Automated image analysis and inference of gene function from high - Content screens

The study of tumor biology and heterogeneity is of high importance for identifying viable cancer therapeutics. A high content RNAi screen is carried out to identify genes that induce varied tumor morphologies. We present a novel automated pipeline to identify and interpret gene function by extracting morphological features of tumor cell aggregates, in large scale 3D RNAi screens. We use a “bag of words” based clustering approach to distinguish multiple phenotypes. Functional analysis of genes underlying the phenotypic clusters reveals the role of growth and invasion modulators in shaping tumor cell morphology and heterogeneity.