RNA structure prediction: an overview of methods

Targeting non-coding RNAs: Perspectives and challenges of in-silico approaches

The growing information currently available on the central role of non-coding RNAs (ncRNAs) including microRNAs (miRNAS) and long non-coding RNAs (lncRNAs) for chronic and degenerative human diseases makes them attractive therapeutic targets. RNAs carry out different functional roles in human biology and are deeply deregulated in several diseases. So far, different attempts to therapeutically target the 3D RNA structures with small molecules have been reported. In this scenario, the development of computational tools suitable for describing RNA structures and their potential interactions with small molecules is gaining more and more interest. Here, we describe the most suitable strategies to study ncRNAs through computational tools. We focus on methods capable of predicting 2D and 3D ncRNA structures. Furthermore, we describe computational tools to identify, design and optimize small molecule ncRNA binders. This review aims to outline the state of the art and perspectives of computational methods for ncRNAs over the past decade.

Machine learning modeling of RNA structures: methods, challenges and future perspectives

The three-dimensional structure of RNA molecules plays a critical role in a wide range of cellular processes encompassing functions from riboswitches to epigenetic regulation. These RNA structures are incredibly dynamic and can indeed be described aptly as an ensemble of structures that shifts in distribution depending on different cellular conditions. Thus, the computational prediction of RNA structure poses a unique challenge, even as computational protein folding has seen great advances. In this review, we focus on a variety of machine learning-based methods that have been developed to predict RNA molecules' secondary structure, as well as more complex tertiary structures. We survey commonly used modeling strategies, and how many are inspired by or incorporate thermodynamic principles. We discuss the shortcomings that various design decisions entail and propose future directions that could build off these methods to yield more robust, accurate RNA structure predictions.

Machine learning for RNA 2D structure prediction benchmarked on experimental data

Since the 1980s, dozens of computational methods have addressed the problem of predicting RNA secondary structure. Among them are those that follow standard optimization approaches and, more recently, machine learning (ML) algorithms. The former were repeatedly benchmarked on various datasets. The latter, on the other hand, have not yet undergone extensive analysis that could suggest to the user which algorithm best fits the problem to be solved. In this review, we compare 15 methods that predict the secondary structure of RNA, of which 6 are based on deep learning (DL), 3 on shallow learning (SL) and 6 control methods on non-ML approaches. We discuss the ML strategies implemented and perform three experiments in which we evaluate the prediction of (I) representatives of the RNA equivalence classes, (II) selected Rfam sequences and (III) RNAs from new Rfam families. We show that DL-based algorithms (such as SPOT-RNA and UFold) can outperform SL and traditional methods if the data distribution is similar in the training and testing set. However, when predicting 2D structures for new RNA families, the advantage of DL is no longer clear, and its performance is inferior or equal to that of SL and non-ML methods.

Sequence similarity governs generalizability of de novo deep learning models for RNA secondary structure prediction

Making no use of physical laws or co-evolutionary information, de novo deep learning (DL) models for RNA secondary structure prediction have achieved far superior performances than traditional algorithms. However, their statistical underpinning raises the crucial question of generalizability. We present a quantitative study of the performance and generalizability of a series of de novo DL models, with a minimal two-module architecture and no post-processing, under varied similarities between seen and unseen sequences. Our models demonstrate excellent expressive capacities and outperform existing methods on common benchmark datasets. However, model generalizability, i.e., the performance gap between the seen and unseen sets, degrades rapidly as the sequence similarity decreases. The same trends are observed from several recent DL and machine learning models. And an inverse correlation between performance and generalizability is revealed collectively across all learning-based models with wide-ranging architectures and sizes. We further quantitate how generalizability depends on sequence and structure identity scores via pairwise alignment, providing unique quantitative insights into the limitations of statistical learning. Generalizability thus poses a major hurdle for deploying de novo DL models in practice and various pathways for future advances are discussed.

RNA independent fragment partition method based on deep learning for RNA secondary structure prediction

The non-coding RNA secondary structure largely determines its function. Hence, accuracy in structure acquisition is of great importance. Currently, this acquisition primarily relies on various computational methods. The prediction of the structures of long RNA sequences with high precision and reasonable computational cost remains challenging. Here, we propose a deep learning model, RNA-par, which could partition an RNA sequence into several independent fragments (i-fragments) based on its exterior loops. Each i-fragment secondary structure predicted individually could be further assembled to acquire the complete RNA secondary structure. In the examination of our independent test set, the average length of the predicted i-fragments was 453 nt, which was considerably shorter than that of complete RNA sequences (848 nt). The accuracy of the assembled structures was higher than that of the structures predicted directly using the state-of-the-art RNA secondary structure prediction methods. This proposed model could serve as a preprocessing step for RNA secondary structure prediction for enhancing the predictive performance (especially for long RNA sequences) and reducing the computational cost. In the future, predicting the secondary structure of long-sequence RNA with high accuracy can be enabled by developing a framework combining RNA-par with various existing RNA secondary structure prediction algorithms. Our models, test codes and test data are provided at https://github.com/mianfei71/RNAPar .

Linear-Time Algorithms for RNA Structure Prediction

RNA secondary structure prediction is widely used to understand RNA function. Existing dynamic programming-based algorithms, both the classical minimum free energy (MFE) methods and partition function methods, suffer from a major limitation: their runtimes scale cubically with the RNA length, and this slowness limits their use in genome-wide applications. Inspired by incremental parsing for context-free grammars in computational linguistics, we designed linear-time heuristic algorithms, LinearFold and LinearPartition, to approximate the MFE structure, partition function and base pairing probabilities. These programs are orders of magnitude faster than Vienna RNAfold and CONTRAfold on long sequences. More interestingly, LinearFold and LinearPartition lead to more accurate predictions on the longest sequence families for which the structures are well established (16S and 23S Ribosomal RNAs), as well as improved accuracies for long-range base pairs (500 +  nucleotides apart). This chapter provides protocols for using LinearFold and LinearPartition for secondary structure prediction.

Artificial intelligence methods enhance the discovery of RNA interactions

Understanding how RNAs interact with proteins, RNAs, or other molecules remains a challenge of main interest in biology, given the importance of these complexes in both normal and pathological cellular processes. Since experimental datasets are starting to be available for hundreds of functional interactions between RNAs and other biomolecules, several machine learning and deep learning algorithms have been proposed for predicting RNA-RNA or RNA-protein interactions. However, most of these approaches were evaluated on a single dataset, making performance comparisons difficult. With this review, we aim to summarize recent computational methods, developed in this broad research area, highlighting feature encoding and machine learning strategies adopted. Given the magnitude of the effect that dataset size and quality have on performance, we explored the characteristics of these datasets. Additionally, we discuss multiple approaches to generate datasets of negative examples for training. Finally, we describe the best-performing methods to predict interactions between proteins and specific classes of RNA molecules, such as circular RNAs (circRNAs) and long non-coding RNAs (lncRNAs), and methods to predict RNA-RNA or RNA-RBP interactions independently of the RNA type.

