Fig. 6: High m6A detection accuracy of DeepRM is validated by multiple orthogonal experimental methods.
From: Comprehensive discovery of m6A sites in the human transcriptome at single-molecule resolution
a A schematic illustration of the transcriptome-wide in vitro transcription (IVT) experiment to estimate the false positive rate. b The fraction of A sites identified as m6A sites in the HEK293T IVT transcriptome (left bars), that is the false positive rate, and the non-IVT whole transcriptome (right bars). c A Venn diagram of m6A sites detected by Dorado and DeepRM in the HEK293T whole transcriptome (left) and fraction of sites validated via GLORI (right). d SELECT result for sites detected by DeepRM (see “Methods”). Three sites (KDM4A A3515, NTN1 A5645, MKI67 A980) were identified by both DeepRM and Dorado, and the other three sites (KHDRBS1 A1533, KAT6A A777, MICU1 A1871) were identified by DeepRM but not Dorado. The mean and SEM of 2-ΔΔCt values of wild type (WT) and IVT samples are shown for each site, with the one-sided Student’s t test results (n = 3 for each sample). p-values: 2.75e-6 (KDM4A A3515), 8.30e-4 (NTN1 A5645), 4.41e-3 (MKI67 A980), 7.91e-5 (KHDRBS1 A1533), 2.06e-3 (KAT6A A777), and 2.20e-3 (MICU1 A1871). *: p < 0.05, **: p < 0.01, ***: p < 0.001, n.s.: not significant (p ≥ 0.05).
