RNA-Puzzles toolkit: a computational resource of RNA 3D structure benchmark datasets, structure manipulation, and evaluation tools

trRosettaRNA: automated prediction of RNA 3D structure with transformer network

RNA 3D structure prediction is a long-standing challenge. Inspired by the recent breakthrough in protein structure prediction, we developed trRosettaRNA, an automated deep learning-based approach to RNA 3D structure prediction. The trRosettaRNA pipeline comprises two major steps: 1D and 2D geometries prediction by a transformer network; and 3D structure folding by energy minimization. Benchmark tests suggest that trRosettaRNA outperforms traditional automated methods. In the blind tests of the 15th Critical Assessment of Structure Prediction (CASP15) and the RNA-Puzzles experiments, the automated trRosettaRNA predictions for the natural RNAs are competitive with the top human predictions. trRosettaRNA also outperforms other deep learning-based methods in CASP15 when measured by the Z-score of the Root-Mean-Square Deviation. Nevertheless, it remains challenging to predict accurate structures for synthetic RNAs with an automated approach. We hope this work could be a good start toward solving the hard problem of RNA structure prediction with deep learning.

Assessment of three-dimensional RNA structure prediction in CASP15

The prediction of RNA three-dimensional structures remains an unsolved problem. Here, we report assessments of RNA structure predictions in CASP15, the first CASP exercise that involved RNA structure modeling. Forty two predictor groups submitted models for at least one of twelve RNA-containing targets. These models were evaluated by the RNA-Puzzles organizers and, separately, by a CASP-recruited team using metrics (GDT, lDDT) and approaches (Z-score rankings) initially developed for assessment of proteins and generalized here for RNA assessment. The two assessments independently ranked the same predictor groups as first (AIchemy_RNA2), second (Chen), and third (RNAPolis and GeneSilico, tied); predictions from deep learning approaches were significantly worse than these top ranked groups, which did not use deep learning. Further analyses based on direct comparison of predicted models to cryogenic electron microscopy (cryo-EM) maps and X-ray diffraction data support these rankings. With the exception of two RNA-protein complexes, models submitted by CASP15 groups correctly predicted the global fold of the RNA targets. Comparisons of CASP15 submissions to designed RNA nanostructures as well as molecular replacement trials highlight the potential utility of current RNA modeling approaches for RNA nanotechnology and structural biology, respectively. Nevertheless, challenges remain in modeling fine details such as non-canonical pairs, in ranking among submitted models, and in prediction of multiple structures resolved by cryo-EM or crystallography.

NuFold: A Novel Tertiary RNA Structure Prediction Method Using Deep Learning with Flexible Nucleobase Center Representation

RNA is not only playing a core role in the central dogma as mRNA between DNA and protein, but also many non-coding RNAs have been discovered to have unique and diverse biological functions. As genome sequences become increasingly available and our knowledge of RNA sequences grows, the study of RNA's structure and function has become more demanding. However, experimental determination of three-dimensional RNA structures is both costly and time-consuming, resulting in a substantial disparity between RNA sequence data and structural insights. In response to this challenge, we propose a novel computational approach that harnesses state-of-the-art deep learning architecture NuFold to accurately predict RNA tertiary structures. This approach aims to offer a cost-effective and efficient means of bridging the gap between RNA sequence information and structural comprehension. NuFold implements a nucleobase center representation, which allows it to reproduce all possible nucleotide conformations accurately.

Integrating end-to-end learning with deep geometrical potentials for ab initio RNA structure prediction

RNAs are fundamental in living cells and perform critical functions determined by their tertiary architectures. However, accurate modeling of 3D RNA structure remains a challenging problem. We present a novel method, DRfold, to predict RNA tertiary structures by simultaneous learning of local frame rotations and geometric restraints from experimentally solved RNA structures, where the learned knowledge is converted into a hybrid energy potential to guide RNA structure assembly. The method significantly outperforms previous approaches by >73.3% in TM-score on a sequence-nonredundant dataset containing recently released structures. Detailed analyses showed that the major contribution to the improvements arise from the deep end-to-end learning supervised with the atom coordinates and the composite energy function integrating complementary information from geometry restraints and end-to-end learning models. The open-source DRfold program with fast training protocol allows large-scale application of high-resolution RNA structure modeling and can be further improved with future expansion of RNA structure databases.

