Critical assessment of high-throughput standalone methods for secondary structure prediction

Taxonomy-specific assessment of intrinsic disorder predictions at residue and region levels in higher eukaryotes, protists, archaea, bacteria and viruses

Intrinsic disorder predictors were evaluated in several studies including the two large CAID experiments. However, these studies are biased towards eukaryotic proteins and focus primarily on the residue-level predictions. We provide first-of-its-kind assessment that comprehensively covers the taxonomy and evaluates predictions at the residue and disordered region levels. We curate a benchmark dataset that uniformly covers eukaryotic, archaeal, bacterial, and viral proteins. We find that predictive performance differs substantially across taxonomy, where viruses are predicted most accurately, followed by protists and higher eukaryotes, while bacterial and archaeal proteins suffer lower levels of accuracy. These trends are consistent across predictors. We also find that current tools, except for flDPnn, struggle with reproducing native distributions of the numbers and sizes of the disordered regions. Moreover, analysis of two variants of disorder predictions derived from the AlphaFold2 predicted structures reveals that they produce accurate residue-level propensities for archaea, bacteria and protists. However, they underperform for higher eukaryotes and generally struggle to accurately identify disordered regions. Our results motivate development of new predictors that target bacteria and archaea and which produce accurate results at both residue and region levels. We also stress the need to include the region-level assessments in future assessments.

The N-terminal intrinsically disordered region of Ncb5or docks with the cytochrome b5 core to form a helical motif that is of ancient origin

NADH cytochrome b5 oxidoreductase (Ncb5or) is a cytosolic ferric reductase implicated in diabetes and neurological conditions. Ncb5or comprises cytochrome b5 (b5 ) and cytochrome b5 reductase (b5 R) domains separated by a CHORD-Sgt1 (CS) linker domain. Ncb5or redox activity depends on proper inter-domain interactions to mediate electron transfer from NADH or NADPH via FAD to heme. While full-length human Ncb5or has proven resistant to crystallization, we have succeeded in obtaining high-resolution atomic structures of the b5 domain and a construct containing the CS and b5 R domains (CS/b5 R). Ncb5or also contains an N-terminal intrinsically disordered region of 50 residues that has no homologs in other protein families in animals but features a distinctive, conserved L34 MDWIRL40 motif also present in reduced lateral root formation (RLF) protein in rice and increased recombination center 21 in baker's yeast, all attaching to a b5 domain. After unsuccessful attempts at crystallizing a human Ncb5or construct comprising the N-terminal region naturally fused to the b5 domain, we were able to obtain a high-resolution atomic structure of a recombinant rice RLF construct corresponding to residues 25-129 of human Ncb5or (52% sequence identity; 74% similarity). The structure reveals Trp120 (corresponding to invariant Trp37 in Ncb5or) to be part of an 11-residue α-helix (S116 QMDWLKLTRT126 ) packing against two of the four helices in the b5 domain that surround heme (α2 and α5). The Trp120 side chain forms a network of interactions with the side chains of four highly conserved residues corresponding to Tyr85 and Tyr88 (α2), Cys124 (α5), and Leu47 in Ncb5or. Circular dichroism measurements of human Ncb5or fragments further support a key role of Trp37 in nucleating the formation of the N-terminal helix, whose location in the N/b5 module suggests a role in regulating the function of this multi-domain redox enzyme. This study revealed for the first time an ancient origin of a helical motif in the N/b5 module as reflected by its existence in a class of cytochrome b5 proteins from three kingdoms among eukaryotes.

Availability of web servers significantly boosts citations rates of bioinformatics methods for protein function and disorder prediction

Motivation

Development of bioinformatics methods is a long, complex and resource-hungry process. Hundreds of these tools were released. While some methods are highly cited and used, many suffer relatively low citation rates. We empirically analyze a large collection of recently released methods in three diverse protein function and disorder prediction areas to identify key factors that contribute to increased citations.

Results

We show that provision of a working web server significantly boosts citation rates. On average, methods with working web servers generate three times as many citations compared to tools that are available as only source code, have no code and no server, or are no longer available. This observation holds consistently across different research areas and publication years. We also find that differences in predictive performance are unlikely to impact citation rates. Overall, our empirical results suggest that a relatively low-cost investment into the provision and long-term support of web servers would substantially increase the impact of bioinformatics tools.

Empirical Study of Protein Feature Representation on Deep Belief Networks Trained With Small Data for Secondary Structure Prediction

Protein secondary structure (SS) prediction is a classic problem of computational biology and is widely used in structural characterization and to infer homology. While most SS predictors have been trained on thousands of sequences, a previous approach had developed a compact model of training proteins that used a C-Alpha, C-Beta Side Chain (CABS)-algorithm derived energy based feature representation. Here, the previous approach is extended to Deep Belief Networks (DBN). Deep learning methods are notorious for requiring large datasets and there is a wide consensus that training deep models from scratch on small datasets, works poorly. By contrast, we demonstrate a simple DBN architecture containing a single hidden layer, trained only on the CB513 dataset. Testing on an independent set of G Switch proteins improved the Q ₃ score of the previous compact model by almost 3%. The findings are further confirmed by comparison to several deep learning models which are trained on thousands of proteins. Finally, the DBN performance is also compared with Position Specific Scoring Matrix (PSSM)-profile based feature representation. The importance of (i) structural information in protein feature representation and (ii) complementary small dataset learning approaches for detection of structural fold switching are demonstrated.

