A quantitative framework for structural interpretation of DMS reactivity

Abstract

Dimethyl sulfate (DMS) chemical mapping is widely used for probing RNA structure, with low reactivity interpreted as Watson-Crick (WC) base pairs and high reactivity as unpaired nucleotides. Despite its widespread use, a quantitative understanding of how DMS reactivity relates to specific RNA 3D structural features remains incomplete. To address this gap, we systematically analyzed DMS reactivity patterns with a massive library of 7,500 RNA constructs containing two-way junctions with known 3D structures. Our results reveal that DMS reactivity exists on a continuous spectrum rather than discrete high and low bins. Approximately 10% overlap in reactivity between WC and non-WC nucleotides demonstrates that simple thresholds cannot accurately determine base-pairing status. In flanking WC pairs, DMS reactivity correlates with base stacking strength and junction dynamics. For non-WC nucleotides, increased hydrogen bonding and decreased solvent accessibility led to WC-like DMS protection. Most significantly, we discover that DMS reactivity in non-canonical pairs strongly correlates with atomic distances and base pair geometry, enabling discrimination between different 3D conformations. These quantitative relationships establish novel metrics for evaluating RNA structural models and provide a new framework for incorporating DMS reactivity patterns into structure prediction algorithms.