Revealing therapeutic single-cell transcriptomic signatures using a simple classifier

Here we present Cell-Sign, a detection algorithm that can classify cells according to drug treatments at the single-cell level. The method allows the identification of individual cells that remain hidden in conventional dimensionality reduction approaches and reveals both the drug a cell was exposed to and the duration of exposure.

We show how our approach can be used to identify drug targets by comparing single cell Perturb-seq data with drug signatures from the Library of Integrated Network-based Cellular Signatures (LINCS). Cell-Sign can contribute to highly multiplexed single-cell drug discovery and the identification of novel drug targets.