RNA 3D structure modeling by fragment assembly with small-angle X-ray scattering restraints

SUMMARY

Structure determination is a key step in the functional characterization of many non-coding RNA molecules. High-resolution RNA 3D structure determination efforts, however, are not keeping up with the pace of discovery of new non-coding RNA sequences. This increases the importance of computational approaches and low-resolution experimental data, such as from the small-angle X-ray scattering experiments. We present RNA Masonry, a computer program and a web service for a fully automated modeling of RNA 3D structures. It assemblies RNA fragments into geometrically plausible models that meet user-provided secondary structure constraints, restraints on tertiary contacts, and small-angle X-ray scattering data. We illustrate the method description with detailed benchmarks and its application to structural studies of viral RNAs with SAXS restraints.

AVAILABILITY AND IMPLEMENTATION

The program web server is available at http://iimcb.genesilico.pl/rnamasonry. The source code is available at https://gitlab.com/gchojnowski/rnamasonry.

Predicting 3D structures and stabilities for complex RNA pseudoknots in ion solutions

RNA pseudoknots are a kind of important tertiary motif, and the structures and stabilities of pseudoknots are generally critical to the biological functions of RNAs with the motifs. In this work, we have carefully refined our previously developed coarse-grained model with salt effect through involving a new coarse-grained force field and a replica-exchange Monte Carlo algorithm, and employed the model to predict structures and stabilities of complex RNA pseudoknots in ion solutions beyond minimal H-type pseudoknots. Compared with available experimental data, the newly refined model can successfully predict 3D structures from sequences for the complex RNA pseudoknots including SARS-CoV-2 programming-1 ribosomal frameshifting element and Zika virus xrRNA, and can reliably predict the thermal stabilities of RNA pseudoknots with various sequences and lengths over broad ranges of monovalent/divalent salts. In addition, for complex pseudoknots including SARS-CoV-2 frameshifting element, our analyses show that their thermally unfolding pathways are mainly dependent on the relative stabilities of unfolded intermediate states, in analogy to those of minimal H-type pseudoknots.

Nucleic acid nanostructures for in vivo applications: The influence of morphology on biological fate

The development of programmable biomaterials for use in nanofabrication represents a major advance for the future of biomedicine and diagnostics. Recent advances in structural nanotechnology using nucleic acids have resulted in dramatic progress in our understanding of nucleic acid-based nanostructures (NANs) for use in biological applications. As the NANs become more architecturally and functionally diverse to accommodate introduction into living systems, there is a need to understand how critical design features can be controlled to impart desired performance in vivo. In this review, we survey the range of nucleic acid materials utilized as structural building blocks (DNA, RNA, and xenonucleic acids), the diversity of geometries for nanofabrication, and the strategies to functionalize these complexes. We include an assessment of the available and emerging characterization tools used to evaluate the physical, mechanical, physiochemical, and biological properties of NANs in vitro. Finally, the current understanding of the obstacles encountered along the in vivo journey is contextualized to demonstrate how morphological features of NANs influence their biological fates. We envision that this summary will aid researchers in the designing novel NAN morphologies, guide characterization efforts, and design of experiments and spark interdisciplinary collaborations to fuel advancements in programmable platforms for biological applications.

cgRNASP: coarse-grained statistical potentials with residue separation for RNA structure evaluation

Knowledge-based statistical potentials are very important for RNA 3-dimensional (3D) structure prediction and evaluation. In recent years, various coarse-grained (CG) and all-atom models have been developed for predicting RNA 3D structures, while there is still lack of reliable CG statistical potentials not only for CG structure evaluation but also for all-atom structure evaluation at high efficiency. In this work, we have developed a series of residue-separation-based CG statistical potentials at different CG levels for RNA 3D structure evaluation, namely cgRNASP, which is composed of long-ranged and short-ranged interactions by residue separation. Compared with the newly developed all-atom rsRNASP, the short-ranged interaction in cgRNASP was involved more subtly and completely. Our examinations show that, the performance of cgRNASP varies with CG levels and compared with rsRNASP, cgRNASP has similarly good performance for extensive types of test datasets and can have slightly better performance for the realistic dataset-RNA-Puzzles dataset. Furthermore, cgRNASP is strikingly more efficient than all-atom statistical potentials/scoring functions, and can be apparently superior to other all-atom statistical potentials and scoring functions trained from neural networks for the RNA-Puzzles dataset. cgRNASP is available at https://github.com/Tan-group/cgRNASP.

