Fold Recognition Research Articles

PRFold-TNN: Protein Fold Recognition With an Ensemble Feature Selection Method Using PageRank Algorithm Based on Transformer.

Understanding the tertiary structures of proteins is of great benefit to function in many aspects of human life. Protein fold recognition is a vital and salient means to know protein structure. Until now, researchers have successively proposed a variety of methods to realize protein fold recognition, but the novel and effective computational method is still needed to handle this problem with the continuous updating of protein structure databases. In this study, we develop a new protein structure dataset named AT and propose the PRFold-TNN model for protein fold recognition. Firstly, different types of feature extraction methods including AAC, HMM, HMM-Bigram and ACC are selected to extract corresponding features for protein sequences. Then an ensemble feature selection method based on PageRank algorithm integrating various tree-based algorithms is used to screen the fusion features. Ultimately, the classifier based on the Transformer model achieves the final prediction. Experiments show that the prediction accuracy is 86.27% on the AT dataset and 88.91% on the independent test set, indicating that the model can demonstrate superior performance and generalization ability in the problem of protein fold recognition. Furthermore, we also carry out research on the DD, EDD and TG benchmark datasets, and make them achieve prediction accuracy of 88.41%, 97.91% and 95.16%, which are at least 3.0%, 0.8% and 2.5% higher than those of the state-of-the-art methods. It can be concluded that the PRFold-TNN model is more prominent.

The glycosylation landscape of prostate cancer tissues and biofluids.

An overview of the role of glycosylation in prostate cancer (PCa) development and progression is presented, focusing on recent advancements in defining the N-glycome through glycomic profiling and glycoproteomic methodologies. Glycosylation is a common post-translational modification typified by oligosaccharides attached N-linked to asparagine or O-linked to serine or threonine on carrier proteins. These attached sugars have crucial roles in protein folding and cellular recognition processes, such that altered glycosylation is a hallmark of cancer pathogenesis and progression. In the past decade, advancements in N-glycan profiling workflows using Matrix Assisted Laser Desorption/Ionization Mass Spectrometry Imaging (MALDI-MSI) technology have been applied to define the spatial distribution of glycans in PCa tissues. Multiple studies applying N-glycan MALDI-MSI to pathology-defined PCa tissues have identified significant alterations in N-glycan profiles associated with PCa progression. N-glycan compositions progressively increase in number, and structural complexity due to increased fucosylation and sialylation. Additionally, significant progress has been made in defining the glycan and glycopeptide compositions of prostatic-derived glycoproteins like prostate-specific antigen in tissues and biofluids. The glycosyltransferases involved in these changes are potential drug targets for PCa, and new approaches in this area are summarized. These advancements will be discussed in the context of the further development of clinical diagnostics and therapeutics targeting glycans and glycoproteins associated with PCa progression. Integration of large scale spatial glycomic data for PCa with other spatial-omic methodologies is now feasible at the tissue and single-cell levels.

Structural and dynamic mechanisms for coupled folding and tRNA recognition of a translational T-box riboswitch

T-box riboswitches are unique riboregulators where gene regulation is mediated through interactions between two highly structured RNAs. Despite extensive structural insights, how RNA-RNA interactions drive the folding and structural transitions of T-box to achieve functional conformations remains unclear. Here, by combining SAXS, single-molecule FRET and computational modeling, we elaborate the folding energy landscape of a translational T-box aptamer consisting of stems I, II and IIA/B, which Mg2+-induced global folding and tRNA binding are cooperatively coupled. smFRET measurements reveal that high Mg2+ stabilizes IIA/B and its stacking on II, which drives the pre-docking of I and II into a competent conformation, subsequent tRNA binding promotes docking of I and II to form a high-affinity tRNA binding groove, of which the essentiality of IIA/B and S-turn in II is substantiated with mutational analysis. We highlight a delicate balance among Mg2+, the intra- and intermolecular RNA-RNA interactions in modulating RNA folding and function.

