Abstract
Recently, the number of RNA tertiary (3D) structures deposited in the PDB database continues to grow. Since structures of molecules evolve more slowly than their sequences, the structural comparison of RNAs can bring more sig-nificant insights into their functions and evolutionary relationships that would not be detected by analyzing sequence information alone. In 2010, our labor-atory developed a tool, called iPARTS, that aligns two RNA 3D structures and determines their structural similarities. The basic steps of our iPARTS are as follows. First, a Ramachandran-like diagram of RNAs was derived by plotting nucleotides of RNA structures in the PDB database on a 2D axis us-ing their two pseudo-torsion angles η and θ. Then, affinity propagation clus-tering algorithm was applied to the η-θ plot to obtain a structural alphabet (SA) of 23 nucleotide conformations. Next, the SA was used to transform RNA 3D structures into 1D sequences of SA letters. Finally, classical se-quence alignment methods were utilized to compare two SA-encoded se-quences and determine their structural similarities. In fact, many new RNA 3D structures have been deposited in the PDB database since we published iPARTS in 2010. In addition, there is a recent study to report that nucleotide 1D sequences of RNAs can provide useful additional information for their structural alignment. In this study, therefore, we first update our current iPARTS by utilizing currently available RNA 3D structures in the PDB data-base and then design two methods to incorporate RNA 1D sequence infor-mation into our iPARTS such that it can more accurately align two RNA 3D structures based on both their sequences and structures. Finally, our experi-mental results demonstrate that the new version of iPARTS, named iPARTS2, indeed outperforms its previous version in terms of accuracies of RNA struc-ture alignment and functional assignment. In addition, both the methods we propose in this study to incorporate 1D sequence information into iPARTS2 further outperform iPARTS2 with considering only 3D structures of RNAs in terms of accuracy of functional assignment.