Computational Modeling of RNA Aptamers: Structure Prediction of the Apo State

RNA aptamers are single-stranded oligonucleotides that bind to specific molecular targets with high affinity and specificity. To design aptamers for new applications, it is critical to understand the ligand binding mechanism in terms of the structure and dynamics of the ligand-bound and apo states. The problem is that most of the NMR or X-ray crystal structures available for RNA aptamers are for ligand-bound states. Available apo state structures, mostly characterized by crystallization under nonphysiological conditions or probed by low resolution techniques, might fail to represent the diverse structural variations of the apo state in solution. Here, we develop an approach to obtain a representative ensemble of apo structures that are based on in silico RNA 3D structure prediction and in vitro experiments that characterize base stacking. Using the neomycin-B aptamer as a case study, an ensemble of structures for the aptamer in the apo (unbound) state are validated and then used to investigate the ligand-binding mechanism for the aptamer in complex with neomycin-B.

RAFFT: Efficient prediction of RNA folding pathways using the fast Fourier transform

We propose a novel heuristic to predict RNA secondary structure formation pathways that has two components: (i) a folding algorithm and (ii) a kinetic ansatz. This heuristic is inspired by the kinetic partitioning mechanism, by which molecules follow alternative folding pathways to their native structure, some much faster than others. Similarly, our algorithm RAFFT starts by generating an ensemble of concurrent folding pathways ending in multiple metastable structures, which is in contrast with traditional thermodynamic approaches that find single structures with minimal free energies. When we constrained the algorithm to predict only 50 structures per sequence, near-native structures were found for RNA molecules of length ≤ 200 nucleotides. Our heuristic has been tested on the coronavirus frameshifting stimulation element (CFSE): an ensemble of 68 distinct structures allowed us to produce complete folding kinetic trajectories, whereas known methods require evaluating millions of sub-optimal structures to achieve this result. Thanks to the fast Fourier transform on which RAFFT (RNA folding Algorithm wih Fast Fourier Transform) is based, these computations are efficient, with complexity O ( L 2logL ).

Using Unassigned NMR Chemical Shifts to Model RNA Secondary Structure

NMR-derived chemical shifts are sensitive probes of RNA structure. However, the need to assign NMR spectra hampers their utility as a direct source of structural information. In this report, we describe a simple method that uses unassigned 2D NMR spectra to model the secondary structure of RNAs. As in the case of assigned chemical shifts, we could use unassigned chemical shift data to reweight conformational libraries such that the highest weighted structure closely resembles their reference NMR structure. Furthermore, the application of our approach to the 3'- and 5'-UTR of the SARS-CoV-2 genome yields structures that are, for the most part, consistent with the secondary structure models derived from chemical probing data. Therefore, we expect the framework we describe here will be useful as a general strategy for rapidly generating preliminary structural RNA models directly from unassigned 2D NMR spectra. As we demonstrated for the 337-nt and 472-nt UTRs of SARS-CoV-2, our approach could be especially valuable for modeling the secondary structures of large RNA.

A Comprehensive Computational Investigation into the Conserved Virulent Proteins of Shigella species Unveils Potential Small-Interfering RNA Candidates as a New Therapeutic Strategy against Shigellosis

Shigella species account for the second-leading cause of deaths due to diarrheal diseases among children of less than 5 years of age. The emergence of multi-drug-resistant Shigella isolates and the lack of availability of Shigella vaccines have led to the pertinence in the efforts made for the development of new therapeutic strategies against shigellosis. Consequently, designing small-interfering RNA (siRNA) candidates against such infectious agents represents a novel approach to propose new therapeutic candidates to curb the rampant rise of anti-microbial resistance in such pathogens. In this study, we analyzed 264 conserved sequences from 15 different conserved virulence genes of Shigella sp., through extensive rational validation using a plethora of first-generation and second-generation computational algorithms for siRNA designing. Fifty-eight siRNA candidates were obtained by using the first-generation algorithms, out of which only 38 siRNA candidates complied with the second-generation rules of siRNA designing. Further computational validation showed that 16 siRNA candidates were found to have a substantial functional efficiency, out of which 11 siRNA candidates were found to be non-immunogenic. Finally, three siRNA candidates exhibited a sterically feasible three-dimensional structure as exhibited by parameters of nucleic acid geometry such as: the probability of wrong sugar puckers, bad backbone confirmations, bad bonds, and bad angles being within the accepted threshold for stable tertiary structure. Although the findings of our study require further wet-lab validation and optimization for therapeutic use in the treatment of shigellosis, the computationally validated siRNA candidates are expected to suppress the expression of the virulence genes, namely: IpgD (siRNA 9) and OspB (siRNA 15 and siRNA 17) and thus act as a prospective tool in the RNA interference (RNAi) pathway. However, the findings of our study require further wet-lab validation and optimization for regular therapeutic use for treatment of shigellosis.

Secondary structure prediction for RNA sequences including N6-methyladenosine

There is increasing interest in the roles of covalently modified nucleotides in RNA. There has been, however, an inability to account for modifications in secondary structure prediction because of a lack of software and thermodynamic parameters. We report the solution for these issues for N⁶-methyladenosine (m⁶A), allowing secondary structure prediction for an alphabet of A, C, G, U, and m⁶A. The RNAstructure software now works with user-defined nucleotide alphabets of any size. We also report a set of nearest neighbor parameters for helices and loops containing m⁶A, using experiments. Interestingly, N⁶-methylation decreases folding stability for adenosines in the middle of a helix, has little effect on folding stability for adenosines at the ends of helices, and increases folding stability for unpaired adenosines stacked on a helix. We demonstrate predictions for an N⁶-methylation-activated protein recognition site from MALAT1 and human transcriptome-wide effects of N⁶-methylation on the probability of adenosine being buried in a helix.

RNA folding free energy nearest neighbor parameters were determined for sequences with the nucleotide m⁶A. The RNAstructure software package can accommodate modified nucleotides, enabling secondary structure prediction of sequences with m⁶A.