ResCNNT-fold: Combining residual convolutional neural network and Transformer for protein fold recognition from language model embeddings

A comprehensive understanding of protein functions holds significant promise for disease research and drug development, and proteins with analogous tertiary structures tend to exhibit similar functions. Protein fold recognition stands as a classical approach in the realm of protein structure investigation. Despite significant advancements made by researchers in this field, the continuous updating of protein databases presents an ongoing challenge in accurately identifying protein fold types. In this study, we introduce a predictor, ResCNNT-fold, for protein fold recognition and employ the LE dataset for testing purpose. ResCNNT-fold leverages a pre-trained language model to obtain embedding representations for protein sequences, which are then processed by the ResCNNT feature extractor, a combination of residual convolutional neural network and Transformer, to derive fold-specific features. Subsequently, the query protein is paired with each protein whose structure is known in the template dataset. For each pair, the similarity score of their fold-specific features is calculated. Ultimately, the query protein is identified as the fold type of the template protein in the pair with the highest similarity score. To further validate the utility and efficacy of the proposed ResCNNT-fold predictor, we conduct a 2-fold cross-validation experiment on the fold level of the LE dataset. Remarkably, this rigorous evaluation yields an exceptional accuracy of 91.57%, which surpasses the best result among other state-of-the-art protein fold recognition methods by an approximate margin of 10%. The excellent performance unequivocally underscores the compelling advantages inherent to our proposed ResCNNT-fold predictor in the realm of protein fold recognition. The source code and data of ResCNNT-fold can be downloaded from https://github.com/Bioinformatics-Laboratory/ResCNNT-fold.

The Realm of Unconventional Noncovalent Interactions in Proteins: Their Significance in Structure and Function.

Proteins and their assemblies are fundamental for living cells to function. Their complex three-dimensional architecture and its stability are attributed to the combined effect of various noncovalent interactions. It is critical to scrutinize these noncovalent interactions to understand their role in the energy landscape in folding, catalysis, and molecular recognition. This Review presents a comprehensive summary of unconventional noncovalent interactions, beyond conventional hydrogen bonds and hydrophobic interactions, which have gained prominence over the past decade. The noncovalent interactions discussed include low-barrier hydrogen bonds, C5 hydrogen bonds, C-H···π interactions, sulfur-mediated hydrogen bonds, n → π* interactions, London dispersion interactions, halogen bonds, chalcogen bonds, and tetrel bonds. This Review focuses on their chemical nature, interaction strength, and geometrical parameters obtained from X-ray crystallography, spectroscopy, bioinformatics, and computational chemistry. Also highlighted are their occurrence in proteins or their complexes and recent advances made toward understanding their role in biomolecular structure and function. Probing the chemical diversity of these interactions, we determined that the variable frequency of occurrence in proteins and the ability to synergize with one another are important not only for ab initio structure prediction but also to design proteins with new functionalities. A better understanding of these interactions will promote their utilization in designing and engineering ligands with potential therapeutic value.

Evolution and phylogenetic distribution of endo-α-mannosidase.

While glycans underlie many biological processes, such as protein folding, cell adhesion, and cell-cell recognition, deep evolution of glycosylation machinery remains an understudied topic. N-linked glycosylation is a conserved process in which mannosidases are key trimming enzymes. One of them is the glycoprotein endo-α-1,2-mannosidase which participates in the initial trimming of mannose moieties from an N-linked glycan inside the cis-Golgi. It is unique as the only endo-acting mannosidase found in this organelle. Relatively little is known about its origins and evolutionary history; so far it was reported to occur only in vertebrates. In this work, a taxon-rich bioinformatic survey to unravel the evolutionary history of this enzyme, including all major eukaryotic clades and a wide representation of animals, is presented. The endomannosidase was found to be more widely distributed in animals and other eukaryotes. The protein motif changes in context of the canonical animal enzyme were tracked. Additionally, the data show the two canonical vertebrate endomannosidase genes, MANEA and MANEAL, arose at the second round of the two vertebrate genome duplications and one more vertebrate paralog, CMANEAL, is uncovered. Finally, a framework where N-glycosylation co-evolved with complex multicellularity is described. A better understanding of the evolution of core glycosylation pathways is pivotal to understanding biology of eukaryotes in general, and the Golgi apparatus in particular. This systematic analysis of the endomannosidase evolution is one step toward this goal.

