De novo basecalling of m6A modifications at single molecule and single nucleotide resolution

Abstract

RNA modifications hold pivotal roles in shaping the fate and function of RNA molecules. Although nanopore sequencing technologies have proven successful at transcriptome-wide detection of RNA modifications, current algorithms are limited to predicting modifications at a per-site level rather than within individual RNA molecules. Herein, we introduce m6ABasecaller, an innovative method enabling direct basecalling of m6A modifications from raw nanopore signals within individual RNA molecules. This approach facilitates de novo prediction of m6A modifications with precision down to the single nucleotide and single molecule levels, without the need of paired knockout or control conditions. Using the m6ABasecaller, we find that the median transcriptome-wide m6A modification stoichiometry is [~]10-15% in human, mouse and zebrafish. Furthermore, we show that m6A modifications affect polyA tail lengths, exhibit a propensity for co-occurrence within the same RNA molecules, and show relatively consistent stoichiometry levels across isoforms. We further validate the m6ABasecaller by treating mESC with increasing concentrations of STM2457, a METTL3 inhibitor as well as in inducible METTL3 knockout systems. Overall, this work demonstrates the feasibility de novo basecalling of m6A modifications, opening novel avenues for the application of nanopore sequencing to samples with limited RNA availability and for which control knockout conditions are unavailable, such as patient-derived samples.