Recent Applications of Deep Learning Methods on Evolution- and Contact-Based Protein Structure Prediction

The new advances in deep learning methods have influenced many aspects of scientific research, including the study of the protein system. The prediction of proteins' 3D structural components is now heavily dependent on machine learning techniques that interpret how protein sequences and their homology govern the inter-residue contacts and structural organization. Especially, methods employing deep neural networks have had a significant impact on recent CASP13 and CASP14 competition. Here, we explore the recent applications of deep learning methods in the protein structure prediction area. We also look at the potential opportunities for deep learning methods to identify unknown protein structures and functions to be discovered and help guide drug-target interactions. Although significant problems still need to be addressed, we expect these techniques in the near future to play crucial roles in protein structural bioinformatics as well as in drug discovery.

Prediction of protein-protein interaction sites through eXtreme gradient boosting with kernel principal component analysis

Predicting protein-protein interaction sites (PPI sites) can provide important clues for understanding biological activity. Using machine learning to predict PPI sites can mitigate the cost of running expensive and time-consuming biological experiments. Here we propose PPISP-XGBoost, a novel PPI sites prediction method based on eXtreme gradient boosting (XGBoost). First, the characteristic information of protein is extracted through the pseudo-position specific scoring matrix (PsePSSM), pseudo-amino acid composition (PseAAC), hydropathy index and solvent accessible surface area (ASA) under the sliding window. Next, these raw features are preprocessed to obtain more optimal representations in order to achieve better prediction. In particular, the synthetic minority oversampling technique (SMOTE) is used to circumvent class imbalance, and the kernel principal component analysis (KPCA) is applied to remove redundant characteristics. Finally, these optimal features are fed to the XGBoost classifier to identify PPI sites. Using PPISP-XGBoost, the prediction accuracy on the training dataset Dset186 reaches 85.4%, and the accuracy on the independent validation datasets Dtestset72, PDBtestset164, Dset_448 and Dset_355 reaches 85.3%, 83.9%, 85.8% and 85.4%, respectively, which all show an increase in accuracy against existing PPI sites prediction methods. These results demonstrate that the PPISP-XGBoost method can further enhance the prediction of PPI sites.

IGPRED: Combination of convolutional neural and graph convolutional networks for protein secondary structure prediction

There is a close relationship between the tertiary structure and the function of a protein. One of the important steps to determine the tertiary structure is protein secondary structure prediction (PSSP). For this reason, predicting secondary structure with higher accuracy will give valuable information about the tertiary structure. Recently, deep learning techniques have obtained promising improvements in several machine learning applications including PSSP. In this article, a novel deep learning model, based on convolutional neural network and graph convolutional network is proposed. PSIBLAST PSSM, HHMAKE PSSM, physico-chemical properties of amino acids are combined with structural profiles to generate a rich feature set. Furthermore, the hyper-parameters of the proposed network are optimized using Bayesian optimization. The proposed model IGPRED obtained 89.19%, 86.34%, 87.87%, 85.76%, and 86.54% Q3 accuracies for CullPDB, EVAset, CASP10, CASP11, and CASP12 datasets, respectively.

An improved DNA-binding hot spot residues prediction method by exploring interfacial neighbor properties

Background

DNA-binding hot spots are dominant and fundamental residues that contribute most of the binding free energy yet accounting for a small portion of protein–DNA interfaces. As experimental methods for identifying hot spots are time-consuming and costly, high-efficiency computational approaches are emerging as alternative pathways to experimental methods.

Results

Herein, we present a new computational method, termed inpPDH, for hot spot prediction. To improve the prediction performance, we extract hybrid features which incorporate traditional features and new interfacial neighbor properties. To remove redundant and irrelevant features, feature selection is employed using a two-step feature selection strategy. Finally, a subset of 7 optimal features are chosen to construct the predictor using support vector machine. The results on the benchmark dataset show that this proposed method yields significantly better prediction accuracy than those previously published methods in the literature. Moreover, a user-friendly web server for inpPDH is well established and is freely available at http://bioinfo.ahu.edu.cn/inpPDH.

Conclusions

We have developed an accurate improved prediction model, inpPDH, for hot spot residues in protein–DNA binding interfaces by given the structure of a protein–DNA complex. Moreover, we identify a comprehensive and useful feature subset including the proposed interfacial neighbor features that has an important strength for identifying hot spot residues. Our results indicate that these features are more effective than the conventional features considered previously, and that the combination of interfacial neighbor features and traditional features may support the creation of a discriminative feature set for efficient prediction of hot spot residues in protein–DNA complexes.

Supplementary information

Supplementary information accompanies this paper at 10.1186/s12859-020-03871-1.

Remodelling structure-based drug design using machine learning

To keep up with the pace of rapid discoveries in biomedicine, a plethora of research endeavors had been directed toward Rational Drug Development that slowly gave way to Structure-Based Drug Design (SBDD). In the past few decades, SBDD played a stupendous role in identification of novel drug-like molecules that are capable of altering the structures and/or functions of the target macromolecules involved in different disease pathways and networks. Unfortunately, post-delivery drug failures due to adverse drug interactions have constrained the use of SBDD in biomedical applications. However, recent technological advancements, along with parallel surge in clinical research have led to the concomitant establishment of other powerful computational techniques such as Artificial Intelligence (AI) and Machine Learning (ML). These leading-edge tools with the ability to successfully predict side-effects of a wide range of drugs have eventually taken over the field of drug design. ML, a subset of AI, is a robust computational tool that is capable of data analysis and analytical model building with minimal human intervention. It is based on powerful algorithms that use huge sets of 'training data' as inputs to predict new output values, which improve iteratively through experience. In this review, along with a brief discussion on the evolution of the drug discovery process, we have focused on the methodologies pertaining to the technological advancements of machine learning. This review, with specific examples, also emphasises the tremendous contributions of ML in the field of biomedicine, while exploring possibilities for future developments.

Accurate prediction of protein-ATP binding residues using position-specific frequency matrix

Knowledge of protein-ATP interaction can help for protein functional annotation and drug discovery. Accurately identifying protein-ATP binding residues is an important but challenging task to gain the knowledge of protein-ATP interactions, especially for the case where only protein sequence information is given. In this study, we propose a novel method, named DeepATPseq, to predict protein-ATP binding residues without using any information about protein three-dimension structure or sequence-derived structural information. In DeepATPseq, the HHBlits-generated position-specific frequency matrix (PSFM) profile is first employed to extract the feature information of each residue. Then, for each residue, the PSFM-based feature is fed into two prediction models, which are generated by the algorithms of deep convolutional neural network (DCNN) and support vector machine (SVM) separately. The final ATP-binding probability of the corresponding residue is calculated by the weighted sum of the outputted values of DCNN-based and SVM-based models. Experimental results on the independent validation data set demonstrate that DeepATPseq could achieve an accuracy of 77.71%, covering 57.42% of all ATP-binding residues, while achieving a Matthew's correlation coefficient value (0.655) that is significantly higher than that of existing sequence-based methods and comparable to that of the state-of-the-art structure-based predictors. Detailed data analysis show that the major advantage of DeepATPseq lies at the combination utilization of DCNN and SVM that helps dig out more discriminative information from the PSFM profiles. The online server and standalone package of DeepATPseq are freely available at: https://jun-csbio.github.io/DeepATPseq/for academic use.

Secondary structure prediction of protein based on multi scale convolutional attention neural networks

To fully extract the local and long-range information of amino acid sequences and enhance the effective information, this research proposes a secondary structure prediction model of protein based on a multi-scale convolutional attentional neural network. The model uses a multi-channel multi-scale parallel architecture to extract amino acid structure features of different granularity according to the window size. The reconstructed feature maps are obtained via multiple convolutional attention blocks. Then, the reconstructed feature map is fused with the input feature map to obtain the enhanced feature map. Finally, the enhanced feature map is fed to the Softmax classifier for prediction. While the traditional cross-entropy loss cannot effectively solve the problem of non-equilibrium training samples, a modified correlated cross-entropy loss function may alleviate this problem. After numerous comparison and ablation experiments, it is verified that the improved model can indeed effectively extract amino acid sequence feature information, alleviate overfitting, and thus improve the overall prediction accuracy.