UFold: fast and accurate RNA secondary structure prediction with deep learning

For many RNA molecules, the secondary structure is essential for the correct function of the RNA. Predicting RNA secondary structure from nucleotide sequences is a long-standing problem in genomics, but the prediction performance has reached a plateau over time. Traditional RNA secondary structure prediction algorithms are primarily based on thermodynamic models through free energy minimization, which imposes strong prior assumptions and is slow to run. Here, we propose a deep learning-based method, called UFold, for RNA secondary structure prediction, trained directly on annotated data and base-pairing rules. UFold proposes a novel image-like representation of RNA sequences, which can be efficiently processed by Fully Convolutional Networks (FCNs). We benchmark the performance of UFold on both within- and cross-family RNA datasets. It significantly outperforms previous methods on within-family datasets, while achieving a similar performance as the traditional methods when trained and tested on distinct RNA families. UFold is also able to predict pseudoknots accurately. Its prediction is fast with an inference time of about 160 ms per sequence up to 1500 bp in length. An online web server running UFold is available at https://ufold.ics.uci.edu. Code is available at https://github.com/uci-cbcl/UFold.

RNA Structure Prediction, Analysis, and Design: An Introduction to Web-Based Tools

Understanding RNA structure has become critical in the study of RNA in their roles as mediators of biological processes. To aid in these studies, computational algorithms that utilize thermodynamics have been developed to predict RNA secondary structure. Due to the importance of intermolecular interactions, the algorithms have been expanded to determine and predict RNA-RNA hybridization. This chapter discusses popular webservers with the tools for RNA secondary structure prediction, RNA-RNA hybridization, and design. We address key features that distinguish common-functioning programs and their purposes for the interests of the user. Ultimately, we hope this review elucidates web-based tools researchers may take advantage of in their investigations of RNA structure and function.

LinearTurboFold: Linear-time global prediction of conserved structures for RNA homologs with applications to SARS-CoV-2

The constant emergence of COVID-19 variants reduces the effectiveness of existing vaccines and test kits. Therefore, it is critical to identify conserved structures in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) genomes as potential targets for variant-proof diagnostics and therapeutics. However, the algorithms to predict these conserved structures, which simultaneously fold and align multiple RNA homologs, scale at best cubically with sequence length and are thus infeasible for coronaviruses, which possess the longest genomes (∼30,000 nt) among RNA viruses. As a result, existing efforts on modeling SARS-CoV-2 structures resort to single-sequence folding as well as local folding methods with short window sizes, which inevitably neglect long-range interactions that are crucial in RNA functions. Here we present LinearTurboFold, an efficient algorithm for folding RNA homologs that scales linearly with sequence length, enabling unprecedented global structural analysis on SARS-CoV-2. Surprisingly, on a group of SARS-CoV-2 and SARS-related genomes, LinearTurboFold's purely in silico prediction not only is close to experimentally guided models for local structures, but also goes far beyond them by capturing the end-to-end pairs between 5' and 3' untranslated regions (UTRs) (∼29,800 nt apart) that match perfectly with a purely experimental work. Furthermore, LinearTurboFold identifies undiscovered conserved structures and conserved accessible regions as potential targets for designing efficient and mutation-insensitive small-molecule drugs, antisense oligonucleotides, small interfering RNAs (siRNAs), CRISPR-Cas13 guide RNAs, and RT-PCR primers. LinearTurboFold is a general technique that can also be applied to other RNA viruses and full-length genome studies and will be a useful tool in fighting the current and future pandemics.

LinearTurboFold: Linear-Time Global Prediction of Conserved Structures for RNA Homologs with Applications to SARS-CoV-2

The constant emergence of COVID-19 variants reduces the effectiveness of existing vaccines and test kits. Therefore, it is critical to identify conserved structures in SARS-CoV-2 genomes as potential targets for variant-proof diagnostics and therapeutics. However, the algorithms to predict these conserved structures, which simultaneously fold and align multiple RNA homologs, scale at best cubically with sequence length, and are thus infeasible for coronaviruses, which possess the longest genomes (∼30,000 nt ) among RNA viruses. As a result, existing efforts on modeling SARS-CoV-2 structures resort to single sequence folding as well as local folding methods with short window sizes, which inevitably neglect long-range interactions that are crucial in RNA functions. Here we present LinearTurboFold, an efficient algorithm for folding RNA homologs that scales linearly with sequence length, enabling unprecedented global structural analysis on SARS-CoV-2. Surprisingly, on a group of SARS-CoV-2 and SARS-related genomes, LinearTurbo-Fold's purely in silico prediction not only is close to experimentally-guided models for local structures, but also goes far beyond them by capturing the end-to-end pairs between 5' and 3' UTRs (∼29,800 nt apart) that match perfectly with a purely experimental work. Furthermore, LinearTurboFold identifies novel conserved structures and conserved accessible regions as potential targets for designing efficient and mutation-insensitive small-molecule drugs, antisense oligonucleotides, siRNAs, CRISPR-Cas13 guide RNAs and RT-PCR primers. LinearTurboFold is a general technique that can also be applied to other RNA viruses and full-length genome studies, and will be a useful tool in fighting the current and future pandemics.

SIGNIFICANCE STATEMENT

Conserved RNA structures are critical for designing diagnostic and therapeutic tools for many diseases including COVID-19. However, existing algorithms are much too slow to model the global structures of full-length RNA viral genomes. We present LinearTurboFold, a linear-time algorithm that is orders of magnitude faster, making it the first method to simultaneously fold and align whole genomes of SARS-CoV-2 variants, the longest known RNA virus (∼30 kilobases). Our work enables unprecedented global structural analysis and captures long-range interactions that are out of reach for existing algorithms but crucial for RNA functions. LinearTurboFold is a general technique for full-length genome studies and can help fight the current and future pandemics.

Review of machine learning methods for RNA secondary structure prediction

Secondary structure plays an important role in determining the function of noncoding RNAs. Hence, identifying RNA secondary structures is of great value to research. Computational prediction is a mainstream approach for predicting RNA secondary structure. Unfortunately, even though new methods have been proposed over the past 40 years, the performance of computational prediction methods has stagnated in the last decade. Recently, with the increasing availability of RNA structure data, new methods based on machine learning (ML) technologies, especially deep learning, have alleviated the issue. In this review, we provide a comprehensive overview of RNA secondary structure prediction methods based on ML technologies and a tabularized summary of the most important methods in this field. The current pending challenges in the field of RNA secondary structure prediction and future trends are also discussed.

