Secondary Structure Prediction Research Articles

Direct Inference of Base-Pairing Probabilities with Neural Networks Improves Prediction of RNA Secondary Structures with Pseudoknots.

Existing approaches to predicting RNA secondary structures depend on how the secondary structure is decomposed into substructures, that is, the architecture, to define their parameter space. However, architecture dependency has not been sufficiently investigated, especially for pseudoknotted secondary structures. In this study, we propose a novel algorithm for directly inferring base-pairing probabilities with neural networks that do not depend on the architecture of RNA secondary structures, and then implement this approach using two maximum expected accuracy (MEA)-based decoding algorithms: Nussinov-style decoding for pseudoknot-free structures and IPknot-style decoding for pseudoknotted structures. To train the neural networks connected to each base pair, we adopt a max-margin framework, called structured support vector machines (SSVM), as the output layer. Our benchmarks for predicting RNA secondary structures with and without pseudoknots show that our algorithm outperforms existing methods in prediction accuracy.

SARS-CoV-2 Variants of Concern and Variations within Their Genome Architecture: Does Nucleotide Distribution and Mutation Rate Alter the Functionality and Evolution of the Virus?

SARS-CoV-2 virus pathogenicity and transmissibility are correlated with the mutations acquired over time, giving rise to variants of concern (VOCs). Mutations can significantly influence the genetic make-up of the virus. Herein, we analyzed the SARS-CoV-2 genomes and sub-genomic nucleotide composition in relation to the mutation rate. Nucleotide percentage distributions of 1397 in-house-sequenced SARS-CoV-2 genomes were enumerated, and comparative analyses (i) within the VOCs and of (ii) recovered and mortality patients were performed. Fisher's test was carried out to highlight the significant mutations, followed by RNA secondary structure prediction and protein modeling for their functional impacts. Subsequently, a uniform dinucleotide composition of AT and GC was found across study cohorts. Notably, the N gene was observed to have a high GC percentage coupled with a relatively higher mutation rate. Functional analysis demonstrated the N gene mutations, C29144T and G29332T, to induce structural changes at the RNA level. Protein secondary structure prediction with N gene missense mutations revealed a differential composition of alpha helices, beta sheets, and coils, whereas the tertiary structure displayed no significant changes. Additionally, the N gene CTD region displayed no mutations. The analysis highlighted the importance of N protein in viral evolution with CTD as a possible target for antiviral drugs.

Phylogenetic inference and secondary structure predictions of Turkish genotypes of Coriandrum sativum (L.) based on ITS4 and ITS5 nrDNA sequences

Phylogenetic inference and secondary structure predictions of Turkish genotypes of Coriandrum sativum (L.) based on ITS4 and ITS5 nrDNA sequences

NLRscape: an atlas of plant NLR proteins.

NLRscape is a webserver that curates a collection of over 80 000 plant protein sequences identified in UniProtKB to contain NOD-like receptor signatures, and hosts in addition a number of tools aimed at the exploration of the complex sequence landscape of this class of plant proteins. Each entry gathers sequence information, domain and motif annotations from multiple third-party sources but also in-house advanced annotations aimed at addressing caveats of the existing broad-based annotations. NLRscape provides a top-down perspective of the NLR sequence landscape but also services for assisting a bottom-up approach starting from a given input sequence. Sequences are clustered by their domain organization layout, global homology and taxonomic spread-in order to allow analysis of how particular traits of an NLR family are scattered within the plant kingdom. Tools are provided for users to locate their own protein of interest in the overall NLR landscape, generate custom clusters centered around it and perform a large number of sequence and structural analyses using included interactive online instruments. Amongst these, we mention: taxonomy distribution plots, homology cluster graphs, identity matrices and interactive MSA synchronizing secondary structure and motif predictions. NLRscape can be found at: https://nlrscape.biochim.ro/.

Protein-DNA Binding Residue Prediction via Bagging Strategy and Sequence-Based Cube-Format Feature.

