Exon-intron boundary inhibits mA deposition, enabling mA distribution hallmark, longer mRNA half-life and flexible protein coding.

Regional bias of N-methyladenosine (mA) mRNA modification avoiding splice site region, calls for an open hypothesis whether exon-intron boundary could affect mA deposition. By deep learning modeling, we find that exon-intron boundary represses a proportion (12% to 34%) of mA deposition at adjacent exons (~100 nt to splice site). Experiments validate that mA signal increases once the host gene does not undergo pre-mRNA splicing to produce the same mRNA. Inhibited mA sites have higher mA enhancers and lower mA silencers locally and show high heterogeneity at different exons genome-widely, with only a small proportion (12% to 15%) of exons showing strong inhibition, enabling more stable mRNAs and flexible protein coding. mA is majorly responsible for why mRNAs with more exons be more stable. Exon junction complex (EJC) only partially contributes to this exon-intron boundary mA inhibition in some short internal exons, highlighting additional factors yet to be identified.