An antisense noncoding RNA enhances translation via localized structural rearrangements of its cognate mRNA

A large portion of eukaryotic genes are associated with noncoding, natural antisense transcripts (NATs). Despite sharing extensive sequence complementarity with their sense mRNAs, mRNA-NAT pairs elusively often evade dsRNA-cleavage and siRNA-triggered silencing. More surprisingly, some NATs enhance translation of their sense mRNAs by yet unknown mechanism(s). Here, we show that translation enhancement of the rice (Oryza sativa) PHOSPHATE1.2 (PHO1.2) mRNA is enabled by specific structural rearrangements guided by its noncoding antisense RNA (cis-NATpho1.2). Their interaction in vitro revealed no evidence of widespread intermolecular dsRNA formation, but rather specific local changes in nucleotide base pairing, leading to higher flexibility of PHO1.2 mRNA at a key high guanine-cytosine�(GC) regulatory region inhibiting translation, ∼350-nt downstream of the start codon. Sense-antisense RNA interaction increased formation of the 80S complex in PHO1.2, possibly by inducing structural rearrangement within this inhibitory region, thus making this mRNA more accessible to 60S. This work presents a framework for nucleotide resolution studies of functional mRNA-antisense pairs.

Predicting dynamic cellular protein-RNA interactions by deep learning using in vivo RNA structures

Interactions with RNA-binding proteins (RBPs) are integral to RNA function and cellular regulation, and dynamically reflect specific cellular conditions. However, presently available tools for predicting RBP-RNA interactions employ RNA sequence and/or predicted RNA structures, and therefore do not capture their condition-dependent nature. Here, after profiling transcriptome-wide in vivo RNA secondary structures in seven cell types, we developed PrismNet, a deep learning tool that integrates experimental in vivo RNA structure data and RBP binding data for matched cells to accurately predict dynamic RBP binding in various cellular conditions. PrismNet results for 168 RBPs support its utility for both understanding CLIP-seq results and largely extending such interaction data to accurately analyze additional cell types. Further, PrismNet employs an "attention" strategy to computationally identify exact RBP-binding nucleotides, and we discovered enrichment among dynamic RBP-binding sites for structure-changing variants (riboSNitches), which can link genetic diseases with dysregulated RBP bindings. Our rich profiling data and deep learning-based prediction tool provide access to a previously inaccessible layer of cell-type-specific RBP-RNA interactions, with clear utility for understanding and treating human diseases.

A novel end-to-end method to predict RNA secondary structure profile based on bidirectional LSTM and residual neural network

Background

Studies have shown that RNA secondary structure, a planar structure formed by paired bases, plays diverse vital roles in fundamental life activities and complex diseases. RNA secondary structure profile can record whether each base is paired with others. Hence, accurate prediction of secondary structure profile can help to deduce the secondary structure and binding site of RNA. RNA secondary structure profile can be obtained through biological experiment and calculation methods. Of them, the biological experiment method involves two ways: chemical reagent and biological crystallization. The chemical reagent method can obtain a large number of prediction data, but its cost is high and always associated with high noise, making it difficult to get results of all bases on RNA due to the limited of sequencing coverage. By contrast, the biological crystallization method can lead to accurate results, yet heavy experimental work and high costs are required. On the other hand, the calculation method is CROSS, which comprises a three-layer fully connected neural network. However, CROSS can not completely learn the features of RNA secondary structure profile since its poor network structure, leading to its low performance.

Results

In this paper, a novel end-to-end method, named as “RPRes, was proposed to predict RNA secondary structure profile based on Bidirectional LSTM and Residual Neural Network.

Conclusions

RPRes utilizes data sets generated by multiple biological experiment methods as the training, validation, and test sets to predict profile, which can compatible with numerous prediction requirements. Compared with the biological experiment method, RPRes has reduced the costs and improved the prediction efficiency. Compared with the state-of-the-art calculation method CROSS, RPRes has significantly improved performance.

Inverse RNA Folding Workflow to Design and Test Ribozymes that Include Pseudoknots

Ribozymes are RNAs that catalyze reactions. They occur in nature, and can also be evolved in vitro to catalyze novel reactions. This chapter provides detailed protocols for using inverse folding software to design a ribozyme sequence that will fold to a known ribozyme secondary structure and for testing the catalytic activity of the sequence experimentally. This protocol is able to design sequences that include pseudoknots, which is important as all naturally occurring full-length ribozymes have pseudoknots. The starting point is the known pseudoknot-containing secondary structure of the ribozyme and knowledge of any nucleotides whose identity is required for function. The output of the protocol is a set of sequences that have been tested for function. Using this protocol, we were previously successful at designing highly active double-pseudoknotted HDV ribozymes.

Deep Learning Method for RNA Secondary Structure Prediction with Pseudoknots Based on Large-Scale Data

Traditional machine learning methods are widely used in the field of RNA secondary structure prediction and have achieved good results. However, with the emergence of large-scale data, deep learning methods have more advantages than traditional machine learning methods. As the number of network layers increases in deep learning, there will often be problems such as increased parameters and overfitting. We used two deep learning models, GoogLeNet and TCN, to predict RNA secondary results. And from the perspective of the depth and width of the network, improvements are made based on the neural network model, which can effectively improve the computational efficiency while extracting more feature information. We process the existing real RNA data through experiments, use deep learning models to extract useful features from a large amount of RNA sequence data and structure data, and then predict the extracted features to obtain each base’s pairing probability. The characteristics of RNA secondary structure and dynamic programming methods are used to process the base prediction results, and the structure with the largest sum of the probability of each base pairing is obtained, and this structure will be used as the optimal RNA secondary structure. We, respectively, evaluated GoogLeNet and TCN models based on 5sRNA, tRNA data, and tmRNA data, and compared them with other standard prediction algorithms. The sensitivity and specificity of the GoogLeNet model on the 5sRNA and tRNA data sets are about 16% higher than the best prediction results in other algorithms. The sensitivity and specificity of the GoogLeNet model on the tmRNA dataset are about 9% higher than the best prediction results in other algorithms. As deep learning algorithms’ performance is related to the size of the data set, as the scale of RNA data continues to expand, the prediction accuracy of deep learning methods for RNA secondary structure will continue to improve.

Hierarchical natural move Monte Carlo refines flexible RNA structures into cryo-EM densities

Ribonucleic acids (RNAs) play essential roles in living cells. Many of them fold into defined three-dimensional (3D) structures to perform functions. Recent advances in single-particle cryo-electron microscopy (cryo-EM) have enabled structure determinations of RNA to atomic resolutions. However, most RNA molecules are structurally flexible, limiting the resolution of their structures solved by cryo-EM. In modeling these molecules, several computational methods are limited by the requirement of massive computational resources and/or the low efficiency in exploring large-scale structural variations. Here we use hierarchical natural move Monte Carlo (HNMMC), which takes advantage of collective motions for groups of nucleic acid residues, to refine RNA structures into their cryo-EM maps, preserving atomic details in the models. After validating the method on a simulated density map of tRNA, we applied it to objectively obtain the model of the folding intermediate for the specificity domain of ribonuclease P from Bacillus subtilis and refine a flexible ribosomal RNA (rRNA) expansion segment from the Mycobacterium tuberculosis (Mtb) ribosome in different conformational states. Finally, we used HNMMC to model atomic details and flexibility for two distinct conformations of the complete genomic RNA (gRNA) inside MS2, a single-stranded RNA virus, revealing multiple pathways for its capsid assembly.