Protein-DNA interactions play an important role in diverse biological processes. Accurately identifying protein-DNA binding residues is a critical but challenging task for protein function annotations and drug design. Although wet-lab experimental methods are the most accurate way to identify protein-DNA binding residues, they are time consuming and labor intensive. There is an urgent need to develop computational methods to rapidly and accurately predict protein-DNA binding residues. In this study, we propose a novel sequence-based method, named PredDBR, for predicting DNA-binding residues. In PredDBR, for each query protein, its position-specific frequency matrix (PSFM), predicted secondary structure (PSS), and predicted probabilities of ligand-binding residues (PPLBR) are first generated as three feature sources. Secondly, for each feature source, the sliding window technique is employed to extract the matrix-format feature of each residue. Then, we design two strategies, i.e., square root (SR) and average (AVE), to separately transform PSFM-based and two predicted feature source-based, i.e., PSS-based and PPLBR-based, matrix-format features of each residue into three corresponding cube-format features. Finally, after serially combining the three cube-format features, the ensemble classifier is generated via applying bagging strategy to multiple base classifiers built by the framework of 2D convolutional neural network. The computational experimental results demonstrate that the proposed PredDBR achieves an average overall accuracy of 93.7% and a Mathew's correlation coefficient of 0.405 on two independent validation datasets and outperforms several state-of-the-art sequenced-based protein-DNA binding residue predictors. The PredDBR web-server is available at https://jun-csbio.github.io/PredDBR/.

Oligomerization-Dependent Beta-Structure Formation in SARS-CoV-2 Envelope Protein.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for the current COVID-19 pandemic. In SARS-CoV-2, the channel-forming envelope (E) protein is almost identical to the E protein in SARS-CoV, and both share an identical α-helical channel-forming domain. Structures for the latter are available in both detergent and lipid membranes. However, models of the extramembrane domains have only been obtained from solution NMR in detergents, and show no β-strands, in contrast to secondary-structure predictions. Herein, we have studied the conformation of purified SARS-CoV-2 E protein in lipid bilayers that mimic the composition of ER-Golgi intermediate compartment (ERGIC) membranes. The full-length E protein at high protein-to-lipid ratios produced a clear shoulder at 1635 cm-1, consistent with the β-structure, but this was absent when the E protein was diluted, which instead showed a band at around 1688 cm-1, usually assigned to β-turns. The results were similar with a mixture of POPC:POPG (2-oleoyl-1-palmitoyl-sn-glycero-3-phosphocholine/3-glycerol) and also when using an E-truncated form (residues 8-65). However, the latter only showed β-structure formation at the highest concentration tested, while having a weaker oligomerization tendency in detergents than in full-length E protein. Therefore, we conclude that E monomer-monomer interaction triggers formation of the β-structure from an undefined structure (possibly β-turns) in at least about 15 residues located at the C-terminal extramembrane domain. Due to its proximity to the channel, this β-structure domain could modulate channel activity or modify membrane structure at the time of virion formation inside the cell.

Formula omitted]-HMM and optimal decoding higher-order structures on sequential data

Decoding higher-order structure on sequential data is an indispensable task in data science. It requires models to have the capability to characterize interdependencies among hidden events that have generated observable data. However, to be able to decode arbitrary structures, such models would need to cope with the intractability arising from computing context-sensitive relations, likely compromising the quality of answers. To address this important issue, the current paper introduces the arbitrary order hidden Markov model (α-HMM), an extension of the HMM that permits decoding of the optimal higher-order structure with an assurance of computational tractability. The advantage of the α-HMMis made possible by an identified principle on how random variables influence each other in a stochastic process. In particular, it is shown that decoding the optimal structure with an α-HMMcan be computed in O(n3)-time for any stochastic process of n random variables. As an application, it is demonstrated the decoding algorithm inspires a simple yet effective algorithm for RNA secondary structure prediction.