Comparative Analysis on Alignment-Based and Pretrained Feature Representations for the Identification of DNA-Binding Proteins

The interaction between DNA and protein is vital for the development of a living body. Previous numerous studies on in silico identification of DNA-binding proteins (DBPs) usually include features extracted from the alignment-based (pseudo) position-specific scoring matrix (PSSM), leading to limited application due to its time-consuming generation. Few researchers have paid attention to the application of pretrained language models at the scale of evolution to the identification of DBPs. To this end, we present comprehensive insights into a comparison study on alignment-based PSSM and pretrained evolutionary scale modeling (ESM) representations in the field of DBP classification. The comparison is conducted by extracting information from PSSM and ESM representations using four unified averaging operations and by performing various feature selection (FS) methods. Experimental results demonstrate that the pretrained ESM representation outperforms the PSSM-derived features in a fair comparison perspective. The pretrained feature presentation deserves wide application to the area of in silico DBP identification as well as other function annotation issues. Finally, it is also confirmed that an ensemble scheme by aggregating various trained FS models can significantly improve the classification performance of DBPs.

Complementarity of the residue-level protein function and structure predictions in human proteins

Sequence-based predictors of the residue-level protein function and structure cover a broad spectrum of characteristics including intrinsic disorder, secondary structure, solvent accessibility and binding to nucleic acids. They were catalogued and evaluated in numerous surveys and assessments. However, methods focusing on a given characteristic are studied separately from predictors of other characteristics, while they are typically used on the same proteins. We fill this void by studying complementarity of a representative collection of methods that target different predictions using a large, taxonomically consistent, and low similarity dataset of human proteins. First, we bridge the gap between the communities that develop structure-trained vs. disorder-trained predictors of binding residues. Motivated by a recent study of the protein-binding residue predictions, we empirically find that combining the structure-trained and disorder-trained predictors of the DNA-binding and RNA-binding residues leads to substantial improvements in predictive quality. Second, we investigate whether diverse predictors generate results that accurately reproduce relations between secondary structure, solvent accessibility, interaction sites, and intrinsic disorder that are present in the experimental data. Our empirical analysis concludes that predictions accurately reflect all combinations of these relations. Altogether, this study provides unique insights that support combining results produced by diverse residue-level predictors of protein function and structure.

Optimized Elastic Network Models With Direct Characterization of Inter-Residue Cooperativity for Protein Dynamics

The elastic network models (ENMs)are known as representative coarse-grained models to capture essential dynamics of proteins. Due to simple designs of the force constants as a decay with spatial distances of residue pairs in many previous studies, there is still much room for the improvement of ENMs. In this article, we directly computed the force constants with the inverse covariance estimation using a ridge-type operater for the precision matrix estimation (ROPE)on a large-scale set of NMR ensembles. Distance-dependent statistical analyses on the force constants were further comprehensively performed in terms of several paired types of sequence and structural information, including secondary structure, relative solvent accessibility, sequence distance and terminal. Various distinguished distributions of the mean force constants highlight the structural and sequential characteristics coupled with the inter-residue cooperativity beyond the spatial distances. We finally integrated these structural and sequential characteristics to build novel ENM variations using the particle swarm optimization for the parameter estimation. The considerable improvements on the correlation coefficient of the mean-square fluctuation and the mode overlap were achieved by the proposed variations when compared with traditional ENMs. This study opens a novel way to develop more accurate elastic network models for protein dynamics.

QUARTERplus: Accurate disorder predictions integrated with interpretable residue-level quality assessment scores

A recent advance in the disorder prediction field is the development of the quality assessment (QA) scores. QA scores complement the propensities produced by the disorder predictors by identifying regions where these predictions are more likely to be correct. We develop, empirically test and release a new QA tool, QUARTERplus, that addresses several key drawbacks of the current QA method, QUARTER. QUARTERplus is the first solution that utilizes QA scores and the associated input disorder predictions to produce very accurate disorder predictions with the help of a modern deep learning meta-model. The deep neural network utilizes the QA scores to identify and fix the regions where the original/input disorder predictions are poor. More importantly, the accurate QUATERplus's predictions are accompanied by easy to interpret residue-level QA scores that reliably quantify their residue-level predictive quality. We provide these interpretable QA scores for QUARTERplus and 10 other popular disorder predictors. Empirical tests on a large and independent (low similarity) test dataset show that QUARTERplus predictions secure AUC = 0.93 and are statistically more accurate than the results of twelve state-of-the-art disorder predictors. We also demonstrate that the new QA scores produced by QUARTERplus are highly correlated with the actual predictive quality and that they can be effectively used to identify regions of correct disorder predictions. This feature empowers the users to easily identify which parts of the predictions generated by the modern disorder predictors are more trustworthy. QUARTERplus is available as a convenient webserver at http://biomine.cs.vcu.edu/servers/QUARTERplus/.