Small molecule-RNA targeting: starting with the fundamentals

The structural and regulatory elements in therapeutically relevant RNAs offer many opportunities for targeting by small molecules, yet fundamental understanding of what drives selectivity in small molecule:RNA recognition has been a recurrent challenge. In particular, RNAs tend to be more dynamic and offer less chemical functionality than proteins, and biologically active ligands must compete with the highly abundant and highly structured RNA of the ribosome. Indeed, the only small molecule drug targeting RNA other than the ribosome was just approved in August 2020, and our recent survey of the literature revealed fewer than 150 reported chemical probes that target non-ribosomal RNA in biological systems. This Feature outlines our efforts to improve small molecule targeting strategies and gain fundamental insights into small molecule:RNA recognition by analyzing patterns in both RNA-biased small molecule chemical space and RNA topological space privileged for differentiation. First, we synthesized libraries based on RNA binding scaffolds that allowed us to reveal general principles in small molecule:recognition and to ask precise chemical questions about drivers of affinity and selectivity. Elaboration of these scaffolds has led to recognition of medicinally relevant RNA targets, including viral and long noncoding RNA structures. More globally, we identified physicochemical, structural, and spatial properties of biologically active RNA ligands that are distinct from those of protein-targeted ligands, and we have provided the dataset and associated analytical tools as part of a publicly available online platform to facilitate RNA ligand discovery. At the same time, we used pattern recognition protocols to identify RNA topologies that can be differentially recognized by small molecules and have elaborated this technique to visualize conformational changes in RNA secondary structure. These fundamental insights into the drivers of RNA recognition in vitro have led to functional targeting of RNA structures in biological systems. We hope that these initial guiding principles, as well as the approaches and assays developed in their pursuit, will enable rapid progress toward the development of RNA-targeted chemical probes and ultimately new therapeutic approaches to a wide range of deadly human diseases.

Fast and accurate structure probability estimation for simultaneous alignment and folding of RNAs with Markov chains

MOTIVATION

Simultaneous alignment and folding (SA&F) of RNAs is the indispensable gold standard for inferring the structure of non-coding RNAs and their general analysis. The original algorithm, proposed by Sankoff, solves the theoretical problem exactly with a complexity of [Formula: see text] in the full energy model. Over the last two decades, several variants and improvements of the Sankoff algorithm have been proposed to reduce its extreme complexity by proposing simplified energy models or imposing restrictions on the predicted alignments.

RESULTS

Here, we introduce a novel variant of Sankoff's algorithm that reconciles the simplifications of PMcomp, namely moving from the full energy model to a simpler base pair-based model, with the accuracy of the loop-based full energy model. Instead of estimating pseudo-energies from unconditional base pair probabilities, our model calculates energies from conditional base pair probabilities that allow to accurately capture structure probabilities, which obey a conditional dependency. This model gives rise to the fast and highly accurate novel algorithm Pankov (Probabilistic Sankoff-like simultaneous alignment and folding of RNAs inspired by Markov chains).

CONCLUSIONS

Pankov benefits from the speed-up of excluding unreliable base-pairing without compromising the loop-based free energy model of the Sankoff's algorithm. We show that Pankov outperforms its predecessors LocARNA and SPARSE in folding quality and is faster than LocARNA.

From structure to function: Route to understanding lncRNA mechanism

RNAs have emerged as a major target for diagnostics and therapeutics approaches. Regulatory nonprotein-coding RNAs (ncRNAs) in particular display remarkable versatility. They can fold into complex structures and interact with proteins, DNA, and other RNAs, thus modulating activity, localization, or interactome of multi-protein complexes. Thus, ncRNAs confer regulatory plasticity and represent a new layer of regulatory control. Interestingly, long noncoding RNAs (lncRNAs) tend to acquire complex secondary and tertiary structures and their function-in many cases-is dependent on structural conservation rather than primary sequence conservation. Whereas for many proteins, structure and its associated function are closely connected, for lncRNAs, the structural domains that determine functionality and its interactome are still not well understood. Numerous approaches for analyzing the structural configuration of lncRNAs have been developed recently. Here, will provide an overview of major experimental approaches used in the field, and discuss the potential benefit of using combinatorial strategies to analyze lncRNA modes of action based on structural information.

Topological constraints of RNA pseudoknotted and loop-kissing motifs: applications to three-dimensional structure prediction

An RNA global fold can be described at the level of helix orientations and relatively flexible loop conformations that connect the helices. The linkage between the helices plays an essential role in determining the structural topology, which restricts RNA local and global folds, especially for RNA tertiary structures involving cross-linked base pairs. We quantitatively analyze the topological constraints on RNA 3D conformational space, in particular, on the distribution of helix orientations, for pseudoknots and loop-loop kissing structures. The result shows that a viable conformational space is predominantly determined by the motif type, helix size, and loop size, indicating a strong topological coupling between helices and loops in RNA tertiary motifs. Moreover, the analysis indicates that (cross-linked) tertiary contacts can cause much stronger topological constraints on RNA global fold than non-cross-linked base pairs. Furthermore, based on the topological constraints encoded in the 2D structure and the 3D templates, we develop a 3D structure prediction approach. This approach can be further combined with structure probing methods to expand the capability of computational prediction for large RNA folds.

Free-Energy Calculation of Ribonucleic Inosines and Its Application to Nearest-Neighbor Parameters

Can current simulations quantitatively predict the stability of ribonucleic acids (RNAs)? In this research, we apply a free-energy perturbation simulation of RNAs containing inosine, a modified ribonucleic base, to the derivation of RNA nearest-neighbor parameters. A parameter set derived solely from 30 simulations was used to predict the free-energy difference of the RNA duplex with a mean unbiased error of 0.70 kcal/mol, which is a level of accuracy comparable to that obtained with parameters derived from 25 experiments. We further show that the error can be lowered to 0.60 kcal/mol by combining the simulation-derived free-energy differences with experimentally measured differences. This protocol can be used as a versatile method for deriving nearest-neighbor parameters of RNAs with various modified bases.