Advances and opportunities in RNA structure experimental determination and computational modeling

Beyond transferring genetic information, RNAs are molecules with diverse functions that include catalyzing biochemical reactions and regulating gene expression. Most of these activities depend on RNAs' specific structures. Therefore, accurately determining RNA structure is integral to advancing our understanding of RNA functions. Here, we summarize the state-of-the-art experimental and computational technologies developed to evaluate RNA secondary and tertiary structures. We also highlight how the rapid increase of experimental data facilitates the integrative modeling approaches for better resolving RNA structures. Finally, we provide our thoughts on the latest advances and challenges in RNA structure determination methods, as well as on future directions for both experimental approaches and artificial intelligence-based computational tools to model RNA structure. Ultimately, we hope the technological advances will deepen our understanding of RNA biology and facilitate RNA structure-based biomedical research such as designing specific RNA structures for therapeutics and deploying RNA-targeting small-molecule drugs.

Computational Pipeline for Reference-Free Comparative Analysis of RNA 3D Structures Applied to SARS-CoV-2 UTR Models

RNA is a unique biomolecule that is involved in a variety of fundamental biological functions, all of which depend solely on its structure and dynamics. Since the experimental determination of crystal RNA structures is laborious, computational 3D structure prediction methods are experiencing an ongoing and thriving development. Such methods can lead to many models; thus, it is necessary to build comparisons and extract common structural motifs for further medical or biological studies. Here, we introduce a computational pipeline dedicated to reference-free high-throughput comparative analysis of 3D RNA structures. We show its application in the RNA-Puzzles challenge, in which five participating groups attempted to predict the three-dimensional structures of 5'- and 3'-untranslated regions (UTRs) of the SARS-CoV-2 genome. We report the results of this puzzle and discuss the structural motifs obtained from the analysis. All simulated models and tools incorporated into the pipeline are open to scientific and academic use.

RNAsolo: a repository of cleaned PDB-derived RNA 3D structures

MOTIVATION

The development of algorithms dedicated to RNA 3D structures contributes to the demand for training, testing, and benchmarking data. A reliable source of such data derived from computational prediction is the RNA-Puzzles repository. In contrast, the largest resource with experimentally determined structures is the Protein Data Bank. However, files in this archive often contain other molecular data in addition to the RNA structure itself, which-to be used by RNA processing algorithms-should be removed.

RESULTS

RNAsolo is a self-updating database dedicated to RNA bioinformatics. It systematically collects experimentally determined RNA 3D structures stored in the PDB, cleans them from non-RNA chains, and groups them into equivalence classes. It allows users to download various subsets of data-clustered by resolution, source, data format, etc. - for further processing and analysis with a single click.

AVAILABILITY

The repository is publicly available at https://rnasolo.cs.put.poznan.pl.

RNAglib: a python package for RNA 2.5 D graphs

SUMMARY

RNA 3D architectures are stabilized by sophisticated networks of (non-canonical) base pair interactions, which can be conveniently encoded as multi-relational graphs and efficiently exploited by graph theoretical approaches and recent progresses in machine learning techniques. RNAglib is a library that eases the use of this representation, by providing clean data, methods to load it in machine learning pipelines and graph-based deep learning models suited for this representation. RNAglib also offers other utilities to model RNA with 2.5 D graphs, such as drawing tools, comparison functions or baseline performances on RNA applications.

AVAILABILITY AND IMPLEMENTATION

The method is distributed as a pip package, RNAglib. Data are available in a repository and can be accessed on rnaglib's web page. The source code, data and documentation are available at https://rnaglib.cs.mcgill.ca.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Evaluation of the stereochemical quality of predicted RNA 3D models in the RNA-Puzzles submissions

In silico prediction is a well-established approach to derive a general shape of an RNA molecule based on its sequence or secondary structure. This paper reports an analysis of the stereochemical quality of the RNA three-dimensional models predicted using dedicated computer programs. The stereochemistry of 1052 RNA 3D structures, including 1030 models predicted by fully automated and human-guided approaches within 22 RNA-Puzzles challenges and reference structures, is analyzed. The evaluation is based on standards of RNA stereochemistry that the Protein Data Bank requires from deposited experimental structures. Deviations from standard bond lengths and angles, planarity, or chirality are quantified. A reduction in the number of such deviations should help in the improvement of RNA 3D structure modeling approaches.

