Structure-aware graph learning predicts RNA editability across tissues and species

Programmable A-to-I RNA editing using endogenous ADAR enzymes is emerging as a therapeutic strategy, but editability remains difficult to predict because ADAR recognition depends on double-stranded RNA geometry and stability rather than sequence alone. We present ADAREDIT, a structure-explicit graph-attention framework that represents each dsRNA substrate as a nucleotide graph with backbone and base-pair edges and augments this representation with typed interactions and a motif-sensitive sequence branch. We trained and evaluated the model on high-confidence inverted Alu duplexes (n = 905) with secondary structures predicted by RNAfold and editing levels measured across 8,603 GTEx RNA-seq samples spanning 47 tissues. Across five tissue contexts and comprehensive cross-tissue transfer experiments, ADAREDIT consistently outperformed sequence-only CNN, transformer, and RNA language model baselines and achieved strong discrimination on combined tissue data (AUROC/AUPRC = 0.96; F1 ≈ 0.90). The same graph representation transferred to evolutionarily distant non-Alu species (sea urchin, acorn worm, and octopus), indicating conserved principles of ADAR substrate recognition. Finally, attention profiles and in silico mutagenesis recapitulated known biochemical constraints, including suppression by an upstream guanosine, and revealed longer-range asymmetric structural influences on editing. The sources of this work are available at our repository: https://github.com/Scientific-Computing-Lab/AdarEdit.