DescribePROT: database of amino acid-level protein structure and function predictions

We present DescribePROT, the database of predicted amino acid-level descriptors of structure and function of proteins. DescribePROT delivers a comprehensive collection of 13 complementary descriptors predicted using 10 popular and accurate algorithms for 83 complete proteomes that cover key model organisms. The current version includes 7.8 billion predictions for close to 600 million amino acids in 1.4 million proteins. The descriptors encompass sequence conservation, position specific scoring matrix, secondary structure, solvent accessibility, intrinsic disorder, disordered linkers, signal peptides, MoRFs and interactions with proteins, DNA and RNAs. Users can search DescribePROT by the amino acid sequence and the UniProt accession number and entry name. The pre-computed results are made available instantaneously. The predictions can be accesses via an interactive graphical interface that allows simultaneous analysis of multiple descriptors and can be also downloaded in structured formats at the protein, proteome and whole database scale. The putative annotations included by DescriPROT are useful for a broad range of studies, including: investigations of protein function, applied projects focusing on therapeutics and diseases, and in the development of predictors for other protein sequence descriptors. Future releases will expand the coverage of DescribePROT. DescribePROT can be accessed at http://biomine.cs.vcu.edu/servers/DESCRIBEPROT/.

Comprehensive Survey and Comparative Assessment of RNA-Binding Residue Predictions with Analysis by RNA Type

With close to 30 sequence-based predictors of RNA-binding residues (RBRs), this comparative survey aims to help with understanding and selection of the appropriate tools. We discuss past reviews on this topic, survey a comprehensive collection of predictors, and comparatively assess six representative methods. We provide a novel and well-designed benchmark dataset and we are the first to report and compare protein-level and datasets-level results, and to contextualize performance to specific types of RNAs. The methods considered here are well-cited and rely on machine learning algorithms on occasion combined with homology-based prediction. Empirical tests reveal that they provide relatively accurate predictions. Virtually all methods perform well for the proteins that interact with rRNAs, some generate accurate predictions for mRNAs, snRNA, SRP and IRES, while proteins that bind tRNAs are predicted poorly. Moreover, except for DRNApred, they confuse DNA and RNA-binding residues. None of the six methods consistently outperforms the others when tested on individual proteins. This variable and complementary protein-level performance suggests that users should not rely on applying just the single best dataset-level predictor. We recommend that future work should focus on the development of approaches that facilitate protein-level selection of accurate predictors and the consensus-based prediction of RBRs.

Prediction of Ion Channels and their Types from Protein Sequences: Comprehensive Review and Comparative Assessment

BACKGROUND

Ion channels are a large and growing protein family. Many of them are associated with diseases, and consequently, they are targets for over 700 drugs. Discovery of new ion channels is facilitated with computational methods that predict ion channels and their types from protein sequences. However, these methods were never comprehensively compared and evaluated.

OBJECTIVE

We offer first-of-its-kind comprehensive survey of the sequence-based predictors of ion channels. We describe eight predictors that include five methods that predict ion channels, their types, and four classes of the voltage-gated channels. We also develop and use a new benchmark dataset to perform comparative empirical analysis of the three currently available predictors.

RESULTS

While several methods that rely on different designs were published, only a few of them are currently available and offer a broad scope of predictions. Support and availability after publication should be required when new methods are considered for publication. Empirical analysis shows strong performance for the prediction of ion channels and modest performance for the prediction of ion channel types and voltage-gated channel classes. We identify a substantial weakness of current methods that cannot accurately predict ion channels that are categorized into multiple classes/types.

CONCLUSION

Several predictors of ion channels are available to the end users. They offer practical levels of predictive quality. Methods that rely on a larger and more diverse set of predictive inputs (such as PSIONplus) are more accurate. New tools that address multi-label prediction of ion channels should be developed.

Deep learning methods in protein structure prediction

Protein Structure Prediction is a central topic in Structural Bioinformatics. Since the '60s statistical methods, followed by increasingly complex Machine Learning and recently Deep Learning methods, have been employed to predict protein structural information at various levels of detail. In this review, we briefly introduce the problem of protein structure prediction and essential elements of Deep Learning (such as Convolutional Neural Networks, Recurrent Neural Networks and basic feed-forward Neural Networks they are founded on), after which we discuss the evolution of predictive methods for one-dimensional and two-dimensional Protein Structure Annotations, from the simple statistical methods of the early days, to the computationally intensive highly-sophisticated Deep Learning algorithms of the last decade. In the process, we review the growth of the databases these algorithms are based on, and how this has impacted our ability to leverage knowledge about evolution and co-evolution to achieve improved predictions. We conclude this review outlining the current role of Deep Learning techniques within the wider pipelines to predict protein structures and trying to anticipate what challenges and opportunities may arise next.

NetSurfP-2.0: Improved prediction of protein structural features by integrated deep learning

The ability to predict local structural features of a protein from the primary sequence is of paramount importance for unraveling its function in absence of experimental structural information. Two main factors affect the utility of potential prediction tools: their accuracy must enable extraction of reliable structural information on the proteins of interest, and their runtime must be low to keep pace with sequencing data being generated at a constantly increasing speed. Here, we present NetSurfP-2.0, a novel tool that can predict the most important local structural features with unprecedented accuracy and runtime. NetSurfP-2.0 is sequence-based and uses an architecture composed of convolutional and long short-term memory neural networks trained on solved protein structures. Using a single integrated model, NetSurfP-2.0 predicts solvent accessibility, secondary structure, structural disorder, and backbone dihedral angles for each residue of the input sequences. We assessed the accuracy of NetSurfP-2.0 on several independent test datasets and found it to consistently produce state-of-the-art predictions for each of its output features. We observe a correlation of 80% between predictions and experimental data for solvent accessibility, and a precision of 85% on secondary structure 3-class predictions. In addition to improved accuracy, the processing time has been optimized to allow predicting more than 1000 proteins in less than 2 hours, and complete proteomes in less than 1 day.

