Item

Abstract

High-content screens (HCS) using chemical and genomic interference based on light microscopy and quantitative image analysis yielded a large amount of multi-parametric (MP) phenotypic data. Such data-sets hold great promise for the understanding of cellular mechanisms by systems biology. However, extracting functional information from data-sets, such as links between cellular processes and the functions of unknown genes, remains challenging. The limitation of HCS analysis lies in the complexity of cellular organization. Here, we assumed that cellular processes have a modular structure, and deconvolved the MP data into separate signals from different cellular modules by Blind Source Separation. We applied a combination of quantitative MP image analysis (QMPIA) and Independent Component Analysis (ICA) to an image-based HCS of endocytosis, the process whereby cells uptake molecules from the outside and distribute them to different sub-cellular organelles. We named our approach Independent Phenotypes Analysis (IPA). Phenotypic traits revealed by IPA are interpretable in terms of perturbation of specific endosomal populations (e.g. specific cargo, specific molecular markers) and of specific functional modules (early stages of endocytosis, recycling, cell cycle, etc.). The profile of perturbation of each gene in such basic phenotypic coordinates intrinsically suggest its possible mode of action.