Effects of gentamicin inducing readthrough premature stop Codons: A study of alpha-L-iduronidase nonsense variants in COS-7 Cells

Mucopolysaccharidosis type I Hurler syndrome (MPS IH) is a severe lysosomal storage disorder caused by alpha-l-iduronidase (IDUA) deficiency. Premature truncation mutations (PTC) are the most common (50%–70%) type of IDUA mutations and correlate with MPS IH. Nonsense suppression therapy is a therapeutic approach that aims to induce stop codon readthrough. The different ability of gentamicin to bind mutant mRNA in readthrough is determined by nucleotide sequence (PTC context: UGA > UAG > UAA) and inserted amino acid including the nucleotide position +4 of the PTC, as well as the mRNA secondary structure. We used COS-7 cells to investigate the functional characteristics of p.Q500X and p.R619X, IDUA variants and the effects of gentamicin in inducing stop codon readthrough of seven IDUA variants including p.Q500X, p.R619X, p.Q70X, p.E299X, p.W312X, p.Q380X, and p.W402X. Moreover, we performed prediction of RNA secondary structure using the online tool RNAfold. We found that cells treated with gentamicin showed significantly enhanced full-length IDUA expression and restored IDUA activity, in a dose-dependent manner, only in cells expressing cDNA with W312X, Q380X, W402X, and R619X. Among the readthrough-responsive variants, we observed UGA PTC in W312X, W402X and R619X; and UAG PTC with C at nucleotide +4 in Q380X. Changes of RNA secondary structure were noted only in mutants with readthrough-responsive variants including W312X, Q380X, W402X, and R619X. Additional preclinical studies of selected PTCs with potential readthrough, using drugs with less oto-nephrotoxicity, in patient's skin fibroblasts and animal model are necessary for the premise of personalized medicine.

Identification of soybean GolS gene family and analysis of expression patterns under salt and drought stresses

Galactinol synthase (GolS) is a key enzyme in the biosynthetic pathway of raffinose family oligosaccharides (RFOs) and plays an important role in plant responses to abiotic stresses. However, the molecular characteristics of the GolS family members in soybean was not well-known. In this study, six members of GmGolS gene family were genome-widely identified, and their physicochemical properties, chromosomal localization, evolutionary relationship, gene structure, conserved motifs, secondary structure, tertiary structure, tissue-specific expression patterns and the expression levels under salt and drought stresses were analyzed. The results showed that six soybean GolS genes were unevenly distributed on four chromosomes, the range of the isoelectric points of six GmGolS proteins was 5.45-6.08, the molecular weight range was 37 567.07-38 817.59 Da, and the number of amino acids was 324-339 aa. The results of subcellular localization showed that 4 proteins were located in the chloroplast, and 2 proteins in the cytoplasm. Phylogenetic tree analysis showed that the members of the soybean GolS gene family were closely adjacent to each other, and were evolutionarily conservative. Six gene members contain 3 or 4 exons. Prediction of secondary and tertiary structures showed that the spatial structure of proteins of all family members was mainly composed of α-helix and random coil structure, with less β-turn and extended chain structure. Tissue-specific expression analysis showed that six GmGolS members expressed to variable degrees in seeds, roots, root hairs, flowers, stems, pods, nodules and leaves. Expression analysis based on qRT-PCR showed that all GmGolS genes showed different degrees of up-regulated expression under salt and drought treatment, indicating that these genes may be related to the response of plants to salt-tolerance and drought-resistance. These results may facilitate subsequent functional analysis of soybean GolS genes.

1H, 13C, 15N backbone resonance assignment of apo and ADP-ribose bound forms of the macro domain of Hepatitis E virus through solution NMR spectroscopy

The genome of Hepatitis E virus (HEV) is 7.2 kilobases long and has three open reading frames. The largest one is ORF1, encoding a non-structural protein involved in the replication process, and whose processing is ill-defined. The ORF1 protein is a multi-modular protein which includes a macro domain (MD). MDs are evolutionarily conserved structures throughout all kingdoms of life. MDs participate in the recognition and removal of ADP-ribosylation, and specifically viral MDs have been identified as erasers of ADP-ribose moieties interpreting them as important players at escaping the early stages of host-immune response. A detailed structural analysis of the apo and bound to ADP-ribose state of the native HEV MD would provide the structural information to understand how HEV MD is implicated in virus-host interplay and how it interacts with its intracellular partner during viral replication. In the present study we present the high yield expression of the native macro domain of HEV and its analysis by solution NMR spectroscopy. The HEV MD is folded in solution and we present a nearly complete backbone and sidechains assignment for apo and bound states. In addition, a secondary structure prediction by TALOS + analysis was performed. The results indicated that HEV MD has a α/β/α topology very similar to that of most viral macro domains.