Accurate inference of the full base-pairing structure of RNA by deep mutational scanning and covariation-induced deviation of activity

Despite the large number of noncoding RNAs in human genome and their roles in many diseases include cancer, we know very little about them due to lack of structural clues. The centerpiece of the structural clues is the full RNA base-pairing structure of secondary and tertiary contacts that can be precisely obtained only from costly and time-consuming 3D structure determination. Here, we performed deep mutational scanning of self-cleaving CPEB3 ribozyme by error-prone PCR and showed that a library of <5 × 104 single-to-triple mutants is sufficient to infer 25 of 26 base pairs including non-nested, nonhelical, and noncanonical base pairs with both sensitivity and precision at 96%. Such accurate inference was further confirmed by a twister ribozyme at 100% precision with only noncanonical base pairs as false negatives. The performance was resulted from analyzing covariation-induced deviation of activity by utilizing both functional and nonfunctional variants for unsupervised classification, followed by Monte Carlo (MC) simulated annealing with mutation-derived scores. Highly accurate inference can also be obtained by combining MC with evolution/direct coupling analysis, R-scape or epistasis analysis. The results highlight the usefulness of deep mutational scanning for high-accuracy structural inference of self-cleaving ribozymes with implications for other structured RNAs that permit high-throughput functional selections.

RNA secondary structure prediction using an ensemble of two-dimensional deep neural networks and transfer learning

The majority of our human genome transcribes into noncoding RNAs with unknown structures and functions. Obtaining functional clues for noncoding RNAs requires accurate base-pairing or secondary-structure prediction. However, the performance of such predictions by current folding-based algorithms has been stagnated for more than a decade. Here, we propose the use of deep contextual learning for base-pair prediction including those noncanonical and non-nested (pseudoknot) base pairs stabilized by tertiary interactions. Since only [Formula: see text]250 nonredundant, high-resolution RNA structures are available for model training, we utilize transfer learning from a model initially trained with a recent high-quality bpRNA dataset of [Formula: see text]10,000 nonredundant RNAs made available through comparative analysis. The resulting method achieves large, statistically significant improvement in predicting all base pairs, noncanonical and non-nested base pairs in particular. The proposed method (SPOT-RNA), with a freely available server and standalone software, should be useful for improving RNA structure modeling, sequence alignment, and functional annotations.

Sequence-dependent RNA helix conformational preferences predictably impact tertiary structure formation

Structured RNAs and RNA complexes underlie biological processes ranging from control of gene expression to protein translation. Approximately 50% of nucleotides within known structured RNAs are folded into Watson-Crick (WC) base pairs, and sequence changes that preserve these pairs are typically assumed to preserve higher-order RNA structure and binding of macromolecule partners. Here, we report that indirect effects of the helix sequence on RNA tertiary stability are, in fact, significant but are nevertheless predictable from a simple computational model called RNAMake-∆∆G. When tested through the RNA on a massively parallel array (RNA-MaP) experimental platform, blind predictions for >1500 variants of the tectoRNA heterodimer model system achieve high accuracy (rmsd 0.34 and 0.77 kcal/mol for sequence and length changes, respectively). Detailed comparison of predictions to experiments support a microscopic picture of how helix sequence changes subtly modulate conformational fluctuations at each base-pair step, which accumulate to impact RNA tertiary structure stability. Our study reveals a previously overlooked phenomenon in RNA structure formation and provides a framework of computation and experiment for understanding helix conformational preferences and their impact across biological RNA and RNA-protein assemblies.

High-throughput determination of RNA structures

RNA performs and regulates a diverse range of cellular processes, with new functional roles being uncovered at a rapid pace. Interest is growing in how these functions are linked to RNA structures that form in the complex cellular environment. A growing suite of technologies that use advances in RNA structural probes, high-throughput sequencing and new computational approaches to interrogate RNA structure at unprecedented throughput are beginning to provide insights into RNA structures at new spatial, temporal and cellular scales.

LinearFold: linear-time approximate RNA folding by 5'-to-3' dynamic programming and beam search

MOTIVATION

Predicting the secondary structure of an ribonucleic acid (RNA) sequence is useful in many applications. Existing algorithms [based on dynamic programming] suffer from a major limitation: their runtimes scale cubically with the RNA length, and this slowness limits their use in genome-wide applications.

RESULTS

We present a novel alternative O(n3)-time dynamic programming algorithm for RNA folding that is amenable to heuristics that make it run in O(n) time and O(n) space, while producing a high-quality approximation to the optimal solution. Inspired by incremental parsing for context-free grammars in computational linguistics, our alternative dynamic programming algorithm scans the sequence in a left-to-right (5'-to-3') direction rather than in a bottom-up fashion, which allows us to employ the effective beam pruning heuristic. Our work, though inexact, is the first RNA folding algorithm to achieve linear runtime (and linear space) without imposing constraints on the output structure. Surprisingly, our approximate search results in even higher overall accuracy on a diverse database of sequences with known structures. More interestingly, it leads to significantly more accurate predictions on the longest sequence families in that database (16S and 23S Ribosomal RNAs), as well as improved accuracies for long-range base pairs (500+ nucleotides apart), both of which are well known to be challenging for the current models.

AVAILABILITY AND IMPLEMENTATION

Our source code is available at https://github.com/LinearFold/LinearFold, and our webserver is at http://linearfold.org (sequence limit: 100 000nt).

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

RNApolis: Computational Platform for RNA Structure Analysis

In the 1970s, computer scientists began to engage in research in the field of structural biology. The first structural databases, as well as models and methods supporting the analysis of biomolecule structures, started to be created. RNA was put at the centre of scientific interest quite late. However, more and more methods dedicated to this molecule are currently being developed. This paper presents RNApolis - a new computing platform, which offers access to seven bioinformatic tools developed to support the RNA structure study. The set of tools include a structural database and systems for predicting, modelling, annotating and evaluating the RNA structure. RNApolis supports research at different structural levels and allows the discovery, establishment, and validation of relationships between the primary, secondary and tertiary structure of RNAs. The platform is freely available at http://rnapolis.pl

Estimating uncertainty in predicted folding free energy changes of RNA secondary structures

Nearest neighbor parameters for estimating the folding stability of RNA are commonly used in secondary structure prediction, for generating folding ensembles of structures, and for analyzing RNA function. Previously, we demonstrated that we could quantify the uncertainties in each nearest neighbor parameter by perturbing the underlying optical melting data within experimental error and rederiving the parameters, which accounts for the substantial correlations that exist between the parameters. In this contribution, we describe a method to estimate uncertainty in the estimated folding stabilities of RNA structures, accounting for correlations in the nearest neighbor parameters. This method is incorporated in the RNA structure software package.