Estimating the Influence of Physicochemical and Biochemical Property Indexes on Selection for Amino Acids Usage in Eukaryotic Cells

Proteins can evolve by accumulating changes on amino acid sequences. These changes are mainly caused by missense mutations on its DNA coding sequences. Mutations with neutral or positive effects on fitness can be maintained while deleterious mutations tend to be eliminated by natural selection. Amino acid changes are influenced by the biophysical, chemical, and biological properties of amino acids. There is a multiplicity of amino acid properties that can influence the function and expression of proteins. Amino acid properties can be expressed into numerical indexes, which can help to predict functional and structural aspects of proteins and allow statistical inferences of selection pressure on amino acid usage. The accuracy of these analyses may be compromised by the existence of several numerical indexes that measure the same amino acid property, and the lack of objective parameters to determine the most accurate and biologically relevant index. In the present study, the gradient consistency test was used in order to estimate the magnitude of directional selection imparted by amino acid biochemical and biophysical properties on protein evolution.

EDLMFC: an ensemble deep learning framework with multi-scale features combination for ncRNA-protein interaction prediction

Background

Non-coding RNA (ncRNA) and protein interactions play essential roles in various physiological and pathological processes. The experimental methods used for predicting ncRNA–protein interactions are time-consuming and labor-intensive. Therefore, there is an increasing demand for computational methods to accurately and efficiently predict ncRNA–protein interactions.

Results

In this work, we presented an ensemble deep learning-based method, EDLMFC, to predict ncRNA–protein interactions using the combination of multi-scale features, including primary sequence features, secondary structure sequence features, and tertiary structure features. Conjoint k-mer was used to extract protein/ncRNA sequence features, integrating tertiary structure features, then fed into an ensemble deep learning model, which combined convolutional neural network (CNN) to learn dominating biological information with bi-directional long short-term memory network (BLSTM) to capture long-range dependencies among the features identified by the CNN. Compared with other state-of-the-art methods under five-fold cross-validation, EDLMFC shows the best performance with accuracy of 93.8%, 89.7%, and 86.1% on RPI1807, NPInter v2.0, and RPI488 datasets, respectively. The results of the independent test demonstrated that EDLMFC can effectively predict potential ncRNA–protein interactions from different organisms. Furtherly, EDLMFC is also shown to predict hub ncRNAs and proteins presented in ncRNA–protein networks of Mus musculus successfully.

Conclusions

In general, our proposed method EDLMFC improved the accuracy of ncRNA–protein interaction predictions and anticipated providing some helpful guidance on ncRNA functions research. The source code of EDLMFC and the datasets used in this work are available at https://github.com/JingjingWang-87/EDLMFC.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-021-04069-9.

Automatic Detection for Multi-Labeled Cardiac Arrhythmia Based on Frame Blocking Preprocessing and Residual Networks

Introduction: Electrocardiograms (ECG) provide information about the electrical activity of the heart, which is useful for diagnosing abnormal cardiac functions such as arrhythmias. Recently, several algorithms based on advanced structures of neural networks have been proposed for auto-detecting cardiac arrhythmias, but their performance still needs to be further improved. This study aimed to develop an auto-detection algorithm, which extracts valid features from 12-lead ECG for classifying multiple types of cardiac states. Method: The proposed algorithm consists of the following components: (i) a preprocessing component that utilizes the frame blocking method to split an ECG recording into frames with a uniform length for all considered ECG recordings; and (ii) a binary classifier based on ResNet, which is combined with the attention-based bidirectional long-short term memory model. Result: The developed algorithm was trained and tested on ECG data of nine types of cardiac states, fulfilling a task of multi-label classification. It achieved an averaged F1-score and area under the curve at 0.908 and 0.974, respectively. Conclusion: The frame blocking and bidirectional long-short term memory model represented an improved algorithm compared with others in the literature for auto-detecting and classifying multi-types of cardiac abnormalities.

TMP- SSurface2: A Novel Deep Learning-Based Surface Accessibility Predictor for Transmembrane Protein Sequence

Transmembrane protein (TMP) is an important type of membrane protein that is involved in various biological membranes related biological processes. As major drug targets, TMPs’ surfaces are highly concerned to form the structural biases of their material-bindings for drugs or other biological molecules. However, the quantity of determinate TMP structures is still far less than the requirements, while artificial intelligence technologies provide a promising approach to accurately identify the TMP surfaces, merely depending on their sequences without any feature-engineering. For this purpose, we present an updated TMP surface residue predictor TMP-SSurface2 which achieved an even higher prediction accuracy compared to our previous version. The method uses an attention-enhanced Bidirectional Long Short Term Memory (BiLSTM) network, benefiting from its efficient learning capability, some useful latent information is abstracted from protein sequences, thus improving the Pearson correlation coefficients (CC) value performance of the old version from 0.58 to 0.66 on an independent test dataset. The results demonstrate that TMP-SSurface2 is efficient in predicting the surface of transmembrane proteins, representing new progress in transmembrane protein structure modeling based on primary sequences. TMP-SSurface2 is freely accessible at https://github.com/NENUBioCompute/TMP-SSurface-2.0.

Prediction of Function Determining and Buried Residues Through Analysis of Saturation Mutagenesis Datasets

Mutational scanning can be used to probe effects of large numbers of point mutations on protein function. Positions affected by mutation are primarily at either buried or at exposed residues directly involved in function, hereafter designated as active-site residues. In the absence of prior structural information, it has not been easy to distinguish between these two categories of residues. We curated and analyzed a set of twelve published deep mutational scanning datasets. The analysis revealed differential patterns of mutational sensitivity and substitution preferences at buried and exposed positions. Prediction of buried-sites solely from the mutational sensitivity data was facilitated by incorporating predicted sequence-based accessibility values. For active-site residues we observed mean sensitivity, specificity and accuracy of 61, 90 and 88% respectively. For buried residues the corresponding figures were 59, 90 and 84% while for exposed non active-site residues these were 98, 44 and 82% respectively. We also identified positions which did not follow these general trends and might require further experimental re-validation. This analysis highlights the ability of deep mutational scans to provide important structural and functional insights, even in the absence of three-dimensional structures determined using conventional structure determination techniques, and also discuss some limitations of the methodology.

QDeep: distance-based protein model quality estimation by residue-level ensemble error classifications using stacked deep residual neural networks

MOTIVATION

Protein model quality estimation, in many ways, informs protein structure prediction. Despite their tight coupling, existing model quality estimation methods do not leverage inter-residue distance information or the latest technological breakthrough in deep learning that has recently revolutionized protein structure prediction.

RESULTS

We present a new distance-based single-model quality estimation method called QDeep by harnessing the power of stacked deep residual neural networks (ResNets). Our method first employs stacked deep ResNets to perform residue-level ensemble error classifications at multiple predefined error thresholds, and then combines the predictions from the individual error classifiers for estimating the quality of a protein structural model. Experimental results show that our method consistently outperforms existing state-of-the-art methods including ProQ2, ProQ3, ProQ3D, ProQ4, 3DCNN, MESHI, and VoroMQA in multiple independent test datasets across a wide-range of accuracy measures; and that predicted distance information significantly contributes to the improved performance of QDeep.

