The Carcinogenome Project

We generated a large toxicogenomics resource comprising ∼11,000 expression profiles corresponding to ~500 chemicals, each annotated with carcinogenicity and genotoxicity information, and profiled in HepG2 (human hepatocellular carcinoma cell line) and MCF10A cells at multiple doses and replicates. The Platform is GPL20573: Broad Institute Human L1000 epsilon https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GPL20573 For questions or assistance with this dataset, please email the CMap support team at: [e-mail] We prioritized chemicals with long-term rodent liver and breast carcinogenicity annotation (derived from the Carcinogenic Potency Database (CPDB)) for inclusion in this experiment. Additional chemicals without carcinogenicity annotation were included on the basis of interest to the Superfund Research Program (environmental toxicants), presence in controversial commercial products (included for predictive purposes), and evidence of binding to the aryl hydrocarbon receptor (AhR), as the AhR is an important mediator of xenobiotics, including carcinogens. Detailed cell culture, plating, treatment and lysis protocols are described in https://assets.clue.io/resources/sop-cell.pdf. Briefly, HepG2 (ATCC® HB-8,065™), MCF10A parental and TP53-/- (Horizon, cat# HD 101-005), were used with medium RPMI1640 (10040CV; Mediatech) supplemented with 10% v/v fetal bovine serum (F4135; Sigma-Aldrich), 1× penicillin-streptomycin-glutamine (Invitrogen 10378016), and incubated at humidified 5% CO2 atmosphere at 37°C. Cell cultures were plated with 4,000 cells (45μL of growth medium) per well on 384-well plates (3707; Corning) and incubated for 24 h before treatment. Cells were treated with 5μL of 1:100 diluted 1,000× stock compound plates to final volume of 50μL and incubated for 24 h before lysis. Each chemical perturbation for the HEPG2 screen was administered at six doses in triplicate wells per dose and chemical combination, starting from 40μM maximum dose (40 mM stock diluted 1:1,000) for NTP chemicals and 20μM for chemicals procured from Sigma-Aldrich, in series of two-fold dilutions. The sole exception to the standard dosage was 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), which had a starting dose of 50 nM due to its extreme potency. Each chemical perturbation for the MCF10A was administered at three doses (100, 33, 10μM) in triplicate wells. The vehicle control used was DMSO with final dose of 14.1 mM. Four positive controls were used (vorinostat, geldanamycin, mitoxantrone, withaferin-a) each in final doses of 10, 3.33, and 1.00μM, respectively. Following treatment and lysis, the gene expression of the HepG2 and MCF10A cells was profiled using the L1000 platform, a high-throughput assay that measures the expression of ∼1,000 landmark genes and computationally infers the expression of nonmeasured transcripts. For each perturbation and landmark gene, we computed the change in gene expression following the perturbation using a moderated z-score procedure as described in the CMap–L1000 workflow. Differential expression values were calculated as moderated z-scores for each landmark gene and each unique perturbation (chemical and dose combination) perturbation, collapsed to a single value across replicates (Subramanian et al. 2017). Each GEO sample (GSM) represents 1-3 replicates of the same treatment. Raw data files can be analyzed at: https://github.com/cmap/cmapJ where the Broad Institute of MIT and Harvard provided the Connectivity Map software for Java to read and store lxb and GCTX files.