Fig. 2: Performance evaluation of ten tools for m6A detection capability.
From: Systematic comparison of tools used for m6A mapping from nanopore direct RNA sequencing
a Receiver operating characteristic (ROC) curve (top) and Precision Recall (PR) curve (bottom) for candidate sites detected by single-mode tools using the miCLIP data as ground truth. Area under the curve (AUC) is indicated in parentheses. b ROC AUC (top) and PR AUC (bottom) for candidate sites detected by single-mode tools using different NGS-based datasets as ground truth. c ROC curve (top) and PR curve (bottom) for candidate sites detected by compare-mode tools using the miCLIP data as ground truth. AUC is indicated in parentheses. d ROC AUC (top) and PR AUC (bottom) for candidate sites detected by compare-mode tools using different NGS-based datasets as ground truth. e Precision curve for top 20,000 m6A sites detected by single-mode tools using the miCLIP data as ground truth. f Percentage for m6A sites supported by various number of NGS-based methods. Top 2000 m6A sites detected by single-mode tools are counted. g Precision curve for top 20,000 m6A sites detected by compare-mode tools using the miCLIP data as ground truth. h Percentage for m6A sites supported by various number of NGS-based methods. Top 2000 m6A sites detected by compare-mode tools are counted. i Metagene plots (left) of the transcriptome-wide m6A distribution for top 2000 m6A sites detected by single-mode tools. The sequence motifs of these sites plotted by Seqlogo are shown on the right. j Metagene plots (left) of the transcriptome-wide m6A distribution for top 2000 m6A sites detected by compare-mode tools. The sequence motifs of these sites plotted by Seqlogo are shown on the right.
