endoplasmic reticulum, nucleoli, Golgi apparatus, actin cytoskeleton, and mitochondria. Multiple quantitative features are extracted from the resulting images and subsequently analysed using approaches such as principal component analysis. Phenotypic profiles of novel compounds can then be compared with those of reference drugs with established mechanisms of action.
This approach has been successfully used to identify active compounds and cluster them according to related phenotypes or reference mechanisms of action. However, when novel compounds do not cluster with any of the reference drugs, their phenotypic profiles can be difficult to interpret. To increase the resolution of the method and gain deeper insight into drug-induced phenotypes, we applied deep-learning-based analysis to Cell Painting images. This approach may support both the clustering of compound-induced phenotypes and the identification of organelle-specific changes in staining patterns.
To this end, we developed and compared several deep-learning models for image segmentation, extraction of single-cell data, and discrimination between drug-induced phenotypes. The models were tested using a reference set of compounds targeting the stained cellular structures. Selected models successfully distinguished individual compounds and broader phenotypic groups. Further development of this approach may enable the identification of organelle-specific effects of novel compounds and provide insights into their possible mechanisms of action.
