Enhanced detection of RNA modifications and read mapping with high-accuracy nanopore RNA basecalling models

The effect of custom basecalling models on biologically relevant m⁶A motifs and at different m⁶A stoichiometries. (A) Comparison of the delta summed errors (ΔSumErr = SumErr_MOD − SumErr_UNM) at the central modified position for 5-mers that fall within the DRACH motif (D = A, G, or U; R = G or A; H = A, C, or U) and others (non-DRACH). Statistical analysis was performed using a two-sided nonparametric Wilcoxon test with “default” as the reference group. (B) Comparison of the ΔSumErr for individual motifs that are DRACH and non-DRACH. Each dot represents a single observation of that 5-mer within the “curlcake” sequences. To compare the relationship between individual 5-mers, a paired two-sided Wilcoxon test was performed. (C) Schematic representation of the experimental approach to obtain “curlcakes” at different m⁶A stoichiometries. ATP and m⁶ATP were mixed at different ratios and used for the in vitro transcription of “curlcakes” with the remaining NTPs, followed by poly(A)-tailing, library preparation, and direct RNA sequencing (DRS). (D) Comparison of error-signature amplitude at different modification stoichiometries between 5-mers that fall within the DRACH motif. The corresponding plot for non-DRACH motifs is reported in Supplemental Figure S5B. Lines were fitted using locally estimated scatterplot smoothing (loess), and lightly shaded areas correspond to the 95% confidence interval. For A,B the box is limited by the lower-quartile Q1 (bottom) and upper-quartile Q3 (top). Whiskers are defined as 1.5 × IQR with outliers represented as individual dots. All statistical tests were corrected for multiple hypothesis testing using the Benjamini–Hochberg correction, and adjusted P-values are displayed: (ns) P > 0.05, (*) P ≤ 0.05, (**) P ≤ 0.01.