RNAStat: An Integrated Tool for Statistical Analysis of RNA 3D Structures

The 3D architectures of RNAs are essential for understanding their cellular functions. While an accurate scoring function based on the statistics of known RNA structures is a key component for successful RNA structure prediction or evaluation, there are few tools or web servers that can be directly used to make comprehensive statistical analysis for RNA 3D structures. In this work, we developed RNAStat, an integrated tool for making statistics on RNA 3D structures. For given RNA structures, RNAStat automatically calculates RNA structural properties such as size and shape, and shows their distributions. Based on the RNA structure annotation from DSSR, RNAStat provides statistical information of RNA secondary structure motifs including canonical/non-canonical base pairs, stems, and various loops. In particular, the geometry of base-pairing/stacking can be calculated in RNAStat by constructing a local coordinate system for each base. In addition, RNAStat also supplies the distribution of distance between any atoms to the users to help build distance-based RNA statistical potentials. To test the usability of the tool, we established a non-redundant RNA 3D structure dataset, and based on the dataset, we made a comprehensive statistical analysis on RNA structures, which could have the guiding significance for RNA structure modeling. The python code of RNAStat, the dataset used in this work, and corresponding statistical data files are freely available at GitHub (https://github.com/RNA-folding-lab/RNAStat).

Entanglements of structure elements revealed in RNA 3D models

Computational methods to predict RNA 3D structure have more and more practical applications in molecular biology and medicine. Therefore, it is crucial to intensify efforts to improve the accuracy and quality of predicted three-dimensional structures. A significant role in this is played by the RNA-Puzzles initiative that collects, evaluates, and shares RNAs built computationally within currently nearly 30 challenges. RNA-Puzzles datasets, subjected to multi-criteria analysis, allow revealing the strengths and weaknesses of computer prediction methods. Here, we study the issue of entangled RNA fragments in the predicted RNA 3D structure models. By entanglement, we mean an arrangement of two structural elements such that one of them passes through the other. We propose the classification of entanglements driven by their topology and components. It distinguishes two general classes, interlaces and lassos, and subclasses characterized by element types-loops, dinucleotide steps, open single-stranded fragments-and puncture multiplicity. Our computational pipeline for entanglement detection, applied for 1,017 non-redundant models from RNA-Puzzles, has shown the frequency of different entanglements and allowed identifying 138 structures with intersected assemblies.

Structural 3D Domain Reconstruction of the RNA Genome from Viruses with Secondary Structure Models

Three-dimensional RNA domain reconstruction is important for the assembly, disassembly and delivery functionalities of a packed proteinaceus capsid. However, to date, the self-association of RNA molecules is still an open problem. Recent chemical probing reports provide, with high reliability, the secondary structure of diverse RNA ensembles, such as those of viral genomes. Here, we present a method for reconstructing the complete 3D structure of RNA genomes, which combines a coarse-grained model with a subdomain composition scheme to obtain the entire genome inside proteinaceus capsids based on secondary structures from experimental techniques. Despite the amount of sampling involved in the folded and also unfolded RNA molecules, advanced microscope techniques can provide points of anchoring, which enhance our model to include interactions between capsid pentamers and RNA subdomains. To test our method, we tackle the satellite tobacco mosaic virus (STMV) genome, which has been widely studied by both experimental and computational communities. We provide not only a methodology to structurally analyze the tertiary conformations of the RNA genome inside capsids, but a flexible platform that allows the easy implementation of features/descriptors coming from both theoretical and experimental approaches.

