Extended Data Fig. 3: mRNA-protein abundance correlation suggests potential prognostic biomarkers.
a, Distribution of Spearman’s correlation coefficients of mRNA and protein abundances of individual genes across patients (n = 150 patients). Yellow and blue, positive and negative correlations, respectively. b, Differential association of the genes having high and low mRNA-protein abundance correlations with KEGG pathways (n = 150 patients). The genes involved in each KEGG pathway were indicated by yellow (positive correlation) and blue (negative correlation) bars. c, Cumulative density distributions of mRNA-protein correlations for the genes with significant (blue) and non-significant (red) survival differences between top and bottom 25% of patients with the highest and the lowest mRNA expression levels, respectively. p = 1.4 × 10−18 by two-sided Kolmogorov-Smirnov test. d, Distributions of survival differences (Chi-square statistic values) for the genes with significant (FDR < 0.01) and non-significant (FDR > 0.1) mRNA–protein correlations. p = 3.2 × 10−18 by two-sided Student’s t-test. In violin plots, the line indicates the median value. n = 150 and 196 patients for mRNA-protein correlation and survival analysis, respectively (c-d). e, Cellular pathways represented by the genes showing negative mRNA-survival correlations. The enrichment significance of each pathway is indicated as –log10 (p-value), where p-value is the p-value for enrichment from ConsensusPathDB. Two dotted lines for p = 0.05 and 0.01. f, Selection of prognostic biomarkers for functional experiments. Among 19 prognostic biomarkers with significant negative (hazard ratio>1) mRNA-survival correlations in the four PDAC cohorts, the candidates with desirable survival curve patterns (red and blue lines, top and bottom 25% of patients with the highest and the lowest mRNA expression levels) in at least 3 PDAC cohorts were selected, resulting in 16 prognostic biomarkers. Among them, 7 prognostic biomarkers that have not been previously reported in pancreatic cancer were selected. Finally, among them, the 5 prognostic biomarkers having expression levels larger than the median expression (log2-FPKM = 3.11) of expressed genes (FPKM > 1) in AsPC1 cells. Among them, KRT19 was excluded due to its involvement in the broad spectrum of functions. The mRNA expression profile of AsPC1 cells was obtained from Cancer Cell Line Encyclopedia (CCLE)65. FPKM, fragments per kilobase of transcript per million.