Protein encoder: An autoencoder-based ensemble feature selection scheme to predict protein secondary structure

Proteins play a vital role in the human body as they perform important metabolic tasks. Experimental identification of protein structure is expensive and time consuming. The prediction of protein secondary structure is significant to identify the protein tertiary structure and its folds. The feature subset selection from high dimensional protein primary sequence is a key to improve the accuracy of Protein Secondary Structure Prediction (PSSP). Therefore, it is essential to select the relevant features from high dimensional data to predict the protein secondary structure. This work presents a novel method for the PSSP problem based on a two-phase feature selection technique. The first stage utilizes an unsupervised autoencoder for feature extractions. Whereas, the second stage is an ensemble of three feature selection methods, namely, generic univariate select, recursive feature elimination, and Pearson's correlation. This phase combines multiple feature subsets using mutual information to select the optimum feature subset. For classification, different resultant subset features are used. These include random forest, decision tree, and multilayer perceptron. Two sets of experiments are performed on five datasets for the assessment of proposed work. The proposed solution is compared with three state-of-the-art methods based on Q3 accuracy, Q8 accuracy, and segment overlap score. Obtained results show that the proposed framework performs better in the majority of the cases than the past contributions. The proposed framework achieves Q8 accuracies of 82%, 80%, 79%, 73%, and 74% and Q3 accuracies of 90%, 90%, 92%, 79%, and 74% on CB6133, CB6133-filtered, CB513, CASP10, and CASP11 datasets, respectively.

G4Atlas: a comprehensive transcriptome-wide G-quadruplex database.

RNA G-quadruplex (rG4) is a vital RNA tertiary structure motif that involves the base pairs on both Hoogsteen and Watson-Crick faces of guanines. rG4 is of great importance in the post-transcriptional regulation of gene expression. Experimental technologies have advanced to identify in vitro and in vivo rG4s across diverse transcriptomes. Building on these recent advances, here we present G4Atlas, the first transcriptome-wide G-quadruplex database, in which we have collated, classified, and visualized transcriptome rG4 experimental data, generated from rG4-seq, chemical profiling and ligand-binding methods. Our comprehensive database includes transcriptome-wide rG4s generated from 82 experimental treatments and 238 samples across ten species. In addition, we have also included RNA secondary structure prediction information across both experimentally identified and unidentified rG4s to enable users to display any potential competitive folding between rG4 and RNA secondary structures. As such, G4Atlas will enable users to explore the general functions of rG4s in diverse biological processes. In addition, G4Atlas lays the foundation for further data-driven deep learning algorithms to examine rG4 structural features.

Comparative Evaluation of Dengue Virus (DENV) Serotypes Infections in Human (Homo sapiens)

Background: Dengue virus is transmitted by the mosquito species Aedes aegypti and causes dengue fever. The single positive-stranded RNA virus of the family Flaviviridae belongs to the genus Flavivirus. There is four dengue virus (DENV) serotypes. (DENV-1, DENV-2, DENV-3, and DENV-4). Each has altered interactions with the antibodies in the human blood serum. Materials and Methods: In the study, we reported that the comparative study of dengue serotypes along with multiple sequence alignment, domain organization, structure modeling, secondary structure prediction, epitope peptide identification, including the phosphorylation sites. Results: A significant objective for future research will be to completely comprehend the auxiliary and proper angles and clarify the various jobs. Conclusion: We present a comparison of Dengue virus serotypes in this study (DENV1, DENV2, DENV3, and DENV4). We looked into dengue serotype differences in this study.

Redefining the catalytic HECT domain boundaries for the HECT E3 ubiquitin ligase family.