Computational Prediction of Secondary and Supersecondary Structures from Protein Sequences

Many new methods for the sequence-based prediction of the secondary and supersecondary structures have been developed over the last several years. These and older sequence-based predictors are widely applied for the characterization and prediction of protein structure and function. These efforts have produced countless accurate predictors, many of which rely on state-of-the-art machine learning models and evolutionary information generated from multiple sequence alignments. We describe and motivate both types of predictions. We introduce concepts related to the annotation and computational prediction of the three-state and eight-state secondary structure as well as several types of supersecondary structures, such as β hairpins, coiled coils, and α-turn-α motifs. We review 34 predictors focusing on recent tools and provide detailed information for a selected set of 14 secondary structure and 3 supersecondary structure predictors. We conclude with several practical notes for the end users of these predictive methods.

HEMEsPred: Structure-Based Ligand-Specific Heme Binding Residues Prediction by Using Fast-Adaptive Ensemble Learning Scheme

Heme is an essential biomolecule that widely exists in numerous extant organisms. Accurately identifying heme binding residues (HEMEs) is of great importance in disease progression and drug development. In this study, a novel predictor named HEMEsPred was proposed for predicting HEMEs. First, several sequence- and structure-based features, including amino acid composition, motifs, surface preferences, and secondary structure, were collected to construct feature matrices. Second, a novel fast-adaptive ensemble learning scheme was designed to overcome the serious class-imbalance problem as well as to enhance the prediction performance. Third, we further developed ligand-specific models considering that different heme ligands varied significantly in their roles, sizes, and distributions. Statistical test proved the effectiveness of ligand-specific models. Experimental results on benchmark datasets demonstrated good robustness of our proposed method. Furthermore, our method also showed good generalization capability and outperformed many state-of-art predictors on two independent testing datasets. HEMEsPred web server was available at http://www.inforstation.com/HEMEsPred/ for free academic use.

Machine learning-enabled discovery and design of membrane-active peptides

Antimicrobial peptides are a class of membrane-active peptides that form a critical component of innate host immunity and possess a diversity of sequence and structure. Machine learning approaches have been profitably employed to efficiently screen sequence space and guide experiment towards promising candidates with high putative activity. In this mini-review, we provide an introduction to antimicrobial peptides and summarize recent advances in machine learning-enabled antimicrobial peptide discovery and design with a focus on a recent work Lee et al. Proc. Natl. Acad. Sci. USA 2016;113(48):13588-13593. This study reports the development of a support vector machine classifier to aid in the design of membrane active peptides. We use this model to discover membrane activity as a multiplexed function in diverse peptide families and provide interpretable understanding of the physicochemical properties and mechanisms governing membrane activity. Experimental validation of the classifier reveals it to have learned membrane activity as a unifying signature of antimicrobial peptides with diverse modes of action. Some of the discriminating rules by which it performs classification are in line with existing "human learned" understanding, but it also unveils new previously unknown determinants and multidimensional couplings governing membrane activity. Integrating machine learning with targeted experimentation can guide both antimicrobial peptide discovery and design and new understanding of the properties and mechanisms underpinning their modes of action.

Effects of ADAMTS14 genetic polymorphism and cigarette smoking on the clinicopathologic development of hepatocellular carcinoma

Background

ADAMTS14 is a member of the ADAMTS (adisintegrin and metalloproteinase with thrombospondin motifs), which are proteolytic enzymes with a variety of further ancillary domain in the C-terminal region for substrate specificity and enzyme localization via extracellular matrix association. However, whether ADAMTS14 genetic variants play a role in hepatocellular carcinoma (HCC) susceptibility remains unknown.

Methodology/Principal findings

Four non-synonymous single-nucleotide polymorphisms (nsSNPs) of the ADAMTS14 gene were examined from 680 controls and 340 patients with HCC. Among 141 HCC patients with smoking behaviour, we found significant associations of the rs12774070 (CC+AA vs CC) and rs61573157 (CT+TT vs CC) variants with a clinical stage of HCC (OR: 2.500 and 2.767; 95% CI: 1.148–5.446 and 1.096–6.483; P = 0.019 and 0.026, respectively) and tumour size (OR: 2.387 and 2.659; 95% CI: 1.098–5.188 and 1.055–6.704; P = 0.026 and 0.034, respectively), but not with lymph node metastasis or other clinical statuses. Moreover, an additional integrated in silico analysis proposed that rs12774070 and rs61573157 affected essential post-translation O-glycosylation site within the 3^rd thrombospondin type 1 repeat and a novel proline-rich region embedded within the C-terminal extension, respectively.

Conclusions

Taken together, our results suggest an involvement of ADAMTS14 SNP rs12774070 and rs61573157 in the liver tumorigenesis and implicate the ADAMTS14 gene polymorphism as a predict factor during the progression of HCC.

