SMRTnet, a deep‑learning framework integrating language models with graph attention networks, predicts small‑molecule–RNA interactions using RNA secondary structure rather than tertiary models. Published with experimental follow‑up, SMRTnet identified nanomolar‑to‑micromolar hits against ten disease‑linked RNA targets and validated a hit that downregulated MYC and inhibited cancer cell growth. The authors benchmarked SMRTnet against existing tools and reported substantially higher hit rates in retrospective and prospective screens. By removing the need for solved RNA tertiary structures, the model broadens the scope of RNA targets accessible to small‑molecule drug discovery and could accelerate lead identification for RNA‑modulating therapeutics.