AVAILABILITY AND IMPLEMENTATION

https://github.com/Bhattacharya-Lab/QDeep.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Applications of Protein Secondary Structure Algorithms in SARS-CoV-2 Research

Since the outset of COVID-19, the pandemic has prompted immediate global efforts to sequence SARS-CoV-2, and over 450 000 complete genomes have been publicly deposited over the course of 12 months. Despite this, comparative nucleotide and amino acid sequence analyses often fall short in answering key questions in vaccine design. For example, the binding affinity between different ACE2 receptors and SARS-COV-2 spike protein cannot be fully explained by amino acid similarity at ACE2 contact sites because protein structure similarities are not fully reflected by amino acid sequence similarities. To comprehensively compare protein homology, secondary structure (SS) analysis is required. While protein structure is slow and difficult to obtain, SS predictions can be made rapidly, and a well-predicted SS structure may serve as a viable proxy to gain biological insight. Here we review algorithms and information used in predicting protein SS to highlight its potential application in pandemics research. We also showed examples of how SS predictions can be used to compare ACE2 proteins and to evaluate the zoonotic origins of viruses. As computational tools are much faster than wet-lab experiments, these applications can be important for research especially in times when quickly obtained biological insights can help in speeding up response to pandemics.

Hybridized distance- and contact-based hierarchical structure modeling for folding soluble and membrane proteins

Crystallography and NMR system (CNS) is currently a widely used method for fragment-free ab initio protein folding from inter-residue distance or contact maps. Despite its widespread use in protein structure prediction, CNS is a decade-old macromolecular structure determination system that was originally developed for solving macromolecular geometry from experimental restraints as opposed to predictive modeling driven by interaction map data. As such, the adaptation of the CNS experimental structure determination protocol for ab initio protein folding is intrinsically anomalous that may undermine the folding accuracy of computational protein structure prediction. In this paper, we propose a new CNS-free hierarchical structure modeling method called DConStruct for folding both soluble and membrane proteins driven by distance and contact information. Rigorous experimental validation shows that DConStruct attains much better reconstruction accuracy than CNS when tested with the same input contact map at varying contact thresholds. The hierarchical modeling with iterative self-correction employed in DConStruct scales at a much higher degree of folding accuracy than CNS with the increase in contact thresholds, ultimately approaching near-optimal reconstruction accuracy at higher-thresholded contact maps. The folding accuracy of DConStruct can be further improved by exploiting distance-based hybrid interaction maps at tri-level thresholding, as demonstrated by the better performance of our method in folding free modeling targets from the 12th and 13th rounds of the Critical Assessment of techniques for protein Structure Prediction (CASP) experiments compared to popular CNS- and fragment-based approaches and energy-minimization protocols, some of which even using much finer-grained distance maps than ours. Additional large-scale benchmarking shows that DConStruct can significantly improve the folding accuracy of membrane proteins compared to a CNS-based approach. These results collectively demonstrate the feasibility of greatly improving the accuracy of ab initio protein folding by optimally exploiting the information encoded in inter-residue interaction maps beyond what is possible by CNS.

Predicting the folded and functional 3-dimensional structure of a protein molecule from its amino acid sequence is of central importance to structural biology. Recently, promising advances have been made in ab initio protein folding due to the reasonably accurate estimation of inter-residue interaction maps at increasingly higher resolutions that range from binary contacts to finer-grained distances. Despite the progress in predicting the interaction maps, approaches for turning the residue-residue interactions projected in these maps into their precise spatial positioning heavily rely on a decade-old experimental structure determination protocol that is not suitable for predictive modeling. This paper presents a new hierarchical structure modeling method, DConStruct, which can better exploit the information encoded in the interaction maps at multiple granularities, from binary contact maps to distance-based hybrid maps at tri-level thresholding, for improved ab initio folding. Multiple large-scale benchmarking experiments show that our proposed method can substantially improve the folding accuracy for both soluble and membrane proteins compared to state-of-the-art approaches. DConStruct is licensed under the GNU General Public License v3 and freely available at https://github.com/Bhattacharya-Lab/DConStruct.

Secondary Structures of Proteins: A Comparison of Models and Experimental Results

Secondary structure predictions of proteins were compared to experimental results by wide-line ¹H NMR. IUPred2A was used to generate predictions of disordered protein or binding regions. Thymosin-β₄ and the stabilin-2 cytoplasmic domain were found to be mainly disordered, in agreement with the experimental results. α-Synuclein variants were predicted to be disordered, as in the experiments, but the A53T mutant showed less predicted disorder, in contrast with the wide-line ¹H NMR result. A disordered binding site was found for thymosin-β₄, whereas the stabilin-2 cytoplasmic domain was indicated as such in its entire length. The last third of the α-synuclein variant's sequence was a disordered binding site. Thymosin-β₄ and the stabilin-2 cytoplasmic domain contained only coils and helices according to five secondary structure prediction methods (SPIDER3-SPOT-1D, PSRSM, MUFold-SSW, Porter 5, and RaptorX). β-Sheets are present in α-synucleins, and they extend to more amino acid residues in the A53T mutant according to the predictions. The latter is verified by experiments. The comparison of the predictions with the experiments suggests that helical parts are buried.

Comprehensive Study on Enhancing Low-Quality Position-Specific Scoring Matrix with Deep Learning for Accurate Protein Structure Property Prediction: Using Bagging Multiple Sequence Alignment Learning

Accurate predictions of protein structure properties, for example, secondary structure and solvent accessibility, are essential in analyzing the structure and function of a protein. Position-specific scoring matrix (PSSM) features are widely used in the structure property prediction. However, some proteins may have low-quality PSSM features due to insufficient homologous sequences, leading to limited prediction accuracy. To address this limitation, we propose an enhancing scheme for PSSM features. We introduce the "Bagging MSA" (multiple sequence alignment) method to calculate PSSM features used to train our model, adopt a convolutional network to capture local context features and bidirectional long short-term memory for long-term dependencies, and integrate them under an unsupervised framework. Structure property prediction models are then built upon such enhanced PSSM features for more accurate predictions. Moreover, we develop two frameworks to evaluate the effectiveness of the enhanced PSSM features, which also bring proposed method into real-world scenarios. Empirical evaluation of CB513, CASP11, and CASP12 data sets indicates that our unsupervised enhancing scheme indeed generates more informative PSSM features for structure property prediction.

SAINT: self-attention augmented inception-inside-inception network improves protein secondary structure prediction

MOTIVATION

Protein structures provide basic insight into how they can interact with other proteins, their functions and biological roles in an organism. Experimental methods (e.g. X-ray crystallography and nuclear magnetic resonance spectroscopy) for predicting the secondary structure (SS) of proteins are very expensive and time consuming. Therefore, developing efficient computational approaches for predicting the SS of protein is of utmost importance. Advances in developing highly accurate SS prediction methods have mostly been focused on 3-class (Q3) structure prediction. However, 8-class (Q8) resolution of SS contains more useful information and is much more challenging than the Q3 prediction.

RESULTS

We present SAINT, a highly accurate method for Q8 structure prediction, which incorporates self-attention mechanism (a concept from natural language processing) with the Deep Inception-Inside-Inception network in order to effectively capture both the short- and long-range interactions among the amino acid residues. SAINT offers a more interpretable framework than the typical black-box deep neural network methods. Through an extensive evaluation study, we report the performance of SAINT in comparison with the existing best methods on a collection of benchmark datasets, namely, TEST2016, TEST2018, CASP12 and CASP13. Our results suggest that self-attention mechanism improves the prediction accuracy and outperforms the existing best alternate methods. SAINT is the first of its kind and offers the best known Q8 accuracy. Thus, we believe SAINT represents a major step toward the accurate and reliable prediction of SSs of proteins.

AVAILABILITY AND IMPLEMENTATION

SAINT is freely available as an open-source project at https://github.com/SAINTProtein/SAINT.

