A key challenge in designing phenotypic screens is the selection of optimal imaging biomarkers. Here, Kang et al. present a method for systematically identifying the ORACL — optimal reporter cell line for annotating compound libraries — the phenotypic profile of which most accurately classifies compounds into multiple, diverse drug classes. The authors use the ORACL to carry out a single-pass phenotypic screen of several small-molecule compound libraries and validate drug class predictions in secondary assays.