Figure 4: Drug response consequences of genetic and transcriptomic variation.

(a) Top: unsupervised hierarchical clustering of 27 MCF7 strains, based on their response to the 55 active compounds in the primary screen. Colors, as in Fig. 1. Bottom: a corresponding heatmap, showing the % of viability change for each compound across strains. Compounds colored based on their mechanism-of-action. (b) Classification of the screened compounds based on their differential activity. Consistent, viability change <−50% for all strains; variable, viability change <−50% for some strains and >−20% for other strains; intermediate, viability change in between these values. (c) Comparison of the similarity in drug response patterns between compounds that share the same mechanism-of-action (n=39) and compounds that work through different mechanisms (n=1,439). One-tailed Wilcoxon rank-sum test. (d) Highly similar differential drug response patterns for three proteasome inhibitors: bortezomib, MG-132 and carfilzomib. Each data point represents the mean of two replicates. The number of data points per strain is mentioned within parentheses. The response pattern with no drug (DMSO control) is presented for comparison. (e) Schematics presenting the analysis performed to evaluate the association between drug response and transcriptional variation. (f) Up-regulation of the KEGG cell cycle signature in strains sensitive to the cell cycle inhibitor alsterpaullone (8 sensitive vs. 15 resistant strains). (g) Up-regulation of mTOR signaling in strains sensitive to the PI3K inhibitor BKM-120 (8 sensitive vs. 5 resistant strains). (h) Up-regulation of the genes that are up-regulated when PTEN is knocked-down in strains sensitive to AKT inhibitor IV (6 sensitive vs. 9 resistant strains). (i) Strains with PTEN mutation (n=12) respond more strongly to AKT inhibitor IV than strains without the mutation (n=14). (j) Strains with ESR1 copy number loss (n=5) grow better in estrogen-depleted medium than strains without ESR1 loss (n=21). (k) Comparison of GSEA-based MoA identification between the MCF7 cohort and the CTD² (n=15) and GDSC (n=19) cohorts, across matched drugs. Two-tailed Fisher’s exact test. For all box plots: bar, median; box, 25^th and 75^th percentiles; whiskers, 1.5*IQR of lower and upper quartile; circles: data points.