TMPSS: A Deep Learning-Based Predictor for Secondary Structure and Topology Structure Prediction of Alpha-Helical Transmembrane Proteins

Alpha transmembrane proteins (αTMPs) profoundly affect many critical biological processes and are major drug targets due to their pivotal protein functions. At present, even though the non-transmembrane secondary structures are highly relevant to the biological functions of αTMPs along with their transmembrane structures, they have not been unified to be studied yet. In this study, we present a novel computational method, TMPSS, to predict the secondary structures in non-transmembrane parts and the topology structures in transmembrane parts of αTMPs. TMPSS applied a Convolutional Neural Network (CNN), combined with an attention-enhanced Bidirectional Long Short-Term Memory (BiLSTM) network, to extract the local contexts and long-distance interdependencies from primary sequences. In addition, a multi-task learning strategy was used to predict the secondary structures and the transmembrane helixes. TMPSS was thoroughly trained and tested against a non-redundant independent dataset, where the Q3 secondary structure prediction accuracy achieved 78% in the non-transmembrane region, and the accuracy of the transmembrane region prediction achieved 90%. In sum, our method showcased a unified model for predicting the secondary structure and topology structure of αTMPs by only utilizing features generated from primary sequences and provided a steady and fast prediction, which promisingly improves the structural studies on αTMPs.

Synergistic role of nucleotides and lipids for the self-assembly of Shs1 septin oligomers

Budding yeast septins are essential for cell division and polarity. Septins assemble as palindromic linear octameric complexes. The function and ultra-structural organization of septins are finely governed by their molecular polymorphism. In particular, in budding yeast, the end subunit can stand either as Shs1 or Cdc11. We have dissected, here, for the first time, the behavior of the Shs1 protomer bound to membranes at nanometer resolution, in complex with the other septins. Using electron microscopy, we have shown that on membranes, Shs1 protomers self-assemble into rings, bundles, filaments or two-dimensional gauzes. Using a set of specific mutants we have demonstrated a synergistic role of both nucleotides and lipids for the organization and oligomerization of budding yeast septins. Besides, cryo-electron tomography assays show that vesicles are deformed by the interaction between Shs1 oligomers and lipids. The Shs1-Shs1 interface is stabilized by the presence of phosphoinositides, allowing the visualization of micrometric long filaments formed by Shs1 protomers. In addition, molecular modeling experiments have revealed a potential molecular mechanism regarding the selectivity of septin subunits for phosphoinositide lipids.

Structure-aware protein solubility prediction from sequence through graph convolutional network and predicted contact map

Protein solubility is significant in producing new soluble proteins that can reduce the cost of biocatalysts or therapeutic agents. Therefore, a computational model is highly desired to accurately predict protein solubility from the amino acid sequence. Many methods have been developed, but they are mostly based on the one-dimensional embedding of amino acids that is limited to catch spatially structural information. In this study, we have developed a new structure-aware method GraphSol to predict protein solubility by attentive graph convolutional network (GCN), where the protein topology attribute graph was constructed through predicted contact maps only from the sequence. GraphSol was shown to substantially outperform other sequence-based methods. The model was proven to be stable by consistent [Formula: see text] of 0.48 in both the cross-validation and independent test of the eSOL dataset. To our best knowledge, this is the first study to utilize the GCN for sequence-based protein solubility predictions. More importantly, this architecture could be easily extended to other protein prediction tasks requiring a raw protein sequence.

Dimerisation of the PICTS complex via LC8/Cut-up drives co-transcriptional transposon silencing in Drosophila

In animal gonads, the PIWI-interacting RNA (piRNA) pathway guards genome integrity in part through the co-transcriptional gene silencing of transposon insertions. In Drosophila ovaries, piRNA-loaded Piwi detects nascent transposon transcripts and instructs heterochromatin formation through the Panoramix-induced co-transcriptional silencing (PICTS) complex, containing Panoramix, Nxf2 and Nxt1. Here, we report that the highly conserved dynein light chain LC8/Cut-up (Ctp) is an essential component of the PICTS complex. Loss of Ctp results in transposon de-repression and a reduction in repressive chromatin marks specifically at transposon loci. In turn, Ctp can enforce transcriptional silencing when artificially recruited to RNA and DNA reporters. We show that Ctp drives dimerisation of the PICTS complex through its interaction with conserved motifs within Panoramix. Artificial dimerisation of Panoramix bypasses the necessity for its interaction with Ctp, demonstrating that conscription of a protein from a ubiquitous cellular machinery has fulfilled a fundamental requirement for a transposon silencing complex.

OCLSTM: Optimized convolutional and long short-term memory neural network model for protein secondary structure prediction

Protein secondary structure prediction is extremely important for determining the spatial structure and function of proteins. In this paper, we apply an optimized convolutional neural network and long short-term memory neural network models to protein secondary structure prediction, which is called OCLSTM. We use an optimized convolutional neural network to extract local features between amino acid residues. Then use the bidirectional long short-term memory neural network to extract the remote interactions between the internal residues of the protein sequence to predict the protein structure. Experiments are performed on CASP10, CASP11, CASP12, CB513, and 25PDB datasets, and the good performance of 84.68%, 82.36%, 82.91%, 84.21% and 85.08% is achieved respectively. Experimental results show that the model can achieve better results.

Broad specificity of immune helminth scFv library to identify monoclonal antibodies targeting Strongyloides

Antibodies have different chemical properties capable of targeting a diverse nature of antigens. Traditionally, immune antibody libraries are perceived to be disease-specific with a skewed repertoire. The complexity during the generation of a combinatorial antibody library allows for a skewed but diverse repertoire to be generated. Strongyloides stercoralis is a parasite that causes strongyloidiasis, a potentially life-threatening disease with a complex diagnosis that impedes effective control and treatment of the disease. This study describes the isolation of monoclonal antibodies against S. stercoralis NIE recombinant protein using an immune antibody phage display library derived from lymphatic filaria-infected individuals. The isolated antibody clones showed both lambda and kappa light chains gene usage, with diverse amino acid distributions. Structural analysis showed that electropositivity and the interface area could determine the binding affinity of the clones with NIE. The successful identification of S. stercoralis antibodies from the filarial immune library highlights the breadth of antibody gene diversification in an immune antibody library that can be applied for closely related infections.

Probabilistic divergence of a template-based modelling methodology from the ideal protocol

Protein structural information is essential for the detailed mapping of a functional protein network. For a higher modelling accuracy and quicker implementation, template-based algorithms have been extensively deployed and redefined. The methods only assess the predicted structure against its native state/template and do not estimate the accuracy for each modelling step. A divergence measure is therefore postulated to estimate the modelling accuracy against its theoretical optimal benchmark. By freezing the domain boundaries, the divergence measures are predicted for the most crucial steps of a modelling algorithm. To precisely refine the score using weighting constants, big data analysis could further be deployed.

PolyFold: An interactive visual simulator for distance-based protein folding

Recent advances in distance-based protein folding have led to a paradigm shift in protein structure prediction. Through sufficiently precise estimation of the inter-residue distance matrix for a protein sequence, it is now feasible to predict the correct folds for new proteins much more accurately than ever before. Despite the exciting progress, a dedicated visualization system that can dynamically capture the distance-based folding process is still lacking. Most molecular visualizers typically provide only a static view of a folded protein conformation, but do not capture the folding process. Even among the selected few graphical interfaces that do adopt a dynamic perspective, none of them are distance-based. Here we present PolyFold, an interactive visual simulator for dynamically capturing the distance-based protein folding process through real-time rendering of a distance matrix and its compatible spatial conformation as it folds in an intuitive and easy-to-use interface. PolyFold integrates highly convergent stochastic optimization algorithms with on-demand customizations and interactive manipulations to maximally satisfy the geometric constraints imposed by a distance matrix. PolyFold is capable of simulating the complex process of protein folding even on modest personal computers, thus making it accessible to the general public for fostering citizen science. Open source code of PolyFold is freely available for download at https://github.com/Bhattacharya-Lab/PolyFold. It is implemented in cross-platform Java and binary executables are available for macOS, Linux, and Windows.