Automated identification and mapping of geological folds in cross sections

Abstract Cross sections carry information on the spatial distribution of rock strata and the development of geological structures, and it is an important data source for three-dimensional (3D) geological modeling. However, the interpretation and mapping of geological structures in sections by means of manual interpretation are inefficient and costly, and the performance varies greatly with the experts’ ability and experience. The objective of this article is to develop an automatic recognition and mapping method for folds in cross sections. This method mainly includes identifying folds based on stratigraphic sequence characteristics (symmetrical and repetitive), classifying fold types based on geometric attributes of folds (interval scheduling, strike, and section morphology), optimizing strata based on the superposition principle and area conservation principle, and constructing the polygon features of folds. Based on experiments in the Parallel Fold Belt of Eastern Sichuan and the central Appalachian fold-thrust belt in the Appalachian Mountains, the method presented in this article can effectively be used for automatic recognition and high-quality mapping of folds in the cross sections. The method provides a good source of geological cross-sectional data for the 3D modeling of geologic bodies.

TemPred: A Novel Protein Template Search Engine to Improve Protein Structure Prediction.

Among new protein structure predictors, the recently developed AlphaFold predictor relies on contact map in line with contact map potential based threading model that basically relies on fold recognition. In parallel, sequence similarity based homology model relies on homologue recognition. Both of these methods rely on sequence-structure or sequence-sequence similarity with protein with known structure in absence of which, as argued in the development of AlphaFold, the structure prediction becomes quite challenging. However, the term, "known structure" depends on the similarity method adopted to identify it, for example, through sequence match yielding homologue or sequence-structure match yielding a fold. Also, quite often, AlphaFold structures are found to be not acceptable by the structure evaluating gold standard parameters. In this context, this work utilized the concept of ordered local physicochemical property, ProtPCV by Pal et al (2020) providing a new similarity criteria to identify the template protein with known structure. Finally a template search engine, TemPred was developed using the ProtPCV similarity criteria. It was intriguing to find that quite often templates generated by TemPred were better than that produced by the conventional search engines. It pointed out the need of combined approach to get better structural model for a protein.

The biological role of core 1β1-3galactosyltransferase (T-synthase) in mucin-type O-glycosylation in Silkworm, Bombyx mori

O-glycosylation of secreted and membrane-bound proteins is an important post-translational modification that affects recognition of cell surface receptors, protein folding, and stability. However, despite the importance of O-linked glycans, their biological functions have not yet been fully elucidated and the synthetic pathway of O-glycosylation has not been investigated in detail, especially in the silkworm. In this study, we aimed to investigate O-glycosylation in silkworms by analyzing the overall structural profiles of mucin-type O-glycans using LC–MS. We found GalNAc or GlcNAc monosaccharide and core 1 disaccharide (Galβ1-3-GalNAcα1-Ser/Thr) were major components of the O-glycan attached to secreted proteins produced in silkworms. Furthermore, we characterized the 1 b1,3-galactosyltransferase (T-synthase) required for synthesis of the core 1 structure, common to many animals. Five transcriptional variants and four protein isoforms were identified in silkworms, and the biological functions of these isoforms were investigated. We found that BmT-synthase isoforms 1 and 2 were localized in the Golgi apparatus in cultured BmN4 cells and functioned both in cultured cells and silkworms. Additionally, a specific functional domain of T-synthase, called the stem domain, was found to be essential for activity and is presumed to be needed for dimer formation and galactosyltransferase activity. Altogether, our results elucidated the O-glycan profile and function of T-synthase in the silkworm. Our findings allow the practical comprehension of O-glycosylation required for employing silkworms as a productive expression system.

A role of salt bridges in mediating drug potency: A lesson from the N-myristoyltransferase inhibitors.