Mapping membrane activity in undiscovered peptide sequence space using machine learning

There are some ∼1,100 known antimicrobial peptides (AMPs), which permeabilize microbial membranes but have diverse sequences. Here, we develop a support vector machine (SVM)-based classifier to investigate ⍺-helical AMPs and the interrelated nature of their functional commonality and sequence homology. SVM is used to search the undiscovered peptide sequence space and identify Pareto-optimal candidates that simultaneously maximize the distance σ from the SVM hyperplane (thus maximize its "antimicrobialness") and its ⍺-helicity, but minimize mutational distance to known AMPs. By calibrating SVM machine learning results with killing assays and small-angle X-ray scattering (SAXS), we find that the SVM metric σ correlates not with a peptide's minimum inhibitory concentration (MIC), but rather its ability to generate negative Gaussian membrane curvature. This surprising result provides a topological basis for membrane activity common to AMPs. Moreover, we highlight an important distinction between the maximal recognizability of a sequence to a trained AMP classifier (its ability to generate membrane curvature) and its maximal antimicrobial efficacy. As mutational distances are increased from known AMPs, we find AMP-like sequences that are increasingly difficult for nature to discover via simple mutation. Using the sequence map as a discovery tool, we find a unexpectedly diverse taxonomy of sequences that are just as membrane-active as known AMPs, but with a broad range of primary functions distinct from AMP functions, including endogenous neuropeptides, viral fusion proteins, topogenic peptides, and amyloids. The SVM classifier is useful as a general detector of membrane activity in peptide sequences.

Protein secondary structure prediction using a small training set (compact model) combined with a Complex-valued neural network approach

BACKGROUND

Protein secondary structure prediction (SSP) has been an area of intense research interest. Despite advances in recent methods conducted on large datasets, the estimated upper limit accuracy is yet to be reached. Since the predictions of SSP methods are applied as input to higher-level structure prediction pipelines, even small errors may have large perturbations in final models. Previous works relied on cross validation as an estimate of classifier accuracy. However, training on large numbers of protein chains compromises the classifier ability to generalize to new sequences. This prompts a novel approach to training and an investigation into the possible structural factors that lead to poor predictions. Here, a small group of 55 proteins termed the compact model is selected from the CB513 dataset using a heuristics-based approach. In a prior work, all sequences were represented as probability matrices of residues adopting each of Helix, Sheet and Coil states, based on energy calculations using the C-Alpha, C-Beta, Side-chain (CABS) algorithm. The functional relationship between the conformational energies computed with CABS force-field and residue states is approximated using a classifier termed the Fully Complex-valued Relaxation Network (FCRN). The FCRN is trained with the compact model proteins.

RESULTS

The performance of the compact model is compared with traditional cross-validated accuracies and blind-tested on a dataset of G Switch proteins, obtaining accuracies of ∼81 %. The model demonstrates better results when compared to several techniques in the literature. A comparative case study of the worst performing chain identifies hydrogen bond contacts that lead to Coil ⇔ Sheet misclassifications. Overall, mispredicted Coil residues have a higher propensity to participate in backbone hydrogen bonding than correctly predicted Coils.

CONCLUSIONS

The implications of these findings are: (i) the choice of training proteins is important in preserving the generalization of a classifier to predict new sequences accurately and (ii) SSP techniques sensitive in distinguishing between backbone hydrogen bonding and side-chain or water-mediated hydrogen bonding might be needed in the reduction of Coil ⇔ Sheet misclassifications.

CONFOLD: Residue-residue contact-guided ab initio protein folding

Predicted protein residue-residue contacts can be used to build three-dimensional models and consequently to predict protein folds from scratch. A considerable amount of effort is currently being spent to improve contact prediction accuracy, whereas few methods are available to construct protein tertiary structures from predicted contacts. Here, we present an ab initio protein folding method to build three-dimensional models using predicted contacts and secondary structures. Our method first translates contacts and secondary structures into distance, dihedral angle, and hydrogen bond restraints according to a set of new conversion rules, and then provides these restraints as input for a distance geometry algorithm to build tertiary structure models. The initially reconstructed models are used to regenerate a set of physically realistic contact restraints and detect secondary structure patterns, which are then used to reconstruct final structural models. This unique two-stage modeling approach of integrating contacts and secondary structures improves the quality and accuracy of structural models and in particular generates better β-sheets than other algorithms. We validate our method on two standard benchmark datasets using true contacts and secondary structures. Our method improves TM-score of reconstructed protein models by 45% and 42% over the existing method on the two datasets, respectively. On the dataset for benchmarking reconstructions methods with predicted contacts and secondary structures, the average TM-score of best models reconstructed by our method is 0.59, 5.5% higher than the existing method. The CONFOLD web server is available at http://protein.rnet.missouri.edu/confold/.

Secondary structure assignment for conformationally irregular peptides: comparison between DSSP, STRIDE and KAKSI

Secondary structure assignment codes were built to explore the regularities associated with the periodic motifs of proteins, such as those in backbone dihedral angles or in hydrogen bonds between backbone atoms. Precise structure assignment is challenging because real-life secondary structures are susceptible to bending, twist, fraying and other deformations that can distance them from their geometrical prototypes. Although results from codes such as DSSP and STRIDE converge in well-ordered structures, the agreement between the secondary structure assignments is known to deteriorate as the conformations become more distorted. Conformationally irregular peptides therefore offer a great opportunity to explore the differences between these codes. This is especially important for unfolded proteins and intrinsically disordered proteins, which are known to exhibit residual and/or transient secondary structure whose characterization is challenging. In this work, we have carried out Molecular Dynamics simulations of (relatively) disordered peptides, specifically gp41659-671 (ELLELDKWASLWN), the homopeptide polyasparagine (N18), and polyasparagine dimers. We have analyzed the resulting conformations with DSSP and STRIDE, based on hydrogen-bond patterns (and dihedral angles for STRIDE), and KAKSI, based on α-Carbon distances; and carefully characterized the differences in structural assignments. The full-sequence Segment Overlap (SOV) scores, that quantify the agreement between two secondary structure assignments, vary from 70% for gp41659-671 (STRIDE as reference) to 49% for N18 (DSSP as reference). Major differences are observed in turns, in the distinction between α and 310 helices, and in short parallel-sheet segments.