SSSCPreds: Deep Neural Network-Based Software for the Prediction of Conformational Variability and Application to SARS-CoV-2

Amino acid mutations that improve protein stability and rigidity can accompany increases in binding affinity. Therefore, conserved amino acids located on a protein surface may be successfully targeted by antibodies. The quantitative deep mutational scanning approach is an excellent technique to understand viral evolution, and the obtained data can be utilized to develop a vaccine. However, the application of the approach to all of the proteins in general is difficult in terms of cost. To address this need, we report the construction of a deep neural network-based program for sequence-based prediction of supersecondary structure codes (SSSCs), called SSSCPrediction (SSSCPred). Further, to predict conformational flexibility or rigidity in proteins, a comparison program called SSSCPreds that consists of three deep neural network-based prediction systems (SSSCPred, SSSCPred100, and SSSCPred200) has also been developed. Using our algorithms we calculated here shows the degree of flexibility for the receptor-binding motif of SARS-CoV-2 spike protein and the rigidity of the unique motif (SSSC: SSSHSSHHHH) at the S2 subunit and has a value independent of the X-ray and Cryo-EM structures. The fact that the sequence flexibility/rigidity map of SARS-CoV-2 RBD resembles the sequence-to-phenotype maps of ACE2-binding affinity and expression, which were experimentally obtained by deep mutational scanning, suggests that the identical SSSC sequences among the ones predicted by three deep neural network-based systems correlate well with the sequences with both lower ACE2-binding affinity and lower expression. The combined analysis of predicted and observed SSSCs with keyword-tagged datasets would be helpful in understanding the structural correlation to the examined system.

HACS1 signaling adaptor protein recognizes a motif in the paired immunoglobulin receptor B cytoplasmic domain

Hematopoietic adaptor containing SH3 and SAM domains-1 (HACS1) is a signaling protein with two juxtaposed protein–protein interaction domains and an intrinsically unstructured region that spans half the sequence. Here, we describe the interaction between the HACS1 SH3 domain and a sequence near the third immunoreceptor tyrosine-based inhibition motif (ITIM3) of the paired immunoglobulin receptor B (PIRB). From surface plasmon resonance binding assays using a mouse and human PIRB ITIM3 phosphopeptides as ligands, the HACS1 SH3 domain and SHP2 N-terminal SH2 domain demonstrated comparable affinities in the micromolar range. Since the PIRB ITIM3 sequence represents an atypical ligand for an SH3 domain, we determined the NMR structure of the HACS1 SH3 domain and performed a chemical shift mapping study. This study showed that the binding site on the HACS1 SH3 domain for PIRB shares many of the same amino acids found in a canonical binding cleft normally associated with polyproline ligands. Molecular modeling suggests that the respective binding sites in PIRB ITIM3 for the HACS1 SH3 domain and the SHP2 SH2 domain are too close to permit simultaneous binding. As a result, the HACS1-PIRB partnership has the potential to amalgamate signaling pathways that influence both immune and neuronal cell fate.

Kwan et al. show the interaction between the HACS1 SH3 domain and a sequence near the third immunoreceptor tyrosine-based inhibition motif of the Paired immunoglobulin receptor B (PIRB). This study suggests that the HACS1-PIRB partnership has the potential to unite signaling pathways that regulate both immune and neuronal cell fate.

Enhancing protein backbone angle prediction by using simpler models of deep neural networks

Protein structure prediction is a grand challenge. Prediction of protein structures via the representations using backbone dihedral angles has recently achieved significant progress along with the on-going surge of deep neural network (DNN) research in general. However, we observe that in the protein backbone angle prediction research, there is an overall trend to employ more and more complex neural networks and then to throw more and more features to the neural networks. While more features might add more predictive power to the neural network, we argue that redundant features could rather clutter the scenario and more complex neural networks then just could counterbalance the noise. From artificial intelligence and machine learning perspectives, problem representations and solution approaches do mutually interact and thus affect performance. We also argue that comparatively simpler predictors can more easily be reconstructed than the more complex ones. With these arguments in mind, we present a deep learning method named Simpler Angle Predictor (SAP) to train simpler DNN models that enhance protein backbone angle prediction. We then empirically show that SAP significantly outperforms existing state-of-the-art methods on well-known benchmark datasets: for some types of angles, the differences are above 3 in mean absolute error (MAE). The SAP program along with its data is available from the website https://gitlab.com/mahnewton/sap.

Amino Acid Encoding Methods for Protein Sequences: A Comprehensive Review and Assessment

As the first step of machine-learning based protein structure and function prediction, the amino acid encoding play a fundamental role in the final success of those methods. Different from the protein sequence encoding, the amino acid encoding can be used in both residue-level and sequence-level prediction of protein properties by combining them with different algorithms. However, it has not attracted enough attention in the past decades, and there are no comprehensive reviews and assessments about encoding methods so far. In this article, we make a systematic classification and propose a comprehensive review and assessment for various amino acid encoding methods. Those methods are grouped into five categories according to their information sources and information extraction methodologies, including binary encoding, physicochemical properties encoding, evolution-based encoding, structure-based encoding, and machine-learning encoding. Then, 16 representative methods from five categories are selected and compared on protein secondary structure prediction and protein fold recognition tasks by using large-scale benchmark datasets. The results show that the evolution-based position-dependent encoding method PSSM achieved the best performance, and the structure-based and machine-learning encoding methods also show some potential for further application, the neural network based distributed representation of amino acids in particular may bring new light to this area. We hope that the review and assessment are useful for future studies in amino acid encoding.

Deep Learning in Proteomics

Proteomics, the study of all the proteins in biological systems, is becoming a data-rich science. Protein sequences and structures are comprehensively catalogued in online databases. With recent advancements in tandem mass spectrometry (MS) technology, protein expression and post-translational modifications (PTMs) can be studied in a variety of biological systems at the global scale. Sophisticated computational algorithms are needed to translate the vast amount of data into novel biological insights. Deep learning automatically extracts data representations at high levels of abstraction from data, and it thrives in data-rich scientific research domains. Here, a comprehensive overview of deep learning applications in proteomics, including retention time prediction, MS/MS spectrum prediction, de novo peptide sequencing, PTM prediction, major histocompatibility complex-peptide binding prediction, and protein structure prediction, is provided. Limitations and the future directions of deep learning in proteomics are also discussed. This review will provide readers an overview of deep learning and how it can be used to analyze proteomics data.

Predicting Hot Spot Residues at Protein-DNA Binding Interfaces Based on Sequence Information

Hot spot residues at protein-DNA binding interfaces are hugely important for investigating the underlying mechanism of molecular recognition. Currently, there are a few tools available for identifying the hot spot residues in the protein-DNA complexes. In addition, the three-dimensional protein structures are needed in these tools. However, it is well known that the three-dimensional structures are unavailable for most proteins. Considering the limitation, we proposed a method, named SPDH, for predicting hot spot residues only based on protein sequences. Firstly, we obtained 133 features from physicochemical property, conservation, predicted solvent accessible surface area and structure. Then, we systematically assessed these features based on various feature selection methods to obtain the optimal feature subset and compared the models using four classical machine learning algorithms (support vector machine, random forest, logistic regression, and k-nearest neighbor) on the training dataset. We found that the variability of physicochemical property features between wild and mutative types was important on improving the performance of the prediction model. On the independent test set, our method achieved the performance with AUC of 0.760 and sensitivity of 0.808, and outperformed other methods. The data and source code can be downloaded at https://github.com/xialab-ahu/SPDH .

