Fig. 4: Concordance among ESR1 mutant breast cancer cell models.

A Scatter plot showing Pearson correlation of average regulatory percentile between 12 ectopic overexpression models (OE) and 32 genome-edited (GE) ESR1 mutant models of all the genes. Genes that more pronouncedly regulated by each model were highlighted (|delta regulatory percentile| >100 between the two models). B Bar graph showing the significantly enriched Hallmark pathways in GE and OE-preferentially regulated genes in (I). C Receiver operating characteristic curve depicting the performance of GE and OE upregulated genes (average percentile > 60%) as signatures in distinguishing ESR1 mutant from WT clinical samples from four merged cohort of 313 samples. D Scatter plot showing Pearson correlation of average regulatory percentile between ESR1 mutant and estrogen treatment of all the genes. Genes are subgrouped into three parts: ligand-dependent genes (average percentile above 50% or below −50% in both conditions, red and blue) and de novo genes (average percentile above 50% in TamR and within −15 to 15% in E2 treatment, green). E Box plot representing the BETA score comparison of each selected gene subgroups in (D). BETA score were calculated based on gained peaks of ESR1 mutant cells merged from 16 ChIP-seq samples normalized to their corresponding controls. Mann–Whitney U test was used. F Scatter plot showing the correlation of −log10 p values of LISA predicted regulators from ligand-dependent and de novo genes in (A). Only significantly enriched regulators were shown, and top targets skewed to each side were labeled. G Scatter plot showing correlation of average regulatory percentile between ESR1 mutant and estrogen treatment of 127 epigenetic modifiers. Targets consistently altered in both conditions or uniquely altered in one of the conditions were labeled. H Line plots representing the intensity of ATAC-seq signals from MCF7 WT and ESR1 mutant cell lines on different histone modification regions from ChIP-seq data of MCF7 cell line. I Scatter plot showing Pearson correlation of average regulatory percentile between Y537S and D538G ESR1 mutant variants of all the genes. Genes that more pronouncedly regulated by Y537S and D538G were highlighted (|delta regulatory percentile| >70 between the two variants). J Scatter plot showing the correlation of −log10 p values of LISA predicted regulators from Y537S and D538G-preferrentially regulated genes in (I). Only significantly enriched regulators were shown and top targets skewed to each side were labeled. K Box plots representing the BETA score comparison of each selected gene subgroups in (I). BETA score were calculated based on gained peaks of ESR1 mutant cells merged from 9 Y537S and 6 D538G ChIP-seq samples normalized to their corresponding controls. Mann–Whitney U test was used.