Bayesian model of protein primary sequence for secondary structure prediction

Determining the primary structure (i.e., amino acid sequence) of a protein has become cheaper, faster, and more accurate. Higher order protein structure provides insight into a protein's function in the cell. Understanding a protein's secondary structure is a first step towards this goal. Therefore, a number of computational prediction methods have been developed to predict secondary structure from just the primary amino acid sequence. The most successful methods use machine learning approaches that are quite accurate, but do not directly incorporate structural information. As a step towards improving secondary structure reduction given the primary structure, we propose a Bayesian model based on the knob-socket model of protein packing in secondary structure. The method considers the packing influence of residues on the secondary structure determination, including those packed close in space but distant in sequence. By performing an assessment of our method on 2 test sets we show how incorporation of multiple sequence alignment data, similarly to PSIPRED, provides balance and improves the accuracy of the predictions. Software implementing the methods is provided as a web application and a stand-alone implementation.

From local structure to a global framework: recognition of protein folds

Protein folding has been a major area of research for many years. Nonetheless, the mechanisms leading to the formation of an active biological fold are still not fully apprehended. The huge amount of available sequence and structural information provides hints to identify the putative fold for a given sequence. Indeed, protein structures prefer a limited number of local backbone conformations, some being characterized by preferences for certain amino acids. These preferences largely depend on the local structural environment. The prediction of local backbone conformations has become an important factor to correctly identifying the global protein fold. Here, we review the developments in the field of local structure prediction and especially their implication in protein fold recognition.

SSThread: Template-free protein structure prediction by threading pairs of contacting secondary structures followed by assembly of overlapping pairs

Acquiring the three-dimensional structure of a protein from its amino acid sequence alone, despite a great deal of work and significant progress on the subject, is still an unsolved problem. SSThread, a new template-free algorithm is described here that consists of making several predictions of contacting pairs of α-helices and β-strands derived from a database of experimental structures using a knowledge-based potential, secondary structure prediction, and contact map prediction followed by assembly of overlapping pair predictions to create an ensemble of core structure predictions whose loops are then predicted. In a set of seven CASP10 targets SSThread outperformed the two leading methods for two targets each. The targets were all β-strand containing structures and most of them have a high relative contact order which demonstrates the advantages of SSThread. The primary bottlenecks based on sets of 74 and 21 test cases are the pair prediction and loop prediction stages.

Sequence-derived structural features driving proteolytic processing

Proteolytic signaling, or regulated proteolysis, is an essential part of many important pathways such as Notch, Wnt, and Hedgehog. How the structure of the cleaved substrate regions influences the efficacy of proteolytic processing remains underexplored. Here, we analyzed the relative importance in proteolysis of various structural features derived from substrate sequences using a dataset of more than 5000 experimentally verified proteolytic events captured in CutDB. Accessibility to the solvent was recognized as an essential property of a proteolytically processed polypeptide chain. Proteolytic events were found nearly uniformly distributed among three types of secondary structure, although with some enrichment in loops. Cleavages in α-helices were found to be relatively abundant in regions apparently prone to unfolding, while cleavages in β-structures tended to be located at the periphery of β-sheets. Application of the same statistical procedures to proteolytic events divided into separate sets according to the catalytic classes of proteases proved consistency of the results and confirmed that the structural mechanisms of proteolysis are universal. The estimated prediction power of sequence-derived structural features, which turned out to be sufficiently high, presents a rationale for their use in bioinformatic prediction of proteolytic events.

Prediction and characterization of cyclic proteins from sequences in three domains of life

Cyclic proteins (CPs) have circular chains with a continuous cycle of peptide bonds. Their unique structural traits result in greater stability and resistance to degradation when compared to their acyclic counterparts. They are also promising targets for pharmaceutical/therapeutic applications. To date, only a few hundred CPs are known, although recent studies suggest that their numbers might be substantially higher. Here we developed a first-of-its-kind, accurate and high-throughput method called CyPred that predicts whether a given protein chain is cyclic. CyPred considers currently well-represented CP families: cyclotides, cyclic defensins, bacteriocins, and trypsin inhibitors. Empirical tests demonstrate that CyPred outperforms commonly used alignment methods. We used CyPred to estimate the incidence of CPs and found ~3500 putative CPs among 5.7+ million chains from 642 fully sequenced proteomes from archaea, bacteria, and eukaryotes. The median number of putative CPs per species ranges from three for archaea proteomes to two for eukaryotes/bacteria, with 7% of archaea, 11% of bacterial, and 16% of eukaryotic proteomes having 10+ CPs. The differences in the estimated fractions of CPs per proteome are as large as three orders of magnitude. Among eukaryotes, animals have higher ratios of CPs compared to fungi, while plants have the largest spread of the ratios. We also show that proteomes enriched in cyclic proteins evolve more slowly than proteomes with fewer cyclic chains. Our results suggest that further research is needed to fully uncover the scope and potential of cyclic proteins. A list of putative CPs and the CyPred method are available at http://biomine.ece.ualberta.ca/CyPred/. This article is part of a Special Issue entitled: Computational Proteomics, Systems Biology & Clinical Implications. Guest Editor: Yudong Cai.