ProteinUnet-An efficient alternative to SPIDER3-single for sequence-based prediction of protein secondary structures

Predicting protein function and structure from sequence remains an unsolved problem in bioinformatics. The best performing methods rely heavily on evolutionary information from multiple sequence alignments, which means their accuracy deteriorates for sequences with a few homologs, and given the increasing sequence database sizes requires long computation times. Here, a single‐sequence‐based prediction method is presented, called ProteinUnet, leveraging an U‐Net convolutional network architecture. It is compared to SPIDER3‐Single model, based on long short‐term memory‐bidirectional recurrent neural networks architecture. Both methods achieve similar results for prediction of secondary structures (both three‐ and eight‐state), half‐sphere exposure, and contact number, but ProteinUnet has two times fewer parameters, 17 times shorter inference time, and can be trained 11 times faster. Moreover, ProteinUnet tends to be better for short sequences and residues with a low number of local contacts. Additionally, the method of loss weighting is presented as an effective way of increasing accuracy for rare secondary structures.

Sequence-to-function deep learning frameworks for engineered riboregulators

While synthetic biology has revolutionized our approaches to medicine, agriculture, and energy, the design of completely novel biological circuit components beyond naturally-derived templates remains challenging due to poorly understood design rules. Toehold switches, which are programmable nucleic acid sensors, face an analogous design bottleneck; our limited understanding of how sequence impacts functionality often necessitates expensive, time-consuming screens to identify effective switches. Here, we introduce Sequence-based Toehold Optimization and Redesign Model (STORM) and Nucleic-Acid Speech (NuSpeak), two orthogonal and synergistic deep learning architectures to characterize and optimize toeholds. Applying techniques from computer vision and natural language processing, we 'un-box' our models using convolutional filters, attention maps, and in silico mutagenesis. Through transfer-learning, we redesign sub-optimal toehold sensors, even with sparse training data, experimentally validating their improved performance. This work provides sequence-to-function deep learning frameworks for toehold selection and design, augmenting our ability to construct potent biological circuit components and precision diagnostics.

Protein profiles: Biases and protocols

The use of evolutionary profiles to predict protein secondary structure, as well as other protein structural features, has been standard practice since the 1990s. Using profiles in the input of such predictors, in place or in addition to the sequence itself, leads to significantly more accurate predictions. While profiles can enhance structural signals, their role remains somewhat surprising as proteins do not use profiles when folding in vivo. Furthermore, the same sequence-based redundancy reduction protocols initially derived to train and evaluate sequence-based predictors, have been applied to train and evaluate profile-based predictors. This can lead to unfair comparisons since profiles may facilitate the bleeding of information between training and test sets. Here we use the extensively studied problem of secondary structure prediction to better evaluate the role of profiles and show that: (1) high levels of profile similarity between training and test proteins are observed when using standard sequence-based redundancy protocols; (2) the gain in accuracy for profile-based predictors, over sequence-based predictors, strongly relies on these high levels of profile similarity between training and test proteins; and (3) the overall accuracy of a profile-based predictor on a given protein dataset provides a biased measure when trying to estimate the actual accuracy of the predictor, or when comparing it to other predictors. We show, however, that this bias can be mitigated by implementing a new protocol (EVALpro) which evaluates the accuracy of profile-based predictors as a function of the profile similarity between training and test proteins. Such a protocol not only allows for a fair comparison of the predictors on equally hard or easy examples, but also reduces the impact of choosing a given similarity cutoff when selecting test proteins. The EVALpro program is available in the SCRATCH suite ( www.scratch.proteomics.ics.uci.edu) and can be downloaded at: www.download.igb.uci.edu/#evalpro.

Predicting the Real-Valued Inter-Residue Distances for Proteins

Predicting protein structure from the amino acid sequence has been a challenge with theoretical and practical significance in biophysics. Despite the recent progresses elicited by improved inter-residue contact prediction, contact-based structure prediction has gradually reached the performance ceiling. New methods have been proposed to predict the inter-residue distance, but unanimously by simplifying the real-valued distance prediction into a multiclass classification problem. Here, a lightweight regression-based distance prediction method is shown, which adopts the generative adversarial network to capture the delicate geometric relationship between residue pairs and thus could predict the continuous, real-valued inter-residue distance rapidly and satisfactorily. The predicted residue distance map allows quick structure modeling by the CNS suite, and the constructed models approach the same level of quality as the other state-of-the-art protein structure prediction methods when tested on CASP13 targets. Moreover, this method can be used directly for the structure prediction of membrane proteins without transfer learning.

Computationally identifying hot spots in protein-DNA binding interfaces using an ensemble approach

BACKGROUND

Protein-DNA interaction governs a large number of cellular processes, and it can be altered by a small fraction of interface residues, i.e., the so-called hot spots, which account for most of the interface binding free energy. Accurate prediction of hot spots is critical to understand the principle of protein-DNA interactions. There are already some computational methods that can accurately and efficiently predict a large number of hot residues. However, the insufficiency of experimentally validated hot-spot residues in protein-DNA complexes and the low diversity of the employed features limit the performance of existing methods.

RESULTS

Here, we report a new computational method for effectively predicting hot spots in protein-DNA binding interfaces. This method, called PreHots (the abbreviation of Predicting Hotspots), adopts an ensemble stacking classifier that integrates different machine learning classifiers to generate a robust model with 19 features selected by a sequential backward feature selection algorithm. To this end, we constructed two new and reliable datasets (one benchmark for model training and one independent dataset for validation), which totally consist of 123 hot spots and 137 non-hot spots from 89 protein-DNA complexes. The data were manually collected from the literature and existing databases with a strict process of redundancy removal. Our method achieves a sensitivity of 0.813 and an AUC score of 0.868 in 10-fold cross-validation on the benchmark dataset, and a sensitivity of 0.818 and an AUC score of 0.820 on the independent test dataset. The results show that our approach outperforms the existing ones.

CONCLUSIONS

PreHots, which is based on stack ensemble of boosting algorithms, can reliably predict hot spots at the protein-DNA binding interface on a large scale. Compared with the existing methods, PreHots can achieve better prediction performance. Both the webserver of PreHots and the datasets are freely available at: http://dmb.tongji.edu.cn/tools/PreHots/ .

DNSS2: Improved ab initio protein secondary structure prediction using advanced deep learning architectures

Accurate prediction of protein secondary structure (alpha-helix, beta-strand and coil) is a crucial step for protein inter-residue contact prediction and ab initio tertiary structure prediction. In a previous study, we developed a deep belief network-based protein secondary structure method (DNSS1) and successfully advanced the prediction accuracy beyond 80%. In this work, we developed multiple advanced deep learning architectures (DNSS2) to further improve secondary structure prediction. The major improvements over the DNSS1 method include (a) designing and integrating six advanced one-dimensional deep convolutional/recurrent/residual/memory/fractal/inception networks to predict 3-state and 8-state secondary structure, and (b) using more sensitive profile features inferred from Hidden Markov model (HMM) and multiple sequence alignment (MSA). Most of the deep learning architectures are novel for protein secondary structure prediction. DNSS2 was systematically benchmarked on independent test data sets with eight state-of-art tools and consistently ranked as one of the best methods. Particularly, DNSS2 was tested on the protein targets of 2018 CASP13 experiment and achieved the Q3 score of 81.62%, SOV score of 72.19%, and Q8 score of 73.28%. DNSS2 is freely available at: https://github.com/multicom-toolbox/DNSS2.

