Targeting Intracellular Organization via Microscopy-Based High-Content Phenotypic Screening and Generative Neural Networks

High-content image-based screening is a powerful technique for identifying phenotypic differences in cell populations with several applications including drug screening. While current computational approaches pool image-based features from different modalities, each of a distinct organelle, we develop a new methodology for assessing spatial dependencies between different organelles and apply it to identify new treatments that affect specific spatial dependencies between organelles. In many diseases, a disruption in the cell`s organization, determined by its organelles composition and the organelle-organelle relationship, leads to impaired cell function. Thus, the discovery of drugs that revert the cell structure and organization to its “healthy” state is imperative for some drug discovery pipelines. The methodology is based on measuring the reconstruction error of generative neural networks that map the different modalities to one another. Our results indicate that this approach is feasible, and that spatial organelle dependencies are more sensitive, specific, interpretable, and reproducible readouts for phenotypic cell screening. Specifically, we show that (1) phenotypes are amplified, making it easier to identify subtle phenotypes, (2) new phenotypes that are missed by traditional analyses can be discovered, and (3) spatial dependencies are differentially determined based on organelles, implying a more specific and interpretable readout. Overall, our methodology is the first to target defects in inter-organelle spatial dependencies and will allow the discovery of the effects each treatment has on specific aspects of cell organization by "breaking" existing relationships between multiple cell structures currently unavailable to researchers.