Changes in cellular gene expression in response to small-molecule or genetic perturbations have yielded signatures that can connect unknown mechanisms of action (MoA) to ones previously established. We hypothesized that differential basal gene expression could be correlated with patterns of small-molecule sensitivity across many cell lines to illuminate the actions of compounds whose MoA are unknown. To test this idea, we correlated the sensitivity patterns of 481 compounds with ∼19,000 basal transcript levels across 823 different human cancer cell lines and identified selective outlier transcripts. This process yielded many novel mechanistic insights, including the identification of activation mechanisms, cellular transporters and direct protein targets. We found that ML239, originally identified in a phenotypic screen for selective cytotoxicity in breast cancer stem-like cells, most likely acts through activation of fatty acid desaturase 2 (FADS2). These data and analytical tools are available to the research community through the Cancer Therapeutics Response Portal.
Access optionsAccess options
Subscribe to Journal
Get full journal access for 1 year
only $14.58 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
This work was supported by the US National Cancer Institute's Cancer Target Discovery and Development (CTD2) Network (grant number U01CA176152, awarded to S.L.S.). We acknowledge the following colleagues and centers for contributing compounds and for valuable critique: Boston University, J. Bradner, P. Brown, C. Chen, J. Clardy, CNIO (Spanish National Cancer Research Center), E.J. Corey, the Drug Synthesis and Chemistry Branch (Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis, National Cancer Institute), Eutropics, J. Gutterman, E. Holson, Karyopharm, M. Meyerson, A. Myers, J. Porco, J. Qi, Sanford-Burnham, M. Serrano-Wu, M. Shair, B. Stockwell, L. Walensky, X. Wang and D. Zaharevitz. We thank A. Deik for conducting lipid profiling measurements; S. Chattopadhay, J. Law, G. Schaefer, M. Stewart, V. Viswanathan and other members of the Schreiber laboratory for advice and helpful discussions; S. Wang for helping curate the Informer Set; K. Emmith, J. Aseidu and the CSofT informatics group for development and support of cell-line and data-tracking software; A. Vrcic and the Broad Compound Management team for handling the Informer Set; J. Boehm, A. Tsherniak, A. Aguirre and the Broad Cancer Program for training and advice; the Broad Biological Samples Platform for providing CCLs; and L. Garraway and the Broad-Novartis Cancer Cell Line Encyclopedia (CCLE) team. S.L.S. is an Investigator of the Howard Hughes Medical Institute.
The cancer cell-line panel. Description of the cancer cell lines (CCLs) profiled in this experiment, including annotations from the CCLE1 for lineage and histology and experimental growth conditions (including media composition).
The small-molecule informer set. Description of the small-molecule informer set profiled in this experiment, including concentrations used, protein target or activity, source and vendor information, and structure.
Area-under-curve values for 481 small molecules across 823 unique CCLs.
All transcripts significantly correlated with small-molecule sensitivity across all CCLs, non-HL CCLs, and HL CCLs. A Bonferroni-corrected significance cutoff of |z| > 5.83, representing a two-tailed distribution with family-wise error rate α < 0.025 in each tail, was used.
Significant lineage-specific correlation values. All significant values where the absolute value of the z-scored Pearson correlation was greater than the value across all CCLs and non-HL CCLs are included.
Correlations for 660 annotated compound-target pairs across all CCLs, non-HL CCLs, and HL CCLs. For compound-target pairs, the transcript rank (of 18,543 transcripts), z-scored Pearson correlation, and significance (|z| > 3.96; Bonferroni-corrected, two-tailed distribution with family-wise error-rate α < 0.025 in each tail) are included.
Expression-sensitivity correlations for TNFRSF12A with 481 small molecules across all CCLs. Included are the number of CCLs tested per small molecule; TNFRSFS12A transcript rank (see Supplementary Fig. 1b); the Pearson expression-sensitivity correlation; the z-scored Pearson correlation; minimum, median, mean, maximum, and standard deviation of z-scored correlations from permutation testing (65,536 permutations); results of Kolmogorov-Smirnov test for normality of permutation results; empirical P-value from permutation testing; and estimated P-value from the normal cumulative distribution function.
Transcript loadings from principal component analyses (PCA). Results are from PC_A1 (first PC from PCA of the entire 18,543 transcript × 481 small molecule correlation matrix across all CCLs), PC_B1 (first PC from PCA of the correlation matrix excluding HL CCLs), and PC_B2 (second PC from PCA of the correlation matrix excluding HL CCLs). The number of small molecules to which each gene was significantly correlated (excluding HL CCLs) was calculated using a Bonferroni-corrected, two-tailed distribution with family-wise error-rate α < 0.025 in each tail (|z| > 5.83).
Small-molecule scores from principal component analyses.
GSEA results with PC_B2 gene loadings as the input using the C2 and C5 gene sets. See www.broadinstitute.org/gsea/doc/GSEAUserGuideTEXT.htm#_Viewing_Analysis_results for analysis description.
Correlations across all 481 small molecules for 654 individual transcripts most correlated with response to at least one small molecule. Included are the number of CCLs tested per small molecule; transcript rank; the Pearson expression-sensitivity correlation; the z-scored Pearson correlation; minimum, median, mean, maximum, and standard deviation of z-scored correlations from permutation testing (≥16,384 permutations); results of Kolmogorov-Smirnov test for normality of permutation results; empirical P-value from permutation testing; and estimated P -value from the normal cumulative distribution function.
GSEA results with austocystin D expression-sensitivity correlations (excluding HL CCLs) as the input using the C2 and C5 gene sets. See www.broadinstitute.org/gsea/doc/GSEAUserGuideTEXT.htm#_Viewing_Analysis_results for analysis description.
Raw results from profiling cellular lipids in NCIH661 cells. Results are from 24-hour treatment with 2 μM ML239, 2 μM SC-26196, co-treatment, or DMSO, with individual replicates numbered across columns. Metabolite IDs match those from The Human Metabolome Database (www.hmdb.ca).
Mapping of accession numbers for the file CCLE_Expression_Entrez_2012-10-18.res to the gene symbols used in this analysis.