The salt bridge is the strongest non-covalent interaction in nature and is known to participate in protein folding, protein-protein interactions, and molecular recognition. However, the role of salt bridges in the context of drug design has remained not well understood. Here, we report that a common feature in the mechanism of inhibition of the N-myristoyltransferases (NMT), promising targets for the treatment of protozoan infections and cancer, is the formation of a salt bridge between a positively charged chemical group of the small molecule and the negatively charged C-terminus of the enzyme. Substituting the inhibitor positively charged amine group with a neutral methylene group prevents the formation of the salt bridge and leads to a dramatic activity loss. Molecular dynamics simulations have revealed that salt bridges stabilize the NMT-ligand complexes by functioning as molecular clips that stabilize the conformation of the protein structure. As such, the creation of salt bridges between the ligands and their protein targets may find an application as a valuable tool in rational drug design.

Structural mechanisms of calmodulin activation of Shigella effector OspC3 to ADP-riboxanate caspase-4/11 and block pyroptosis.

The caspase-4/11-GSDMD pyroptosis axis recognizes cytosolic lipopolysaccharide for antibacterial defenses. Shigella flexneri employs an OspC3 effector to block pyroptosis by catalyzing NAD+-dependent arginine ADP-riboxanation of caspase-4/11. Here, we identify Ca2+-free calmodulin (CaM) that binds and stimulates OspC3 ADP-riboxanase activity. Crystal structures of OspC3-CaM and OspC3-caspase-4 binary complexes reveal unique CaM binding to an OspC3 N-terminal domain featuring an ADP-ribosyltransferase-like fold and specific recognition of caspase-4 by an OspC3 ankryin repeat domain, respectively. CaM-OspC3-caspase-4 ternary complex structures show that NAD+ binding reorganizes the catalytic pocket, in which D231 and D177 activate the substrate arginine for initial ADP-ribosylation and ribosyl 2'-OH in the ADP-ribosylated arginine, respectively, for subsequent deamination. We also determine structures of unmodified and OspC3-ADP-riboxanated caspase-4. Mechanisms derived from this series of structures covering the entire process of OspC3 action are supported by biochemical analyses in vitro and functional validation in S. flexneri-infected mice.

Analysis of miRNA rare variants in amyotrophic lateral sclerosis and in silico prediction of their biological effects.

Background: Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease affecting upper and/or lower motor neurons and characterized by complex etiology. Familial cases show high genetic heterogeneity and sporadic cases (90%) are associated with several genetic and environmental risk factors. Among the genetic risk factors, the contribution of non-coding elements, such as microRNAs (miRNAs), to ALS disease susceptibility remains largely unexplored. Aim: This work aims to identify rare variants in miRNA genes in sporadic ALS (sALS) patients which may cause a defective miRNA maturation or altered target gene recognition by changing miRNA secondary structure or seed sequence, respectively. Methods: Rare variants located in miRNA loci with a minor allele frequency (MAF) < 0.01 were extracted from whole genome sequencing (WGS) data of 100 sALS patients. The secondary pre-miRNA structures were predicted using MiRVas to evaluate the impact of the variants on RNA folding process. Human TargetScan was used to retrieve all the potential target genes of miRNAs with variants in the seed region. Over Representation Analysis (ORA) was conducted to compare the lists of target genes for the reference and mutated miRNAs in the seed sequence. Results: Our analysis identified 86 rare variants in 77 distinct miRNAs and distributed in different parts of the miRNA precursors. The presence of these variants changed miRNA secondary structures in ∼70% of MiRVas predictions. By focusing on the 6 rare variants mapping within the seed sequence, the predicted target genes increased in number compared to the reference miRNA and included novel targets in a proportion ranging from 30 to 82%. Interestingly, ORA revealed significant changes in gene set enrichment only for mutated miR-509-1 and miR-941-3 for which the Gene Ontology term related to "nervous system development" was absent and present, respectively, compared to target lists of the reference miRNA. Conclusion: We here developed a workflow to study miRNA rare variants from WGS data and to predict their biological effects on miRNA folding, maturation and target gene recognition. Although this in silico approach certainly needs functional validation in vitro and in vivo, it may help define the role of miRNA variability in ALS and complex diseases.

