Deconvolution of drug screening data delineates drug sensitivity of stem-like cancer cells in Acute Myeloid Leukemia [RNA-Seq]

Ex-vivo drug sensitivity screening (DSS) only provides a readout on mixtures of cells, potentially occulting important information on clinically relevant cell subtypes. Here we developed a machine-learning framework to deconvolute bulk RNA expression matched with bulk drug sensitivity into cell subtype composition and cell subtype drug sensitivity. We first determined that our method could decipher the cellular composition of bulk samples more accurately than current state-of-the-art methods. We then optimized an algorithm capable of estimating cell subtype- and single-cell-specific drug sensitivity, which we evaluated by performing in-vitro drug studies Overall design: We generated in vitro pools containing various ratios of HL60, SUDHL4, K562, and THP1 human leukemic cells. We generated bulk RNA-seq for 6 mixtures of these cell lines, and 4 RNA-seq of the pure populations. To build the reference dataset, we peformed scRNA-seq on a mix of these four cell lines.