Pairwise visual comparison of small RNA secondary structures with base pair probabilities

BACKGROUND

Predicted RNA secondary structures are typically visualized using dot-plots for base pair binding probabilities and planar graphs for unique structures, such as the minimum free energy structure. These are however difficult to analyze simultaneously.

RESULTS

This work introduces a compact unified view of the most stable conformation of an RNA secondary structure and its base pair probabilities, which is called the Circular Secondary Structure Base Pairs Probabilities Plot (CS2BP2-Plot). Along with our design we provide access to a web server implementation of our solution that facilitates pairwise comparison of short RNA (and DNA) sequences up to 200 base pairs. The web server first calculates the minimum free energy secondary structure and the base pair probabilities for up to 10 RNA or DNA sequences using RNAfold and then provides a two panel comparative view that includes CS2BP2-Plots along with the traditional graph, planar and circular diagrams obtained with VARNA. The CS2BP2-Plots include highlighting of the nucleotide differences between two selected sequences using ClustalW local alignments. We also provide descriptive statistics, dot-bracket secondary structure representations and ClustalW local alignments for compared sequences.

CONCLUSIONS

Using circular diagrams and colour and weight-coded arcs, we demonstrate how a single image can replace the state-of-the-art dual representations (dot-plots and minimum free energy structures) for base-pair probabilities of RNA secondary structures while allowing efficient exploration and comparison of different RNA conformations via a web server front end. With that, we provide the community, especially the biologically oriented, with an intuitive tool for ncRNA visualization. Web-server: https://nrcmonsrv01.nrc.ca/cs2bp2plot.

How to benchmark RNA secondary structure prediction accuracy

RNA secondary structure prediction is widely used. As new methods are developed, these are often benchmarked for accuracy against existing methods. This review discusses good practices for performing these benchmarks, including the choice of benchmarking structures, metrics to quantify accuracy, the importance of allowing flexibility for pairs in the accepted structure, and the importance of statistical testing for significance.

Advances in engineered trans-acting regulatory RNAs and their application in bacterial genome engineering

Small noncoding RNAs, a large class of ancient posttranscriptional regulators, are increasingly recognized and utilized as key modulators of gene expression in a broad range of microorganisms. Owing to their small molecular size and the central role of Watson-Crick base pairing in defining their interactions, structure and function, numerous diverse types of trans-acting RNA regulators that are functional at the DNA, mRNA and protein levels have been experimentally characterized. It has become increasingly clear that most small RNAs play critical regulatory roles in many processes and are, therefore, considered to be powerful tools for genetic engineering and synthetic biology. The trans-acting regulatory RNAs accelerate this ability to establish potential framework for genetic engineering and genome-scale engineering, which allows RNA structure characterization, easier to design and model compared to DNA or protein-based systems. In this review, we summarize recent advances in engineered trans-acting regulatory RNAs that are used in bacterial genome-scale engineering and in novel cellular capabilities as well as their implementation in wide range of biotechnological, biological and medical applications.

DMfold: A Novel Method to Predict RNA Secondary Structure With Pseudoknots Based on Deep Learning and Improved Base Pair Maximization Principle

While predicting the secondary structure of RNA is vital for researching its function, determining RNA secondary structure is challenging, especially for that with pseudoknots. Typically, several excellent computational methods can be utilized to predict the secondary structure (with or without pseudoknots), but they have their own merits and demerits. These methods can be classified into two categories: the multi-sequence method and the single-sequence method. The main advantage of the multi-sequence method lies in its use of the auxiliary sequences to assist in predicting the secondary structure, but it can only successfully predict in the presence of multiple highly homologous sequences. The single-sequence method is associated with the major merit of easy operation (only need the target sequence to predict secondary structure), but its folding parameters are the common features of diversity RNA, which cannot describe the unique characteristics of RNA, thus potentially resulting in the low prediction accuracy in some RNA. In this paper, "DMfold," a method based on the Deep Learning and Improved Base Pair Maximization Principle, is proposed to predict the secondary structure with pseudoknots, which fully absorbs the advantages and avoids some disadvantages of those two methods. Notably, DMfold could predict the secondary structure of RNA by learning similar RNA in the known structures, which uses the similar RNA sequences instead of the highly homogeneous sequences in the multi-sequence method, thereby reducing the requirement for auxiliary sequences. In DMfold, it only needs to input the target sequence to predict the secondary structure. Its folding parameters are fully extracted automatically by deep learning, which could avoid the lack of folding parameters in the single-sequence method. Experiments show that our method is not only simple to operate, but also improves the prediction accuracy compared to multiple excellent prediction methods. A repository containing our code can be found at https://github.com/linyuwangPHD/RNA-Secondary-Structure-Database.

RNA Structural Differentiation: Opportunities with Pattern Recognition

Our awareness and appreciation of the many regulatory roles of RNA have dramatically increased in the past decade. This understanding, in addition to the impact of RNA in many disease states, has renewed interest in developing selective RNA-targeted small molecule probes. However, the fundamental guiding principles in RNA molecular recognition that could accelerate these efforts remain elusive. While high-resolution structural characterization can provide invaluable insight, examples of well-characterized RNA structures, not to mention small molecule:RNA complexes, remain limited. This Perspective provides an overview of the current techniques used to understand RNA molecular recognition when high-resolution structural information is unavailable. We will place particular emphasis on a new method, pattern recognition of RNA with small molecules (PRRSM), that provides rapid insight into critical components of RNA recognition and differentiation by small molecules as well as into RNA structural features.

Bioinformatics Tools and Benchmarks for Computational Docking and 3D Structure Prediction of RNA-Protein Complexes

RNA-protein (RNP) interactions play essential roles in many biological processes, such as regulation of co-transcriptional and post-transcriptional gene expression, RNA splicing, transport, storage and stabilization, as well as protein synthesis. An increasing number of RNP structures would aid in a better understanding of these processes. However, due to the technical difficulties associated with experimental determination of macromolecular structures by high-resolution methods, studies on RNP recognition and complex formation present significant challenges. As an alternative, computational prediction of RNP interactions can be carried out. Structural models obtained by theoretical predictive methods are, in general, less reliable compared to models based on experimental measurements but they can be sufficiently accurate to be used as a basis for to formulating functional hypotheses. In this article, we present an overview of computational methods for 3D structure prediction of RNP complexes. We discuss currently available methods for macromolecular docking and for scoring 3D structural models of RNP complexes in particular. Additionally, we also review benchmarks that have been developed to assess the accuracy of these methods.