Potential Novel N-Glycosylation Patterns Associated with the Emergence of New Genetic Variants of PRRSV-2 in the U.S.

Glycosylation of proteins is a post-translational process where oligosaccharides are attached to proteins, potentially altering their folding, epitope availability, and immune recognition. In Porcine reproductive and respiratory syndrome virus-type 2 (PRRSV-2), positive selection pressure acts on amino acid sites potentially associated with immune escape through glycan shielding. Here, we describe the patterns of potential N-glycosylation sites over time and across different phylogenetic lineages of PRRSV-2 to better understand how these may contribute to patterns of coexistence and emergence of different lineages. We screened 19,179 PRRSV GP5 sequences (2004−2021) in silico for potential N-glycosylated sites. The emergence of novel combinations of N-glycosylated sites coincided with past PRRSV epidemics in the U.S. For lineage L1A, glycosylation at residues 32, 33, 44, 51, and 57 first appeared in 2012, but represented >62% of all L1A sequences by 2015, coinciding with the emergence of the L1A 1-7-4 strain that increased in prevalence from 8 to 86% of all L1A sequences from 2012 to 2015. The L1C 1-4-4 strain that emerged in 2020 also had a distinct N-glycosylation pattern (residues 32, 33, 44, and 51). From 2020 to 2021, this pattern was responsible for 44−47% of the L1C sequences, contrasting to <5% in years prior. Our findings support the hypothesis that antigenic evolution contributes to the sequential dominance of different PRRSV strains and that N-glycosylation patterns may partially account for antigenic differences amongst strains. Further studies on glycosylation and its effect on PRRSV GP5 folding are needed to further understand how glycosylation patterns shape PRRSV occurrence.

Bacterial Outer Membrane Polysaccharide Export (OPX) Proteins Occupy Three Structural Classes with Selective β-Barrel Porin Requirements for Polymer Secretion.

ABSTRACTSecretion of high-molecular-weight polysaccharides across the bacterial envelope is ubiquitous, as it enhances prokaryotic survival in (a)biotic settings. Such polymers are often assembled by Wzx/Wzy- or ABC transporter-dependent schemes implicating outer membrane (OM) polysaccharide export (OPX) proteins in cell-surface polymer translocation. In the social predatory bacterium Myxococcus xanthus, the exopolysaccharide (EPS) pathway WzaX, major spore coat (MASC) pathway WzaS, and biosurfactant polysaccharide (BPS) pathway WzaB were herein found to be truncated OPX homologues of Escherichia coli Wza lacking OM-spanning α-helices. Comparative genomics across all bacteria (>91,000 OPX proteins identified and analyzed), complemented with cryo-electron tomography cell-envelope analyses, revealed such “truncated” WzaX/S/B architecture to be the most common among three defined OPX-protein structural classes independent of periplasm thickness. Fold recognition and deep learning revealed the conserved M. xanthus proteins MXAN_7418/3226/1916 (encoded beside wzaX/S/B, respectively) to be integral OM β-barrels, with structural homology to the poly-N-acetyl-d-glucosamine synthase-dependent pathway porin PgaA. Such bacterial porins were identified near numerous genes for all three OPX protein classes. Interior MXAN_7418/3226/1916 β-barrel electrostatics were found to match properties of their associated polymers. With MXAN_3226 essential for MASC export, and MXAN_7418 herein shown to mediate EPS translocation, we have designated this new secretion machinery component “Wzp” (i.e., Wz porin), with the final step of M. xanthus EPS/MASC/BPS secretion across the OM now proposed to be mediated by WzpX/S/B (i.e., MXAN_7418/3226/1916). Importantly, these data support a novel and widespread secretion paradigm for polysaccharide biosynthesis pathways in which those containing OPX components that cannot span the OM instead utilize β-barrel porins to mediate polysaccharide transport across the OM.IMPORTANCE Diverse bacteria assemble and secrete polysaccharides that alter their physiologies through modulation of motility, biofilm formation, and host immune system evasion. Most such pathways require outer membrane (OM) polysaccharide export (OPX) proteins for sugar-polymer transport to the cell surface. In the prototypic Escherichia coli Group-1-capsule biosynthesis system, eight copies of this canonical OPX protein cross the OM with an α-helix, forming a polysaccharide-export pore. Herein, we instead reveal that most OPX proteins across all bacteria lack this α-helix, raising questions as to the manner by which most secreted polysaccharides actually exit cells. In the model developmental bacterium Myxococcus xanthus, we show this process to depend on OPX-coupled OM-spanning β-barrel porins, with similar porins encoded near numerous OPX genes in diverse bacteria. Knowledge of the terminal polysaccharide secretion step will enable development of antimicrobial compounds targeted to blocking polymer export from outside the cell, thus bypassing any requirements for antimicrobial compound uptake by the cell.