Comprehensively designed consensus of standalone secondary structure predictors improves Q3 by over 3%

Protein fold is defined by a spatial arrangement of three types of secondary structures (SSs) including helices, sheets, and coils/loops. Current methods that predict SS from sequences rely on complex machine learning-derived models and provide the three-state accuracy (Q3) at about 82%. Further improvements in predictive quality could be obtained with a consensus-based approach, which so far received limited attention. We perform first-of-its-kind comprehensive design of a SS consensus predictor (SScon), in which we consider 12 modern standalone SS predictors and utilize Support Vector Machine (SVM) to combine their predictions. Using a large benchmark data-set with 10 random training-test splits, we show that a simple, voting-based consensus of carefully selected base methods improves Q3 by 1.9% when compared to the best single predictor. Use of SVM provides additional 1.4% improvement with the overall Q3 at 85.6% and segment overlap (SOV3) at 83.7%, when compared to 82.3 and 80.9%, respectively, obtained by the best individual methods. We also show strong improvements when the consensus is based on ab-initio methods, with Q3 = 82.3% and SOV3 = 80.7% that match the results from the best template-based approaches. Our consensus reduces the number of significant errors where helix is confused with a strand, provides particularly good results for short helices and strands, and gives the most accurate estimates of the content of individual SSs in the chain. Case studies are used to visualize the improvements offered by the consensus at the residue level. A web-server and a standalone implementation of SScon are available at http://biomine.ece.ualberta.ca/SSCon/ .

AMPLE: a cluster-and-truncate approach to solve the crystal structures of small proteins using rapidly computed ab initio models

Protein ab initio models predicted from sequence data alone can enable the elucidation of crystal structures by molecular replacement. However, the calculation of such ab initio models is typically computationally expensive. Here, a computational pipeline based on the clustering and truncation of cheaply obtained ab initio models for the preparation of structure ensembles is described. Clustering is used to select models and to quantitatively predict their local accuracy, allowing rational truncation of predicted inaccurate regions. The resulting ensembles, with or without rapidly added side chains, solved 43% of all test cases, with an 80% success rate for all-α proteins. A program implementing this approach, AMPLE, is included in the CCP4 suite of programs. It only requires the input of a FASTA sequence file and a diffraction data file. It carries out the modelling using locally installed Rosetta, creates search ensembles and automatically performs molecular replacement and model rebuilding.

Advances in the Prediction of Turn Structures in Peptides and Proteins

Turns are essential for protein structure as they allow the polypeptide chain to fold backup on itself. They also occur within protein binding sites, at proteinprotein interfaces and in small bioactive peptides, where they can play a crucial role for molecular recognition. Turn structures are an important class of protein secondary structure, although relatively little attention is paid to them with respect to helices and β-sheets. Protein structure prediction, functional analysis of proteins and peptides, and computer-aided drug design could all benefit from making use of accurately predicted turn structures from amino acid sequence. Here, recent advances of turn structure prediction and the underlying turn classification will be discussed together with their applications.

Predicting protein-ATP binding sites from primary sequence through fusing bi-profile sampling of multi-view features

Background

Adenosine-5′-triphosphate (ATP) is one of multifunctional nucleotides and plays an important role in cell biology as a coenzyme interacting with proteins. Revealing the binding sites between protein and ATP is significantly important to understand the functionality of the proteins and the mechanisms of protein-ATP complex.

Results

In this paper, we propose a novel framework for predicting the proteins’ functional residues, through which they can bind with ATP molecules. The new prediction protocol is achieved by combination of sequence evolutional information and bi-profile sampling of multi-view sequential features and the sequence derived structural features. The hypothesis for this strategy is single-view feature can only represent partial target’s knowledge and multiple sources of descriptors can be complementary.

Conclusions

Prediction performances evaluated by both 5-fold and leave-one-out jackknife cross-validation tests on two benchmark datasets consisting of 168 and 227 non-homologous ATP binding proteins respectively demonstrate the efficacy of the proposed protocol. Our experimental results also reveal that the residue structural characteristics of real protein-ATP binding sites are significant different from those normal ones, for example the binding residues do not show high solvent accessibility propensities, and the bindings prefer to occur at the conjoint points between different secondary structure segments. Furthermore, results also show that performance is affected by the imbalanced training datasets by testing multiple ratios between positive and negative samples in the experiments. Increasing the dataset scale is also demonstrated useful for improving the prediction performances.

A comparative study on filtering protein secondary structure prediction

Filtering of Protein Secondary Structure Prediction (PSSP) aims to provide physicochemically realistic results, while it usually improves the predictive performance. We performed a comparative study on this challenging problem, utilizing both machine learning techniques and empirical rules and we found that combinations of the two lead to the highest improvement.

