TY - JOUR T1 - Exploring genetic interaction manifolds constructed from rich phenotypes JF - bioRxiv DO - 10.1101/601096 SP - 601096 AU - Thomas M. Norman AU - Max A. Horlbeck AU - Joseph M. Replogle AU - Alex Y. Ge AU - Albert Xu AU - Marco Jost AU - Luke A. Gilbert AU - Jonathan S. Weissman Y1 - 2019/01/01 UR - http://biorxiv.org/content/early/2019/04/07/601096.abstract N2 - Synergistic interactions between gene functions drive cellular complexity. However, the combinatorial explosion of possible genetic interactions (GIs) has necessitated the use of scalar interaction readouts (e.g. growth) that conflate diverse outcomes. Here we present an analytical framework for interpreting manifolds constructed from high-dimensional interaction phenotypes. We applied this framework to rich phenotypes obtained by Perturb-seq (single-cell RNA-seq pooled CRISPR screens) profiling of strong GIs mined from a growth-based, gain-of-function GI map. Exploration of this manifold enabled ordering of regulatory pathways, principled classification of GIs (e.g. identifying true suppressors), and mechanistic elucidation of synthetic lethal interactions, including an unexpected synergy between CBL and CNN1 driving erythroid differentiation. Finally, we apply recommender system machine learning to predict interactions, facilitating exploration of vastly larger GI manifolds.One Sentence Summary Principles and mechanisms of genetic interactions are revealed by rich phenotyping using single-cell RNA sequencing. ER -