ASFold-DNN: Protein Fold Recognition Based on Evolutionary Features With Variable Parameters Using Full Connected Neural Network.

Protein fold recognition contribute to comprehend the function of proteins, which is of great help to the gene therapy of diseases and the development of new drugs. Researchers have been working in this direction and have made considerable achievements, but challenges still exist on low sequence similarity datasets. In this study, we propose the ASFold-DNN framework for protein fold recognition research. Above all, four groups of evolutionary features are extracted from the primary structures of proteins, and a preliminary selection of variable parameter is made for two groups of features including ACC _HMM and SXG _HMM, respectively. Then several feature selection algorithms are selected for comparison and the best feature selection scheme is obtained by changing their internal threshold values. Finally, multiple hyper-parameters of Full Connected Neural Network are fully optimized to construct the best model. DD, EDD and TG datasets with low sequence similarities are chosen to evaluate the performance of the models constructed by the framework, and the final prediction accuracy are 85.28, 95.00 and 88.84 percent, respectively. Furthermore, the ASTRAL186 and LE datasets are introduced to further verify the generalization ability of our proposed framework. Comprehensive experimental results prove that the ASFold-DNN framework is more prominent than the state-of-the-art studies on protein fold recognition. The source code and data of ASFold-DNN can be downloaded from https://github.com/Bioinformatics-Laboratory/project/tree/master/ASFold.

Learning Protein Embedding to Improve Protein Fold Recognition Using Deep Metric Learning.

Protein fold recognition refers to predicting the most likely fold type of the query protein and is a critical step of protein structure and function prediction. With the popularity of deep learning in bioinformatics, protein fold recognition has obtained impressive progress. In this study, to extract the fold-specific feature to improve protein fold recognition, we proposed a unified deep metric learning framework based on a joint loss function, termed NPCFold. In addition, we also proposed an integrated machine learning model based on the similarity of proteins in various properties, termed NPCFoldpro. Benchmark experiments show both NPCFold and NPCFoldpro outperform existing protein fold recognition methods at the fold level, indicating that our proposed strategies of fusing loss functions and fusing features could improve the fold recognition level.

Three Dimension Structure Modeling of The Superoxide Dismutase (SOD) of Rice (Oryza sativa) Using Fold Recognition Method Using Phyre2 Web Server

Determining the 3D structure of proteins using laboratory instrumentation is time-consuming and expensive. The in silico method can be used as an alternative to predict the 3D structure of proteins, such as the fold recognition method. This study aims to create a 3D structural model of rice's (Oryza sativa) protein superoxide dismutase (SOD). The 3D structure modeling of the protein was carried out with the Phyre2 web server. The protein sequence was obtained from the UniProt KB database with the code A0A6F8FUX1. The results showed that the suitable template used to build the model was the template with the code c1unfX. The c1unfX template has a coverage value of 80%, 100% confidence, and i.d. of 51%. Validation of the model with the PROCHECK program showed that the most favored area on the Ramachandran Plot was 87.8%, and the disallowed area was 1.1%. The disallowed area, which is still below 15%, indicates that the three-dimensional structure model of the SOD protein built from the c1unfX template has good a value .