An RNA-centric historical narrative around the Protein Data Bank

Some of the amazing contributions brought to the scientific community by the Protein Data Bank (PDB) are described. The focus is on nucleic acid structures with a bias toward RNA. The evolution and key roles in science of the PDB and other structural databases for nucleic acids illustrate how small initial ideas can become huge and indispensable resources with the unflinching willingness of scientists to cooperate globally. The progress in the understanding of the molecular interactions driving RNA architectures followed the rapid increase in RNA structures in the PDB. That increase was consecutive to improvements in chemical synthesis and purification of RNA molecules, as well as in biophysical methods for structure determination and computer technology. The RNA modeling efforts from the early beginnings are also described together with their links to the state of structural knowledge and technological development. Structures of RNA and of its assemblies are physical objects, which, together with genomic data, allow us to integrate present-day biological functions and the historical evolution in all living species on earth.

IsRNA1: De Novo Prediction and Blind Screening of RNA 3D Structures

Modeling structures and functions of large ribonucleic acid (RNAs) especially with complicated topologies is highly challenging due to the inefficiency of large conformational sampling and the presence of complicated tertiary interactions. To address this problem, one highly promising approach is coarse-grained modeling. Here, following an iterative simulated reference state approach to decipher the correlations between different structural parameters, we developed a potent coarse-grained RNA model named as IsRNA1 for RNA studies. Molecular dynamics simulations in the IsRNA1 can predict the native structures of small RNAs from a sequence and fold medium-sized RNAs into near-native tertiary structures with the assistance of secondary structure constraints. A large-scale benchmark test on RNA 3D structure prediction shows that IsRNA1 exhibits improved performance for relatively large RNAs of complicated topologies, such as large stem-loop structures and structures containing long-range tertiary interactions. The advantages of IsRNA1 include the consideration of the correlations between the different structural variables, the appropriate characterization of canonical base-pairing and base-stacking interactions, and the better sampling for the backbone conformations. Moreover, a blind screening protocol was developed based on IsRNA1 to identify good structural models from a pool of candidates without prior knowledge of the native structures.

Saccharomyces cerevisiae Ecm2 Modulates the Catalytic Steps of pre-mRNA Splicing

Genetic, biochemical, and structural studies have elucidated the molecular basis for spliceosome catalysis. Splicing is RNA catalyzed and the essential snRNA and protein factors are well-conserved. However, little is known about how non-essential components of the spliceosome contribute to the reaction and modulate the activities of the fundamental core machinery. Ecm2 is a non-essential yeast splicing factor that is a member of the Prp19-related complex of proteins. Cryo-electron microscopy (cryo-EM) structures have revealed that Ecm2 binds the U6 snRNA and is entangled with Cwc2, a factor previously found to promote a catalytically active conformation of the spliceosome. These structures also indicate that Ecm2 and the U2 snRNA likely form a transient interaction during 5' splice site (SS) cleavage. We have characterized genetic interactions between ECM2 and alleles of splicing factors that alter the catalytic steps in splicing. In addition, we have studied how loss of ECM2 impacts splicing of pre-mRNAs containing non-consensus or competing SS. Our results show that ECM2 functions during the catalytic stages of splicing. Our data are consistent with Ecm2 facilitating the formation and stabilization of the 1st-step catalytic site, promoting 2nd-step catalysis, and permiting alternate 5' SS usage. We propose that Cwc2 and Ecm2 can each fine-tune the spliceosome active site in unique ways. Their interaction network may act as a conduit through which splicing of certain pre-mRNAs, such as those containing weak or alternate splice sites, can be regulated.

AnnapuRNA: A scoring function for predicting RNA-small molecule binding poses

RNA is considered as an attractive target for new small molecule drugs. Designing active compounds can be facilitated by computational modeling. Most of the available tools developed for these prediction purposes, such as molecular docking or scoring functions, are parametrized for protein targets. The performance of these methods, when applied to RNA-ligand systems, is insufficient. To overcome these problems, we developed AnnapuRNA, a new knowledge-based scoring function designed to evaluate RNA-ligand complex structures, generated by any computational docking method. We also evaluated three main factors that may influence the structure prediction, i.e., the starting conformer of a ligand, the docking program, and the scoring function used. We applied the AnnapuRNA method for a post-hoc study of the recently published structures of the FMN riboswitch. Software is available at https://github.com/filipspl/AnnapuRNA.