Identifying molecular recognition features in intrinsically disordered regions of proteins by transfer learning

MOTIVATION

Protein intrinsic disorder describes the tendency of sequence residues to not fold into a rigid three-dimensional shape by themselves. However, some of these disordered regions can transition from disorder to order when interacting with another molecule in segments known as molecular recognition features (MoRFs). Previous analysis has shown that these MoRF regions are indirectly encoded within the prediction of residue disorder as low-confidence predictions [i.e. in a semi-disordered state P(D)≈0.5]. Thus, what has been learned for disorder prediction may be transferable to MoRF prediction. Transferring the internal characterization of protein disorder for the prediction of MoRF residues would allow us to take advantage of the large training set available for disorder prediction, enabling the training of larger analytical models than is currently feasible on the small number of currently available annotated MoRF proteins. In this paper, we propose a new method for MoRF prediction by transfer learning from the SPOT-Disorder2 ensemble models built for disorder prediction.

RESULTS

We confirm that directly training on the MoRF set with a randomly initialized model yields substantially poorer performance on independent test sets than by using the transfer-learning-based method SPOT-MoRF, for both deep and simple networks. Its comparison to current state-of-the-art techniques reveals its superior performance in identifying MoRF binding regions in proteins across two independent testing sets, including our new dataset of >800 protein chains. These test chains share <30% sequence similarity to all training and validation proteins used in SPOT-Disorder2 and SPOT-MoRF, and provide a much-needed large-scale update on the performance of current MoRF predictors. The method is expected to be useful in locating functional disordered regions in proteins.

AVAILABILITY AND IMPLEMENTATION

SPOT-MoRF and its data are available as a web server and as a standalone program at: http://sparks-lab.org/jack/server/SPOT-MoRF/index.php.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Deep Robust Framework for Protein Function Prediction Using Variable-Length Protein Sequences

The order of amino acids in a protein sequence enables the protein to acquire a conformation suitable for performing functions, thereby motivating the need to analyze these sequences for predicting functions. Although machine learning based approaches are fast compared to methods using BLAST, FASTA, etc., they fail to perform well for long protein sequences (with more than 300 amino acids). In this paper, we introduce a novel method for construction of two separate feature sets for protein using bi-directional long short-term memory network based on the analysis of fixed 1) single-sized segments and 2) multi-sized segments. The model trained on the proposed feature set based on multi-sized segments is combined with the model trained using state-of-the-art Multi-label Linear Discriminant Analysis (MLDA) features to further improve the accuracy. Extensive evaluations using separate datasets for biological processes and molecular functions demonstrate not only improved results for long sequences, but also significantly improve the overall accuracy over state-of-the-art method. The single-sized approach produces an improvement of +3.37 percent for biological processes and +5.48 percent for molecular functions over the MLDA based classifier. The corresponding numbers for multi-sized approach are +5.38 and +8.00 percent. Combining the two models, the accuracy further improves to +7.41 and +9.21 percent, respectively.

Sequence-Based Prediction of Metamorphic Behavior in Proteins

An increasing number of proteins have been demonstrated in recent years to adopt multiple three-dimensional folds with different functions. These metamorphic proteins are characterized by having two or more folds with significant differences in their secondary structure, in which each fold is stabilized by a distinct local environment. So far, ∼90 metamorphic proteins have been identified in the Protein Databank, but we and others hypothesize that a far greater number of metamorphic proteins remain undiscovered. In this work, we introduce a computational model to predict metamorphic behavior in proteins using only knowledge of the sequence. In this model, secondary structure prediction programs are used to calculate diversity indices, which are measures of uncertainty in predicted secondary structure at each position in the sequence; these are then used to assign protein sequences as likely to be metamorphic versus monomorphic (i.e., having just one fold). We constructed a reference data set to train our classification method, which includes a novel compilation of 136 likely monomorphic proteins and a set of 201 metamorphic protein structures taken from the literature. Our model is able to classify proteins as metamorphic versus monomorphic with a Matthews correlation coefficient of ∼0.36 and true positive/true negative rates of ∼65%/80%, suggesting that it is possible to predict metamorphic behavior in proteins using only sequence information.

A deep attention network for predicting amino acid signals in the formation of [Formula: see text]-helices

The secondary and tertiary structure of a protein has a primary role in determining its function. Even though many folding prediction algorithms have been developed in the past decades - mainly based on the assumption that folding instructions are encoded within the protein sequence - experimental techniques remain the most reliable to establish protein structures. In this paper, we searched for signals related to the formation of [Formula: see text]-helices. We carried out a statistical analysis on a large dataset of experimentally characterized secondary structure elements to find over- or under-occurrences of specific amino acids defining the boundaries of helical moieties. To validate our hypothesis, we trained various Machine Learning models, each equipped with an attention mechanism, to predict the occurrence of [Formula: see text]-helices. The attention mechanism allows to interpret the model's decision, weighing the importance the predictor gives to each part of the input. The experimental results show that different models focus on the same subsequences, which can be seen as codes driving the secondary structure formation.

autoBioSeqpy: A Deep Learning Tool for the Classification of Biological Sequences

Deep learning has proven to be a powerful method with applications in various fields including image, language, and biomedical data. Thanks to the libraries and toolkits such as TensorFlow, PyTorch, and Keras, researchers can use different deep learning architectures and data sets for rapid modeling. However, the available implementations of neural networks using these toolkits are usually designed for a specific research and are difficult to transfer to other work. Here, we present autoBioSeqpy, a tool that uses deep learning for biological sequence classification. The advantage of this tool is its simplicity. Users only need to prepare the input data set and then use a command line interface. Then, autoBioSeqpy automatically executes a series of customizable steps including text reading, parameter initialization, sequence encoding, model loading, training, and evaluation. In addition, the tool provides various ready-to-apply and adapt model templates to improve the usability of these networks. We introduce the application of autoBioSeqpy on three biological sequence problems: the prediction of type III secreted proteins, protein subcellular localization, and CRISPR/Cas9 sgRNA activity. autoBioSeqpy is freely available with examples at https://github.com/jingry/autoBioSeqpy.

OPUS-TASS: a protein backbone torsion angles and secondary structure predictor based on ensemble neural networks

Motivation

Predictions of protein backbone torsion angles (ϕ and ψ) and secondary structure from sequence are crucial subproblems in protein structure prediction. With the development of deep learning approaches, their accuracies have been significantly improved. To capture the long-range interactions, most studies integrate bidirectional recurrent neural networks into their models. In this study, we introduce and modify a recently proposed architecture named Transformer to capture the interactions between the two residues theoretically with arbitrary distance. Moreover, we take advantage of multitask learning to improve the generalization of neural network by introducing related tasks into the training process. Similar to many previous studies, OPUS-TASS uses an ensemble of models and achieves better results.

Results

OPUS-TASS uses the same training and validation sets as SPOT-1D. We compare the performance of OPUS-TASS and SPOT-1D on TEST2016 (1213 proteins) and TEST2018 (250 proteins) proposed in the SPOT-1D paper, CASP12 (55 proteins), CASP13 (32 proteins) and CASP-FM (56 proteins) proposed in the SAINT paper, and a recently released PDB structure collection from CAMEO (93 proteins) named as CAMEO93. On these six test sets, OPUS-TASS achieves consistent improvements in both backbone torsion angles prediction and secondary structure prediction. On CAMEO93, SPOT-1D achieves the mean absolute errors of 16.89 and 23.02 for ϕ and ψ predictions, respectively, and the accuracies for 3- and 8-state secondary structure predictions are 87.72 and 77.15%, respectively. In comparison, OPUS-TASS achieves 16.56 and 22.56 for ϕ and ψ predictions, and 89.06 and 78.87% for 3- and 8-state secondary structure predictions, respectively. In particular, after using our torsion angles refinement method OPUS-Refine as the post-processing procedure for OPUS-TASS, the mean absolute errors for final ϕ and ψ predictions are further decreased to 16.28 and 21.98, respectively.

Availability and implementation

The training and the inference codes of OPUS-TASS and its data are available at https://github.com/thuxugang/opus_tass.