Learning Proteome Domain Folding Using LSTMs in an Empirical Kernel Space

The recognition of protein structural folds is the starting point for protein function inference and for many structural prediction tools. We previously introduced the idea of using empirical comparisons to create a data-augmented feature space called PESS (Protein Empirical Structure Space)1 as a novel approach for protein structure prediction. Here, we extend the previous approach by generating the PESS feature space over fixed-length subsequences of query peptides, and applying a sequential neural network model, with one long short-term memory cell layer followed by a fully connected layer. Using this approach, we show that only a small group of domains as a training set is needed to achieve near state-of-the-art accuracy on fold recognition. Our method improves on the previous approach by reducing the training set required and improving the model’s ability to generalize across species, which will help fold prediction for newly discovered proteins.

Pemodelan Calponin Ikan Gabus (Channa striata) dengan Phyre2 dan Interaksi dengan Protein Lain

Knowledge of the three-dimensional structure of proteins is very important to understand their character and function at the molecular level. Determination of protein structure in the laboratory is expensive and relatively difficult because it requires sophisticated instruments and takes a long time. As an alternative, an in silico approach can be used to predict the three-dimensional structure of proteins, namely the fold recognition method. The purpose of this study was to create a three-dimensional structure model of snakehead fish (Channa striata) calponin protein using the Phyre2 web server. The target protein sequence was obtained from UniProt KB with code K9LH64. The results of the study showed that the template for building the model was the c1wynA code. The results of the model (MP) evaluation obtained a coverage value was 45%, a confidence value was 100%, and an identity value was 69%. Validation of the model (MP) using the PROCHECK program showed the model was in the most favoured residue plot of 82.4%. This model deserves to be used as a model for calponin protein of snakehead fish, because the disallowed area was below 15%. The results of the analysis of interactions with other proteins using STRING-DB obtained the interaction model (MP) with Actin assembly-inducing protein (ACTA2) was 0.791; and Tropomyosin (TPM1) was 0.786. The results of molecular docking using PatchDock obtained the value of Atomic Contact Energy (ACE) for the calponin protein model (MP)-ACTA2 was 372.76 kJ/mol; Calponin model protein (MP)-TPM1 was 406.93 kJ/mol. The interactions that occured in the calponin model (MP)-ACTA2 were hydrogen bonds at residues Gly-41, Glu-60, Arg-64, Arg-66, Pro-67, Gly-68, Lys-71 and hydrophobic bonds at Arg-64, Lys- 71, Lys-72, Ile-73, Va-138. The interactions model of calponin model (MP)-TPM1 were hydrogen bonding at residues Ser-13, Lys-71, Ala-95, Tyr-96, Ser-136, Arg-137, Arg-150 and hydrophobic bonds at residues Leu-12, Ala-14, Lys-71, Ala-95, Val-138, Arg-150, Phe-152. Keywords: fold recognition, Phyre2 , 3D structure protein, calponin, protein-protein interaction

Performing protein fold recognition by exploiting a stack convolutional neural network with the attention mechanism

Protein fold recognition is a critical step in protein structure and function prediction, and aims to ascertain the most likely fold type of the query protein. As a typical pattern recognition problem, designing a powerful feature extractor and metric function to extract relevant and representative fold-specific features from protein sequences is the key to improving protein fold recognition. In this study, we propose an effective sequence-based approach, called RattnetFold, to identify protein fold types. The basic concept of RattnetFold is to employ a stack convolutional neural network with the attention mechanism that acts as a feature extractor to extract fold-specific features from protein residue-residue contact maps. Moreover, based on the fold-specific features, we leverage metric learning to project fold-specific features into a subspace where similar proteins are closer together and name this approach RattnetFoldPro. Benchmarking experiments illustrate that RattnetFold and RattnetFoldPro enable the convolutional neural networks to efficiently learn the underlying subtle patterns in residue-residue contact maps, thereby improving the performance of protein fold recognition. An online web server of RattnetFold and the benchmark datasets are freely available at http://csbio.njust.edu.cn/bioinf/rattnetfold/.