There are 28 unique human members of the homologous to E6AP C-terminus (HECT) E3 ubiquitin ligase family. Each member of the HECT E3 ubiquitin ligases contains a conserved bilobal HECT domain of approximately 350 residues found near their C-termini that is responsible for their respective ubiquitylation activities. Recent studies have begun to elucidate specific roles that each HECT E3 ubiquitin ligase has in various cancers, age-induced neurodegeneration, and neurological disorders. New structural models have been recently released for some of the HECT E3 ubiquitin ligases, but many HECT domain structures have yet to be examined due to chronic insolubility and/or protein folding issues. Building on these recently published structural studies coupled with our in-house experiments discussed in the present study, we suggest that the addition of ∼50 conserved residues preceding the N-terminal to the current UniProt defined boundaries of the HECT domain are required for isolating soluble, stable, and active HECT domains. We show using in silico bioinformatic analyses coupled with secondary structural prediction software that this predicted N-terminal α-helix found in all 28 human HECT E3 ubiquitin ligases forms an obligate amphipathic α-helix that binds to a hydrophobic pocket found within the HECT N-terminal lobe. The present study brings forth the proposal to redefine the residue boundaries of the HECT domain to include this N-terminal extension that will likely be critical for future biochemical, structural, and therapeutic studies on the HECT E3 ubiquitin ligase family.

ProtTrans: Toward Understanding the Language of Life Through Self-Supervised Learning

Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models (LMs) taken from Natural Language Processing (NLP). These LMs reach for new prediction frontiers at low inference costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The protein LMs (pLMs) were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores. Dimensionality reduction revealed that the raw pLM-embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks: (1) a per-residue (per-token) prediction of protein secondary structure (3-state accuracy Q3=81%-87%); (2) per-protein (pooling) predictions of protein sub-cellular location (ten-state accuracy: Q10=81%) and membrane versus water-soluble (2-state accuracy Q2=91%). For secondary structure, the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without multiple sequence alignments (MSAs) or evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that pLMs learned some of the grammar of the language of life. All our models are available through https://github.com/agemagician/ProtTrans.

Targeted Sequencing of Cytokine-Induced PI3K-Related Genes in Ulcerative Colitis, Colorectal Cancer and Colitis-Associated Cancer.

Chronic relapsing inflammatory bowel disease is strongly linked to an increased risk of colitis-associated cancer (CAC). One of the well-known inflammatory carcinogenesis pathways, phosphatidylinositol 3-kinase (PI3K), was identified to be a crucial mechanism in long-standing ulcerative colitis (UC). The goal of this study was to identify somatic variants in the cytokine-induced PI3K-related genes in UC, colorectal cancer (CRC) and CAC. Thirty biopsies (n = 8 long-standing UC, n = 11 CRC, n = 8 paired normal colorectal mucosa and n = 3 CAC) were subjected to targeted sequencing on 13 PI3K-related genes using Illumina sequencing and the SureSelectXT Target Enrichment System. The Genome Analysis Toolkit was used to analyze variants, while ANNOVAR was employed to detect annotations. There were 5116 intronic, 355 exonic, 172 untranslated region (UTR) and 59 noncoding intronic variations detected across all samples. Apart from a very small number of frameshifts, the distribution of missense and synonymous variants was almost equal. We discovered changed levels of IL23R, IL12Rß1, IL12Rß2, TYK2, JAK2 and OSMR in more than 50% of the samples. The IL23R variant in the UTR region, rs10889677, was identified to be a possible variant that might potentially connect CAC with UC and CRC. Additional secondary structure prediction using RNAfold revealed that mutant structures were more unstable than wildtype structures. Further functional research on the potential variants is, therefore, highly recommended since it may provide insight on the relationship between inflammation and cancer risk in the cytokine-induced PI3K pathway.

A novel gyrase inhibitor from toxin-antitoxin system expressed by Staphylococcus aureus.

