Fig. 4

Cell type deconvolution of RNA-seq and DNA methylation data (a) heatmap presenting scores from two independent methods—MethylResolver (utilizing EPIC DNA methylation data, upper panel) and MCPcounter (utilizing RNA-seq data, lower panel) both point to higher immune infiltration in chordoma I cluster, significant differences in U-Mann–Whitney test are marked with * (b) Correlation matrix of immune signatures from both methods (Kendall correlation); coefficients are shown in the middle of each cell, significant ones are black (c) Plot presenting signature correlation for cytotoxic T-lymphocytes from both methods, correlation coefficient: 0.73, adjusted p value: 2.8e−142, cluster differences adjusted p values: 0.0012 and 0.0003 for MethylResolver and MCPcounter, respectively (d) Representative examples of immunohistochemical staining of selected chordoma C and chordoma I samples with antibodies against CD3, CD4 and CD8. (e) Comparison of chordomas (including chordoma subtypes classified according to DNA methylation profile) with existing signatures for variable human cancer types from ESTIMATE method (signatures for immune and stromal components of the tumor)