Advances in RNA 3D Structure Modeling Using Experimental Data

Abstract

RNA is a unique bio-macromolecule that can both record genetic information and perform biological functions in a variety of molecular processes, including transcription, splicing, translation, and even regulating protein function. RNAs adopt specific three-dimensional conformations to enable their functions. Experimental determination of high-resolution RNA structures using x-ray crystallography is both laborious and demands expertise, thus, hindering our comprehension of RNA structural biology. The computational modeling of RNA structure was a milestone in the birth of bioinformatics. Although computational modeling has been greatly improved over the last decade showing many successful cases, the accuracy of such computational modeling is not only length-dependent but also varies according to the complexity of the structure. To increase credibility, various experimental data were integrated into computational modeling. In this review, we summarize the experiments that can be integrated into RNA structure modeling as well as the computational methods based on these experimental data. We also demonstrate how computational modeling can help the experimental determination of RNA structure. We highlight the recent advances in computational modeling which can offer reliable structure models using high-throughput experimental data.

Keywords: 3D shape; RNA structure; RNA-puzzles; chemical probing; structure prediction.

FIGURE 1

Graphical illustration of the experimental data that can be used in RNA structure modeling. (A) A graphical summary shows how structural information derived from biophysical and biochemical experiments can be used for structure modeling. Different experiments indicate different types of structural information: the three-dimensional shape of a molecule can be given by x-ray crystallography, cryo-EM, or SAS; pairwise interactions, including base-pair interactions and atomic contacts, are indicated from NMR or mutate-and-map; and features of a single nucleotide are inferred from chemical probing. (B) A scheme shows that the next-generation sequencing technique can be applied to the cDNA sets generated from the biochemical probing experiments in order to increase the throughput of the experiments. Chemical probing reagents modify the exposed nucleotides and result in adducts (orange hexagons). Adducts interrupt the reverse transcription known as RT-stop, while RT-mutate means that reverse transcription introduces a mismatched nucleotide (black dots) at the position of the adduct under special conditions. The sequencing results of the cDNA sets from RT-stop/RT-mutate can be transformed back to structural restraints.

FIGURE 2

An example of structure modeling based on M2-seq (or MOHCA-seq). This scheme shows the workflow of M2-seq and MOHCA-seq experiments. Both approaches are based on the assumption that the mutated nucleotide in a base-pair tends to become more exposed and more detectable by chemical mapping. M2-seq uses DMS to probe the unpaired nucleotides introduced by error-prone PCR, while sequencing data can be analyzed by the M2-net algorithm based on the RT-stop mechanism. MOHCA-seq uses 2′-NH2-2′-dATP and isothio cyanobenzyl-Fe(III)EDTA to introduce Fe adducts into the RNA. While the Fenton reaction damages other nucleotides nearby the reacting Fe-modified nucleotide. Sequencing data analyzed by the MAPseeker algorithm highlights the proximally tertiary structure interactions.

FIGURE 3

Examples of non-canonical interactions. (A) The UA-U-rich RNA triple helix in MALAT1_th11 RNA, PDB id 6SVS. (B) The simple k-turn with two trans-sugar-Hoogsteen G:A base pairs, PDB id 6HCT. (C) The structure of the G18 > G2:C39 triple interaction in glutamine-II riboswitch, PDB id 6QN3.

FIGURE 4

The prediction of non-canonical base-pairs. (A) The comparison between Watson–Crick and non-Watson–Crick base-pairs prediction in terms of interaction network fidelity (INF; Parisien et al., 2009) in RNA-Puzzles. (B) The superimposition of recurrent kink-turn modules. (C) The graphical abstraction of the kink-turn module. (D) The module abstraction in the Basepairing program.

FIGURE 5

The approaches of RNA 3D structure modeling. Comparative modeling, fragment assembly, and de novo modeling are the basic approaches in predicting RNA structures. Comparative modeling is based on the availability of one or more homologous structures, while fragment assembly uses known structural fragments to assemble the structure. De novo modeling searches for the best conformation in the space considering the physical or empirical force-fields. Experimental data generates different types of restraints in structure modeling and can be applied in different modeling approaches. Details of the restraints and the related modeling approaches are explained in the main text.

FIGURE 6

Using chemical mapping-based structure modeling to model low-resolution cryo-EM data (Ribosolve). The Ribosolve approach is a recent example of integrative modeling using moderate-resolution cryo-EM maps, chemical mapping, and Rosetta computational modeling. M2-seq probes the base-pair interactions, while cryo-EM restrains the three-dimensional shape of the structure topology. Force-field-based modeling optimizes the structure based on the learned knowledge of the force-field derived from known RNA structures. The generated models are assessed by modeling convergence, while mutate-map-rescue provides an alternative to optimize the secondary structure information and improve the modeling.