Tree_RNA-Align: RNA Secondary Structure Clustering and Classification Based on Tree-Structure Alignment

Abstract

Clustering and classification of RNA secondary structures are central to understanding RNA function. However, widely used alignment methods, such as LocARNA and bpRNA-align, are not explicitly designed to exploit the hierarchical relationships among RNA structural elements, limiting their applicability to complex, multi-branched structures. In this study, we introduce Tree_RNA-Align, a novel method for RNA secondary structure clustering and classification based on a tree-structure alignment algorithm. The method transforms dot–bracket structures into tree representations, in which multibranch loops and stems serve as nodes, thereby preserving the hierarchical relationships among structural elements. It integrates a bottom–up hierarchical comparison for clustering with a top–down comparison for classification and prediction. Notably, classification experiments on five RNA families (16S rRNA, group_I_intron, RNase P, SRP, and tmRNA) achieved a micro-averaged F1-score of 0.933 (range across families: 0.760–0.981) and effectively identified representative structures within each family. Overall, these results suggest that incorporating classification can further improve RNA secondary structure prediction, demonstrating the utility of Tree_RNA-Align for structural analysis and functional annotation.