Note: We have developed an extremely accurate method capable of detecting m6A at the single-molecule level. For details, see https://www.biorxiv.org/content/10.1101/2023.11.16.567334v1.
basecalling and alignment
These ONT tools were executed with appropriate parameters to output results of all sites in RRACH motifs (coverage >= 5).
The results of each tool were organized into a standard format:
transcript pos score coverage id motif
ENSMUST00000000001.4 63 0.478300 27 ENSMUST00000000001.4|63 AGACC
The transcriptome coordinates were converted to the genomic coordinates. Sites derived from multiple transcriptomic coordinates but aligned to the same genomic coordinate were merged as following: for Tombo/Tombo_com, MINES, Nanom6A, m6Anet and Epinano/Epinano_delta, the scores (fraction-modified/probability-modified) were weighted averaging according coverage; for Nanocompore, Xpore, DiffErr, DRUMMER and ELIGOS, the best scores (p-value/z score) were selected as representative.
plot ROC and PR curves
The optimal cut-off of each tool for m6A detection was determined by varing cut-offs and calculated F1 scores.