In silico analysis

The list of drug candidates was obtained by querying the Connectivity Map database (CMap build 02) with both LSC-R (Eppert et al.¹⁹) and an additional signature (LSC-Ng) derived from 78 additional AML (Ng et al.²⁰) as well as HSC-R (Eppert et al.¹⁹). The LSC-R up and down probes sets are the 100 probes most significantly correlated with LSCs and 100 probes most correlated with non-LSCs from Eppert et al.¹⁹. The LSC-Ng probes sets are derived from an early build of the LSC expression data from Ng et al.²⁰. The top and bottom 250 genes associated with LSCs (determined by t-statistic) were converted to Affymetrix probe IDs (Supplementary Table 1). The HSC-R probes lists are derived from the list of probes associated with HSCs (top and bottom) with a p value of ≤0.05. The molecules displaying a negative mean enrichment score (ES) with a p value of ≤0.1 for the LSC signatures and that were not associated with a negative ES in HSC-R were considered for in vitro screening.