Unsupervised image-based profiling identifies a functional cluster containing known components of the dynein machinery and novel factors. (A) Workflow for phenotypic profiling using images collected from the secondary screen. (B) UMAP plot for phenotypes of genes selected from the primary screen. The highlighted clusters of genes were manually curated and annotated based on data in the UNIPROT database and primary literature. See Table S7 for source data. (C) Phenotypic feature heatmap of the dynein–dynactin gene cluster. Features are grouped in the x-axis according to the marker (see Data S1 and Table S8 for names of individual features). Genes encoding dynein and dynactin components, as well as the associated proteins BICD2 and LIS1, are labeled in different shades of blue. FHF component genes are shown in magenta. Novel genes are labeled in gray. The scales of rZ values (central reference = NTC) were adjusted based on minimum and maximum values of individual features. “Cytoplasm” refers to features associated with the background Hoechst staining in the cytoplasm.