Supplementary information

Supplementary data are available at Bioinformatics online.

DeepHelicon: Accurate prediction of inter-helical residue contacts in transmembrane proteins by residual neural networks

Accurate prediction of amino acid residue contacts is an important prerequisite for generating high-quality 3D models of transmembrane (TM) proteins. While a large number of compositional, evolutionary, and structural properties of proteins can be used to train contact prediction methods, recent research suggests that coevolution between residues provides the strongest indication of their spatial proximity. We have developed a deep learning approach, DeepHelicon, to predict inter-helical residue contacts in TM proteins by considering only coevolutionary features. DeepHelicon comprises a two-stage supervised learning process by residual neural networks for a gradual refinement of contact maps, followed by variance reduction by an ensemble of models. We present a benchmark study of 12 contact predictors and conclude that DeepHelicon together with the two other state-of-the-art methods DeepMetaPSICOV and Membrain2 outperforms the 10 remaining algorithms on all datasets and at all settings. On a set of 44 TM proteins with an average length of 388 residues DeepHelicon achieves the best performance among all benchmarked methods in predicting the top L/5 and L/2 inter-helical contacts, with the mean precision of 87.42% and 77.84%, respectively. On a set of 57 relatively small TM proteins with an average length of 298 residues DeepHelicon ranks second best after DeepMetaPSICOV. DeepHelicon produces the most accurate predictions for large proteins with more than 10 transmembrane helices. Coevolutionary features alone allow to predict inter-helical residue contacts with an accuracy sufficient for generating acceptable 3D models for up to 30% of proteins using a fully automated modeling method such as CONFOLD2.

DNN-Dom: predicting protein domain boundary from sequence alone by deep neural network

MOTIVATION

Accurate delineation of protein domain boundary plays an important role for protein engineering and structure prediction. Although machine-learning methods are widely used to predict domain boundary, these approaches often ignore long-range interactions among residues, which have been proven to improve the prediction performance. However, how to simultaneously model the local and global interactions to further improve domain boundary prediction is still a challenging problem.

RESULTS

This article employs a hybrid deep learning method that combines convolutional neural network and gate recurrent units' models for domain boundary prediction. It not only captures the local and non-local interactions, but also fuses these features for prediction. Additionally, we adopt balanced Random Forest for classification to deal with high imbalance of samples and high dimensions of deep features. Experimental results show that our proposed approach (DNN-Dom) outperforms existing machine-learning-based methods for boundary prediction. We expect that DNN-Dom can be useful for assisting protein structure and function prediction.

AVAILABILITY AND IMPLEMENTATION

The method is available as DNN-Dom Server at http://isyslab.info/DNN-Dom/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

SPRINT-Gly: predicting N- and O-linked glycosylation sites of human and mouse proteins by using sequence and predicted structural properties

MOTIVATION

Protein glycosylation is one of the most abundant post-translational modifications that plays an important role in immune responses, intercellular signaling, inflammation and host-pathogen interactions. However, due to the poor ionization efficiency and microheterogeneity of glycopeptides identifying glycosylation sites is a challenging task, and there is a demand for computational methods. Here, we constructed the largest dataset of human and mouse glycosylation sites to train deep learning neural networks and support vector machine classifiers to predict N-/O-linked glycosylation sites, respectively.

RESULTS

The method, called SPRINT-Gly, achieved consistent results between ten-fold cross validation and independent test for predicting human and mouse glycosylation sites. For N-glycosylation, a mouse-trained model performs equally well in human glycoproteins and vice versa, however, due to significant differences in O-linked sites separate models were generated. Overall, SPRINT-Gly is 18% and 50% higher in Matthews correlation coefficient than the next best method compared in N-linked and O-linked sites, respectively. This improved performance is due to the inclusion of novel structure and sequence-based features.

AVAILABILITY AND IMPLEMENTATION

http://sparks-lab.org/server/SPRINT-Gly/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

CoRNeA: A Pipeline to Decrypt the Inter-Protein Interfaces from Amino Acid Sequence Information

Decrypting the interface residues of the protein complexes provides insight into the functions of the proteins and, hence, the overall cellular machinery. Computational methods have been devised in the past to predict the interface residues using amino acid sequence information, but all these methods have been majorly applied to predict for prokaryotic protein complexes. Since the composition and rate of evolution of the primary sequence is different between prokaryotes and eukaryotes, it is important to develop a method specifically for eukaryotic complexes. Here, we report a new hybrid pipeline for predicting the protein-protein interaction interfaces in a pairwise manner from the amino acid sequence information of the interacting proteins. It is based on the framework of Co-evolution, machine learning (Random Forest), and Network Analysis named CoRNeA trained specifically on eukaryotic protein complexes. We use Co-evolution, physicochemical properties, and contact potential as major group of features to train the Random Forest classifier. We also incorporate the intra-contact information of the individual proteins to eliminate false positives from the predictions keeping in mind that the amino acid sequence of a protein also holds information for its own folding and not only the interface propensities. Our prediction on example datasets shows that CoRNeA not only enhances the prediction of true interface residues but also reduces false positive rates significantly.

Structural basis of ER-associated protein degradation mediated by the Hrd1 ubiquitin ligase complex

Misfolded luminal endoplasmic reticulum (ER) proteins undergo ER-associated degradation (ERAD-L): They are retrotranslocated into the cytosol, polyubiquitinated, and degraded by the proteasome. ERAD-L is mediated by the Hrd1 complex (composed of Hrd1, Hrd3, Der1, Usa1, and Yos9), but the mechanism of retrotranslocation remains mysterious. Here, we report a structure of the active Hrd1 complex, as determined by cryo-electron microscopy analysis of two subcomplexes. Hrd3 and Yos9 jointly create a luminal binding site that recognizes glycosylated substrates. Hrd1 and the rhomboid-like Der1 protein form two "half-channels" with cytosolic and luminal cavities, respectively, and lateral gates facing one another in a thinned membrane region. These structures, along with crosslinking and molecular dynamics simulation results, suggest how a polypeptide loop of an ERAD-L substrate moves through the ER membrane.

Multifaceted analysis of training and testing convolutional neural networks for protein secondary structure prediction

Protein secondary structure prediction remains a vital topic with broad applications. Due to lack of a widely accepted standard in secondary structure predictor evaluation, a fair comparison of predictors is challenging. A detailed examination of factors that contribute to higher accuracy is also lacking. In this paper, we present: (1) new test sets, Test2018, Test2019, and Test2018-2019, consisting of proteins from structures released in 2018 and 2019 with less than 25% identity to any protein published before 2018; (2) a 4-layer convolutional neural network, SecNet, with an input window of ±14 amino acids which was trained on proteins ≤25% identical to proteins in Test2018 and the commonly used CB513 test set; (3) an additional test set that shares no homologous domains with the training set proteins, according to the Evolutionary Classification of Proteins (ECOD) database; (4) a detailed ablation study where we reverse one algorithmic choice at a time in SecNet and evaluate the effect on the prediction accuracy; (5) new 4- and 5-label prediction alphabets that may be more practical for tertiary structure prediction methods. The 3-label accuracy (helix, sheet, coil) of the leading predictors on both Test2018 and CB513 is 81-82%, while SecNet's accuracy is 84% for both sets. Accuracy on the non-homologous ECOD set is only 0.6 points (83.9%) lower than the results on the Test2018-2019 set (84.5%). The ablation study of features, neural network architecture, and training hyper-parameters suggests the best accuracy results are achieved with good choices for each of them while the neural network architecture is not as critical as long as it is not too simple. Protocols for generating and using unbiased test, validation, and training sets are provided. Our data sets, including input features and assigned labels, and SecNet software including third-party dependencies and databases, are downloadable from dunbrack.fccc.edu/ss and github.com/sh-maxim/ss.