Toxin-antitoxin (TA) systems consist of a toxin inhibiting essential cellular functions (such as DNA, RNA and protein synthesis), and its cognate antitoxin neutralizing the toxicity. Recently, we identified a TA system termed TsbA/TsbT in the Staphylococcus aureus genome. The induction of the tsbT gene in Escherichia coli halted both DNA and RNA synthesis, reduced supercoiled plasmid and resulted in increasingly relaxed DNA. These results suggested that DNA gyrase was the target of TsbT. In addition, TsbT inhibited both E. coli and S. aureus DNA gyrase activity and induced linearization of plasmid DNA in vitro. Taken together, these results demonstrate that the TsbT toxin targets DNA gyrase in vivo. Site-directed mutagenesis experiments showed that the E27 and D37 residues in TsbT are critical for toxicity. Secondary structure prediction combining the analysis of vacuum-ultraviolet circular-dichroism spectroscopy and neural network method demonstrated that the 22nd-32nd residues of TsbT form an α-helix structure, and that the E27 residue is located around the centre of the α-helix segment. These findings give new insights not only into S. aureus TA systems, but also into bacterial toxins targeting DNA topoisomerases.

Propensities of Some Amino Acid Pairings in α-Helices Vary with Length.

The results of secondary structure prediction methods are widely used in applications in biotechnology and bioinformatics. However, the accuracy limit of these methods could be improved up to 92%. One approach to achieve this goal is to harvest information from the primary structure of the peptide. This study aims to contribute to this goal by investigating the variations in propensity of amino acid pairings to α-helices in globular proteins depending on helix length. (n):(n + 4) residue pairings were determined using a comprehensive peptide data set according to backbone hydrogen bond criterion which states that backbone hydrogen bond is the dominant driving force of protein folding. Helix length is limited to 13 to 26 residues. Findings of this study show that propensities of ALA:GLY and GLY:GLU pairings to α-helix in globular protein increase with increasing helix length but of ALA:ALA and ALA:VAL decrease. While the frequencies of ILE:ALA, LEU:ALA, LEU:GLN, LEU:GLU, LEU:LEU, MET:ILE and VAL:LEU pairings remain roughly constant with length, the 25 residue pairings have varying propensities in narrow helix lengths. The remaining pairings have no prominent propensity to α-helices.

SnoDB 2.0: an enhanced interactive database, specializing in human snoRNAs.

snoDB is an interactive database of human small nucleolar RNAs (snoRNAs) that includes up-to-date information on snoRNA features, genomic location, conservation, host gene, snoRNA-RNA targets andsnoRNA abundance and provides links to other resources. In the second edition of this database (snoDB 2.0), we added an entirely new section on ribosomal RNA (rRNA) chemical modifications guided by snoRNAs with easy navigation between the different rRNA versions used in the literature and experimentally measured levels of modification. We also included new layers of information, including snoRNA motifs, secondary structure prediction, snoRNA-protein interactions, copy annotations and low structure bias expression data in a wide panel of tissues and cell lines to bolster functional probing of snoRNA biology. Version 2.0 features updated identifiers, more links to external resources and duplicate entry resolution. As a result, snoDB 2.0, which is freely available at https://bioinfo-scottgroup.med.usherbrooke.ca/snoDB/, represents a one-stop shop for snoRNA features, rRNA modification targets, functional impact and potential regulators.

ProteinGLUE multi-task benchmark suite for self-supervised protein modeling.

Self-supervised language modeling is a rapidly developing approach for the analysis of protein sequence data. However, work in this area is heterogeneous and diverse, making comparison of models and methods difficult. Moreover, models are often evaluated only on one or two downstream tasks, making it unclear whether the models capture generally useful properties. We introduce the ProteinGLUE benchmark for the evaluation of protein representations: a set of seven per-amino-acid tasks for evaluating learned protein representations. We also offer reference code, and we provide two baseline models with hyperparameters specifically trained for these benchmarks. Pre-training was done on two tasks, masked symbol prediction and next sentence prediction. We show that pre-training yields higher performance on a variety of downstream tasks such as secondary structure and protein interaction interface prediction, compared to no pre-training. However, the larger base model does not outperform the smaller medium model. We expect the ProteinGLUE benchmark dataset introduced here, together with the two baseline pre-trained models and their performance evaluations, to be of great value to the field of protein sequence-based property prediction. Availability: code and datasets from https://github.com/ibivu/protein-glue.