Fig. 2
Primary tumor sample/cell line correlations driven by tumor purity. a Correlations between cell lines and high purity primary tumor samples (red) are significantly higher than correlations between cell lines and low purity primary tumor samples (turquoise) in 15/20 tumor types using the one-sided Wilcoxon test. P-values are indicated by symbols above the violin plots with ns corresponding to p-value > 0.05, one star corresponding to p-value < = 0.05, two stars corresponding to p-value < = 0.01, three stars corresponding to p-value < = 0.001, and four stars corresponding to p-value < = 0.0001. The median correlation coefficients are depicted by the horizontal black lines in the violin plots. b STRING analysis of protein–protein interactions for the 95 genes upregulated in primary tumor samples in all 20 of the analyzed tumor types (PPI enrichment p-value < 1.0e-16). Line thickness denotes confidence of the interaction, and only high confidence interactions are shown. The PPI network is enriched for immune response pathway genes (false discovery rate = 5.51e-06). c Gene Set Enrichment Analysis (GSEA) between primary tumor samples and cell lines using hallmark gene sets from MSigDB. NES are shown for pathways with FDR < 5%. Blue boxes indicate enrichment in cell lines, and red boxes indicate enrichment in primary tumor samples. Gene sets related to cell-cycle progression are enriched in cell lines across tumor types, and immune pathways are enriched in primary tumors. d GSEA of hallmarks of cancer pathways. Genome instability is enriched in cell lines across all tumor types, and tumor-promoting inflammation is enriched in primary tumors