Drug development is a lengthy and complicated process, which requires costly experiments on a very large number of chemical compounds. The identification of chemical molecules with desired properties can be facilitated by computational methods. Several methods were developed for computer-aided design of drugs that target protein molecules. However, recently the ribonucleic acid (RNA) emerged as an attractive target for the development of new drugs. Unfortunately, the portfolio of the computer methods that can be applied to study RNA and its interactions with small chemical molecules is very limited. This situation motivated us to develop a new computational method, with which to predict RNA-small molecule interactions. To this end, we collected the information on the statistics of interactions in experimentally determined structures of complexes formed by RNA with small molecules. We then used the statistical data to train machine learning methods aiming to distinguish between RNA-ligand interactions observed experimentally and other interactions that can be observed in theoretical analyses, but are not observed in nature. The resulting method called AnnapuRNA is superior to other similar tools and can be used to predict preferred ligands of RNA molecules and how RNA and small molecules interact with each other.

Advances in RNA 3D Structure Modeling Using Experimental Data

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.

RaptRanker: in silico RNA aptamer selection from HT-SELEX experiment based on local sequence and structure information

Aptamers are short single-stranded RNA/DNA molecules that bind to specific target molecules. Aptamers with high binding-affinity and target specificity are identified using an in vitro procedure called high throughput systematic evolution of ligands by exponential enrichment (HT-SELEX). However, the development of aptamer affinity reagents takes a considerable amount of time and is costly because HT-SELEX produces a large dataset of candidate sequences, some of which have insufficient binding-affinity. Here, we present RNA aptamer Ranker (RaptRanker), a novel in silico method for identifying high binding-affinity aptamers from HT-SELEX data by scoring and ranking. RaptRanker analyzes HT-SELEX data by evaluating the nucleotide sequence and secondary structure simultaneously, and by ranking according to scores reflecting local structure and sequence frequencies. To evaluate the performance of RaptRanker, we performed two new HT-SELEX experiments, and evaluated binding affinities of a part of sequences that include aptamers with low binding-affinity. In both datasets, the performance of RaptRanker was superior to Frequency, Enrichment and MPBind. We also confirmed that the consideration of secondary structures is effective in HT-SELEX data analysis, and that RaptRanker successfully predicted the essential subsequence motifs in each identified sequence.

FARFAR2: Improved De Novo Rosetta Prediction of Complex Global RNA Folds

Predicting RNA three-dimensional structures from sequence could accelerate understanding of the growing number of RNA molecules being discovered across biology. Rosetta's Fragment Assembly of RNA with Full-Atom Refinement (FARFAR) has shown promise in community-wide blind RNA-Puzzle trials, but lack of a systematic and automated benchmark has left unclear what limits FARFAR performance. Here, we benchmark FARFAR2, an algorithm integrating RNA-Puzzle-inspired innovations with updated fragment libraries and helix modeling. In 16 of 21 RNA-Puzzles revisited without experimental data or expert intervention, FARFAR2 recovers native-like structures more accurate than models submitted during the RNA-Puzzles trials. Remaining bottlenecks include conformational sampling for >80-nucleotide problems and scoring function limitations more generally. Supporting these conclusions, preregistered blind models for adenovirus VA-I RNA and five riboswitch complexes predicted native-like folds with 3- to 14 Å root-mean-square deviation accuracies. We present a FARFAR2 webserver and three large model archives (FARFAR2-Classics, FARFAR2-Motifs, and FARFAR2-Puzzles) to guide future applications and advances.

RNA-Puzzles Round IV: 3D structure predictions of four ribozymes and two aptamers

RNA-Puzzles is a collective endeavor dedicated to the advancement and improvement of RNA 3D structure prediction. With agreement from crystallographers, the RNA structures are predicted by various groups before the publication of the crystal structures. We now report the prediction of 3D structures for six RNA sequences: four nucleolytic ribozymes and two riboswitches. Systematic protocols for comparing models and crystal structures are described and analyzed. In these six puzzles, we discuss (i) the comparison between the automated web servers and human experts; (ii) the prediction of coaxial stacking; (iii) the prediction of structural details and ligand binding; (iv) the development of novel prediction methods; and (v) the potential improvements to be made. We show that correct prediction of coaxial stacking and tertiary contacts is essential for the prediction of RNA architecture, while ligand binding modes can only be predicted with low resolution and simultaneous prediction of RNA structure with accurate ligand binding still remains out of reach. All the predicted models are available for the future development of force field parameters and the improvement of comparison and assessment tools.