Fig. 1
Identification of candidate m6A target miRNAs in serum. A The workflow of the establishment of serum m6A-miRNAs signature for cancer detection as well as the validation process. B Functional annotation for the included m6A target miRNAs using GO enrichment analysis. All the biological processes selected were statistically significant. The color depth of column represented P value and length represented enriched gene counts. C According to the criteria of FDR < 0.05 and |fold changes| > 1.23, the volcano plot showing the 18 candidate m6A target miRNAs, which were identified by LASSO, presented a higher expression level in cancer samples compared with non-cancer serum controls. Red dot, up-regulated miRNAs; Gray dot, not significant miRNAs. D The heatmap plotted for the expression of 18 candidate miRNAs using unsupervised hierarchical clustering in both cancer and non-cancer control groups. Yellow represented up-regulation and blue represented down-regulation. The cancer types were utilized as sample annotations. E Principal component analysis (PCA) for the 18 candidate m6A target miRNAs in cancer and non-cancer control. Two independent clusters were identified, suggesting the 18 miRNAs could well distinguished cancer samples from non-cancer controls. Red dot, cancer sample; Blue dot, non-cancer control sample. F-G ROC curve showing the performance of each candidate miRNA individually detecting cancer patients in the training cohort. H The performance of each candidate miRNA individually detecting cancer patients was validated in the internal validation cohort. The radar chart summarized the area under curve (AUC) and the AUC ranged from 0.667 to 0.94