Extended Data Fig. 3: Case study of small molecule–RNA interactions using different computational methods.
From: Predicting small molecule–RNA interactions without RNA tertiary structures
a-f. Heatmap of structural similarity (measured by the Tanimoto similarity) between theophylline and its 30 decoys across different computational methods, including SMRTnet (a), RNAmigos2 (b), AutoDock Vina (c), NLDock (d), RLDOCK (e), and rDock (f). The red dashed box in (a) represents two groups clustered by ranking score. g-h. Chemical structure of theophylline and its six top-ranked decoys (g) and ten bottom-ranked decoys (h) predicted by SMRTnet. SMRTnet predicted binding scores and names of decoys are listed below. The red dashed box highlights the pyrimidine dione-like structure, where the triazole in decoy_13 serves as a chemical equivalent of pyrimidine.
