Scalable, compressed phenotypic screening using pooled perturbations

High-throughput phenotypic screens using biochemical perturbationsand high-content readouts are constrained by limitations of scale. Toaddress this, we establish a method of pooling exogenous perturbationsfollowed by computational deconvolution to reduce required sample size,labor and cost. We demonstrate the increased efficiency of compressedexperimental designs compared to conventional approaches throughbenchmarking with a bioactive small-molecule library and a high-contentimaging readout. We then apply compressed screening in two biologicaldiscovery campaigns. In the first, we use early-passage pancreatic cancerorganoids to map transcriptional responses to a library of recombinanttumor microenvironment protein ligands, uncovering reproduciblephenotypic shifts induced by specific ligands distinct from canonicalreference signatures and correlated with clinical outcome.

In the second, weidentify the pleotropic modulatory effects of a chemical compound librarywith known mechanisms of action on primary human peripheral bloodmononuclear cell immune responses. In sum, our approach empowersphenotypic screens with information-rich readouts to advance drugdiscovery efforts and basic biological inquiry.