Fig. 1

MethCORR inferred RNA expression in ten cancer types. a Overview of the MethCORR method. (1) Each TCGA cohort with matched RNA expression and DNA methylation data is independently used for the MethCORR method. (2) The expression of each RNA is correlated to the methylation level of each CpG site across all discovery samples. (3) The ≤ 100 most positive and ≤ 100 most negative expression-correlated CpG sites specific for each RNA constitute the MethCORR matrix. (4) The methylation level of the RNA expression-correlated CpG sites from the MethCORR matrix is used to calculate inferred RNA expression for each gene in fresh-frozen and FFPE samples. The iRNA expression profiles can be used for transcriptional-like analysis. b Scatterplots with intra-sample correlations between observed RNA expression and iRNA expression for a representative sample (median R²) from the TCGA BRCA, PRAD, and LUAD validation samples. c Scatterplot with RNA expression-iRNA expression squared correlations (R²) for all validation samples for each of the ten TCGA cohorts. d Left: Table with squared correlation (R²) and root mean square error (RMSE) for correlations between observed RNA expression in fresh-frozen tissue and iRNA expression calculated in matched FFPE tissue or observed RNA expression in matched FFPE tissue. Correlations are shown for all validation samples with matched tissue for the BRCA (n = 3), PRAD (n = 3), LUAD (n = 9), BLCA (n = 3), KIRC (n = 3), and UCEC (n = 4) cohorts. Right: Scatterplots with correlations between observed RNA expression in fresh-frozen tissue and iRNA expression calculated in matched FFPE tissue for a representative independent validation sample from the TCGA BRCA, PRAD, and LUAD cohorts. e Scatterplot with the first- (PC1; X-axis) and second principal component (PC2; Y-axis) from a PCA performed with RNA expression for 25 fresh-frozen cancer samples and matched FFPE RNA sequencing data (top) or calculated iRNA expression (bottom). The analysis was performed with common MethCORR genes for the 25 cancer samples (six cancers; n = 2374 common MethCORR genes)