TANGLE: two-level support vector regression approach for protein backbone torsion angle prediction from primary sequences

Protein backbone torsion angles (Phi) and (Psi) involve two rotation angles rotating around the C_α-N bond (Phi) and the C_α-C bond (Psi). Due to the planarity of the linked rigid peptide bonds, these two angles can essentially determine the backbone geometry of proteins. Accordingly, the accurate prediction of protein backbone torsion angle from sequence information can assist the prediction of protein structures. In this study, we develop a new approach called TANGLE (Torsion ANGLE predictor) to predict the protein backbone torsion angles from amino acid sequences. TANGLE uses a two-level support vector regression approach to perform real-value torsion angle prediction using a variety of features derived from amino acid sequences, including the evolutionary profiles in the form of position-specific scoring matrices, predicted secondary structure, solvent accessibility and natively disordered region as well as other global sequence features. When evaluated based on a large benchmark dataset of 1,526 non-homologous proteins, the mean absolute errors (MAEs) of the Phi and Psi angle prediction are 27.8° and 44.6°, respectively, which are 1% and 3% respectively lower than that using one of the state-of-the-art prediction tools ANGLOR. Moreover, the prediction of TANGLE is significantly better than a random predictor that was built on the amino acid-specific basis, with the p-value<1.46e-147 and 7.97e-150, respectively by the Wilcoxon signed rank test. As a complementary approach to the current torsion angle prediction algorithms, TANGLE should prove useful in predicting protein structural properties and assisting protein fold recognition by applying the predicted torsion angles as useful restraints. TANGLE is freely accessible at http://sunflower.kuicr.kyoto-u.ac.jp/~sjn/TANGLE/.

SPINE X: improving protein secondary structure prediction by multistep learning coupled with prediction of solvent accessible surface area and backbone torsion angles

Accurate prediction of protein secondary structure is essential for accurate sequence alignment, three-dimensional structure modeling, and function prediction. The accuracy of ab initio secondary structure prediction from sequence, however, has only increased from around 77 to 80% over the past decade. Here, we developed a multistep neural-network algorithm by coupling secondary structure prediction with prediction of solvent accessibility and backbone torsion angles in an iterative manner. Our method called SPINE X was applied to a dataset of 2640 proteins (25% sequence identity cutoff) previously built for the first version of SPINE and achieved a 82.0% accuracy based on 10-fold cross validation (Q(3)). Surpassing 81% accuracy by SPINE X is further confirmed by employing an independently built test dataset of 1833 protein chains, a recently built dataset of 1975 proteins and 117 CASP 9 targets (critical assessment of structure prediction techniques) with an accuracy of 81.3%, 82.3% and 81.8%, respectively. The prediction accuracy is further improved to 83.8% for the dataset of 2640 proteins if the DSSP assignment used above is replaced by a more consistent consensus secondary structure assignment method. Comparison to the popular PSIPRED and CASP-winning structure-prediction techniques is made. SPINE X predicts number of helices and sheets correctly for 21.0% of 1833 proteins, compared to 17.6% by PSIPRED. It further shows that SPINE X consistently makes more accurate prediction in helical residues (6%) without over prediction while PSIPRED makes more accurate prediction in coil residues (3-5%) and over predicts them by 7%. SPINE X Server and its training/test datasets are available at http://sparks.informatics.iupui.edu/

Computational prediction of secondary and supersecondary structures

The sequence-based prediction of the secondary and supersecondary structures enjoys strong interest and finds applications in numerous areas related to the characterization and prediction of protein structure and function. Substantial efforts in these areas over the last three decades resulted in the development of accurate predictors, which take advantage of modern machine learning models and availability of evolutionary information extracted from multiple sequence alignment. In this chapter, we first introduce and motivate both prediction areas and introduce basic concepts related to the annotation and prediction of the secondary and supersecondary structures, focusing on the β hairpin, coiled coil, and α-turn-α motifs. Next, we overview state-of-the-art prediction methods, and we provide details for 12 modern secondary structure predictors and 4 representative supersecondary structure predictors. Finally, we provide several practical notes for the users of these prediction tools.

SOMPNN: an efficient non-parametric model for predicting transmembrane helices

Accurately predicting the transmembrane helices (TMH) in a helical membrane protein is an important but challenging task. Recent researches have demonstrated that statistics-based methods are promising routes to improve the TMH prediction accuracy. However, most of existing TMH predictors are parametric models and they have to make assumptions of several or even hundreds of adjustable parameters based on the underlying probability distribution, which is difficult when no a priori knowledge is available. Besides the performances of these parametric predictors significantly depend on the estimated parameters, some of them need to exploit the entire training dataset in the prediction stage, which will lead to low prediction efficiency and this problem will become even worse when dealing with large-scale dataset. In this paper, we propose a novel SOMPNN model for prediction of TMH that features by minimal parameter assumptions requirement and high computational efficiency. In the SOMPNN model, a self-organizing map (SOM) is used to adaptively learn the helices distribution knowledge hidden in the training data, and then a probabilistic neural network (PNN) is adopted to predict TMH segments based on the knowledge learned by SOM. Experimental results on two benchmark datasets show that the proposed SOMPNN outperforms most existing popular TMH predictors and is promising to be extended to deal with other complicated biological problems. The datasets and the source codes of SOMPNN are available at http://www.csbio.sjtu.edu.cn/bioinf/SOMPNN/.