We began with a collection of 16,527 unannotated small molecules for which we had generated L1000 profiles. These compounds were derived from a variety of sources, including the Broad Institute's diversity oriented synthesis (DOS) library and the NIH's Molecular Libraries Probe Production Centers Network (MLPCN). We focused on the 2,418 compounds whose 75th quantile of TAS was as least 0.2 and whose signatures had a median pairwise WTCS of at least 0.3 across cell lines, indicating a robust transcriptional response in at least a subset of cell lines. We termed these compounds Discovery, and attempted to assign functional annotations via comparison with annotated drugs and genes in the L1000 Touchstone (reference) part of the data.
