Creating the world’s largest public Cell Painting gene/compound image resource via the JUMP-Consortium

Abstract: Srinivas Niranj Chandrasekaran & Anne E. Carpenter

The JUMP-CP Consortium (Joint Undertaking for Morphological Profiling-Cell Painting) is creating a public Cell Painting dataset of over 140,000 genetic and chemical perturbations and establishing experimental and computational best practices for image-based profiling. In Cell Painting, we stain many cellular components and extract thousands of morphological features from each cell’s image. We then harvest similarities in these “profiles” to identify, at a single-cell level, how diseases, drugs, and genes affect cells, which can uncover small molecules’ mechanism of action, discover disease-associated phenotypes, identify the functional impact of disease-associated alleles, and identify novel therapeutics, among many other applications.

Our pilot experiments have improved the experimental and computational workflow; improvements and suggested controls are now provided publicly at our project site (https://jump-cellpainting.broadinstitute.org/results). The full dataset will be public November 2022. We have in the meantime publicly released a pilot dataset called CPJUMP1, which was carefully designed to have 160 genes, each targeted by two CRISPR guides, one overexpression vector and two compounds that are annotated as targeting that gene. Each perturbation was screened in replicates and under different experimental conditions, such as cell type and time point. We used CPJUMP1 to optimize experimental conditions and are continuing to use it to optimize data analysis, including machine learning strategies. Apart from learning effective representations of cell states, it can also be used for training gene-compound matching methods. Similarly, the different experimental conditions enable the development of style-transfer methods.