Evaluation of global HIV/SIV envelope gp120 RNA structure and evolution within and among infected hosts

Lentiviral RNA genomes contain structural elements that play critical roles in viral replication. Although structural features of 5'-untranslated regions have been well characterized, attempts to identify important structures in other genomic regions by Selective 2'-Hydroxyl Acylation analyzed by Primer Extension (SHAPE) have led to conflicting structural and mechanistic conclusions. Previous approaches accounted neither for sequence heterogeneity that is ubiquitous in viral populations, nor for selective constraints operating at the protein level. We developed an approach that augments SHAPE with phylogenetic analyses and applied it to investigate structure in coding regions (cRNA) within the HIV and SIV envelope genes. Analysis of single-genome SHAPE data with phylogenetic information from diverse lentiviral sequences argues against the conservation of a putative global gp120 RNA structure but points to the existence of core RNA sub-structures. Our findings establish a framework for considering sequence heterogeneity and protein function in de novo RNA structure inference approaches.

Amino acid repeats avert mRNA folding through conservative substitutions and synonymous codons, regardless of codon bias

A significant number of proteins in all living species contains amino acid repeats (AARs) of various lengths and compositions, many of which play important roles in protein structure and function. Here, I have surveyed select homopolymeric single [(A)n] and double [(AB)n] AARs in the human proteome. A close examination of their codon pattern and analysis of RNA structure propensity led to the following set of empirical rules: (1) One class of amino acid repeats (Class I) uses a mixture of synonymous codons, some of which approximate the codon bias ratio in the overall human proteome; (2) The second class (Class II) disregards the codon bias ratio, and appears to have originated by simple repetition of the same codon (or just a few codons); and finally, (3) In all AARs (including Class I, Class II, and the in-betweens), the codons are chosen in a manner that precludes the formation of RNA secondary structure. It appears that the AAR genes have evolved by orchestrating a balance between codon usage and mRNA secondary structure. The insights gained here should provide a better understanding of AAR evolution and may assist in designing synthetic genes.

TurboFold II: RNA structural alignment and secondary structure prediction informed by multiple homologs

This paper presents TurboFold II, an extension of the TurboFold algorithm for predicting secondary structures for multiple RNA homologs. TurboFold II augments the structure prediction capabilities of TurboFold by additionally providing multiple sequence alignments. Probabilities for alignment of nucleotide positions between all pairs of input sequences are iteratively estimated in TurboFold II by incorporating information from both the sequence identity and secondary structures. A multiple sequence alignment is obtained from these probabilities by using a probabilistic consistency transformation and a hierarchically computed guide tree. To assess TurboFold II, its sequence alignment and structure predictions were compared with leading tools, including methods that focus on alignment alone and methods that provide both alignment and structure prediction. TurboFold II has comparable alignment accuracy with MAFFT and higher accuracy than other tools. TurboFold II also has comparable structure prediction accuracy as the original TurboFold algorithm, which is one of the most accurate methods. TurboFold II is part of the RNAstructure software package, which is freely available for download at http://rna.urmc.rochester.edu under a GPL license.

An RNA structure-mediated, posttranscriptional model of human α-1-antitrypsin expression

Protein and mRNA expression are in most cases poorly correlated, which suggests that the posttranscriptional regulatory program of a cell is an important component of gene expression. This regulatory network is still poorly understood, including how RNA structure quantitatively contributes to translational control. We present here a series of structural and functional experiments that together allow us to derive a quantitative, structure-dependent model of translation that accurately predicts translation efficiency in reporter assays and primary human tissue for a complex and medically important protein, α-1-antitrypsin. Our model demonstrates the importance of accurate, experimentally derived RNA structural models partnered with Kozak sequence information to explain protein expression and suggests a strategy by which α-1-antitrypsin expression may be increased in diseased individuals.

Chronic obstructive pulmonary disease (COPD) affects over 65 million individuals worldwide, where α-1-antitrypsin deficiency is a major genetic cause of the disease. The α-1-antitrypsin gene, SERPINA1, expresses an exceptional number of mRNA isoforms generated entirely by alternative splicing in the 5′-untranslated region (5′-UTR). Although all SERPINA1 mRNAs encode exactly the same protein, expression levels of the individual mRNAs vary substantially in different human tissues. We hypothesize that these transcripts behave unequally due to a posttranscriptional regulatory program governed by their distinct 5′-UTRs and that this regulation ultimately determines α-1-antitrypsin expression. Using whole-transcript selective 2′-hydroxyl acylation by primer extension (SHAPE) chemical probing, we show that splicing yields distinct local 5′-UTR secondary structures in SERPINA1 transcripts. Splicing in the 5′-UTR also changes the inclusion of long upstream ORFs (uORFs). We demonstrate that disrupting the uORFs results in markedly increased translation efficiencies in luciferase reporter assays. These uORF-dependent changes suggest that α-1-antitrypsin protein expression levels are controlled at the posttranscriptional level. A leaky-scanning model of translation based on Kozak translation initiation sequences alone does not adequately explain our quantitative expression data. However, when we incorporate the experimentally derived RNA structure data, the model accurately predicts translation efficiencies in reporter assays and improves α-1-antitrypsin expression prediction in primary human tissues. Our results reveal that RNA structure governs a complex posttranscriptional regulatory program of α-1-antitrypsin expression. Crucially, these findings describe a mechanism by which genetic alterations in noncoding gene regions may result in α-1-antitrypsin deficiency.

Structural signatures of thermal adaptation of bacterial ribosomal RNA, transfer RNA, and messenger RNA

Temperature adaptation of bacterial RNAs is a subject of both fundamental and practical interest because it will allow a better understanding of molecular mechanism of RNA folding with potential industrial application of functional thermophilic or psychrophilic RNAs. Here, we performed a comprehensive study of rRNA, tRNA, and mRNA of more than 200 bacterial species with optimal growth temperatures (OGT) ranging from 4°C to 95°C. We investigated temperature adaptation at primary, secondary and tertiary structure levels. We showed that unlike mRNA, tRNA and rRNA were optimized for their structures at compositional levels with significant tertiary structural features even for their corresponding randomly permutated sequences. tRNA and rRNA are more exposed to solvent but remain structured for hyperthermophiles with nearly OGT-independent fluctuation of solvent accessible surface area within a single RNA chain. mRNA in hyperthermophiles is essentially the same as random sequences without tertiary structures although many mRNA in mesophiles and psychrophiles have well-defined tertiary structures based on their low overall solvent exposure with clear separation of deeply buried from partly exposed bases as in tRNA and rRNA. These results provide new insight into temperature adaptation of different RNAs.