STAR3D: a stack-based RNA 3D structural alignment tool.

Ping Ge,Shaojie Zhang

doi:10.1093/nar/gkv697

Abstract

The various roles of versatile non-coding RNAs typically require the attainment of complex high-order structures. Therefore, comparing the 3D structures of RNA molecules can yield in-depth understanding of their functional conservation and evolutionary history. Recently, many powerful tools have been developed to align RNA 3D structures. Although some methods rely on both backbone conformations and base pairing interactions, none of them consider the entire hierarchical formation of the RNA secondary structure. One of the major issues is that directly applying the algorithms of matching 2D structures to the 3D coordinates is particularly time-consuming. In this article, we propose a novel RNA 3D structural alignment tool, STAR3D, to take into full account the 2D relations between stacks without the complicated comparison of secondary structures. First, the 3D conserved stacks in the inputs are identified and then combined into a tree-like consensus. Afterward, the loop regions are compared one-to-one in accordance with their relative positions in the consensus tree. The experimental results show that the prediction of STAR3D is more accurate for both non-homologous and homologous RNAs than other state-of-the-art tools with shorter running time.

Highlights

Non-coding RNAs play diverse cellular functions in biological systems [1,2,3,4]
Elastic Shape Analysis (ESA) models the RNA 3D structures not as sequences but as curves in a four-dimensional space: the atomic coordinates are in 3D space and the sequence information is encoded as an additional dimension [18]
The inputs of STAR3D are the atomic coordinates of two polymer RNA chains, which are presented in the corresponding Protein Data Bank (PDB) files

Summary

Introduction

Non-coding RNAs (ncRNAs) play diverse cellular functions in biological systems [1,2,3,4]. With the rapid growth of RNA deposition in the Protein Data Bank (PDB) [12], a number of tools have been developed for the alignments of RNA 3D structures They can be categorized into two groups. Similar to LaJolla, FRIEs [15] uses the matching of k-mer RNA fragments In this method, a large set of training fragments from the PDB are clustered into tens of classes based on their structural properties. The matches between n-body cliques (in which n member nucleotides satisfy that all pair-wise spatial distances are within a threshold) are determined by the superimposition of their atomic coordinates With this local structural equivalence, the optimal global alignment is generated by using 3D least squares fitting.

Methods

Results

Conclusion