Direct profiling of non-adenosines in poly(A) tails of endogenous and therapeutic mRNAs with Ninetails

The advent of single-molecule sequencing technologies facilitated the analysis of thehomopolymeric tracts of nucleic acids. Direct RNA sequencing (DRS), which avoids amplification bias, holds great potential for the study ofpoly(A) tails of mRNAs. Although non-adenosines within poly(A) tails affect the stability and translation of mRNA, there is no computational method to detect composite tails in DRS data. To address this gap, we introduce the Ninetails R package, a neural network-based tool that identifies and quantifies non-adenosines in poly(A) tails with high accuracy. Examination of different biological contexts revealed widespread non-adenosine decorations, with insertion frequencies differing according to mRNA class, cell type, and species. Among many other factors, the frequency is influenced by the origin of poly(A) tails, with TENT5-polymerases substrates and mitochondrially encoded mRNAs enriched in composite tails. Notably, we also show that the composition of poly(A) tails in mRNA vaccines is dynamic and that the manufacturing protocol of synthetic mRNAs affects the purity of poly(A) tails.