Fig. 3

Functional evaluation of the emCpG gene targets. A Pearson correlation between FOXA1 expression and tumour purity in BRCA patients. B Pearson correlation between SPI1 expression and tumour purity in KIRP patients. As a tumour purity variable, we used cumulative purity estimates from BRCA and KIRP patients, respectively, reported by Aran et al. [52]. The scatterplots compare the tumour purity (x-axis; cumulative purity estimate) and expression of the TFs (y-axis). The blue lines represent the fitted Pearson linear relationship with the grey zone representing the 95% confidence interval (Pearson R coefficients and associated p values are provided in the top-left corner). The expression of FOXA1 in breast cancer patients shows positive correlation indicating that the signals observed in the GO term and pathway enrichments are coming from the tumour cells themselves. The expression of SPI1 in kidney renal papillary cell carcinoma patients shows negative correlation indicating that the signals observed in the GO term and pathway enrichments may be coming from the tumour microenvironment. C Functional enrichment analysis for genes linked to emCpGs associated with cancer cell emTFs (i.e. emTFs whose expression positively correlate with tumour purity as in A.). D Functional enrichment analysis for genes linked to emCpGs associated with immune cell emTFs (i.e. emTFs whose expression negatively correlates with tumour purity as in B.). Functional enrichments in C–D were performed using the Hallmark sets from MSigDB [49]