Assessment of contact predictions in CASP12: Co-evolution and deep learning coming of age

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/.

A systematic analysis of the beta hairpin motif in the Protein Data Bank

The beta hairpin motif is a ubiquitous protein structural motif that can be found in molecules across the tree of life. This motif, which is also popular in synthetically designed proteins and peptides, is known for its stability and adaptability to broad functions. Here, we systematically probe all 49,000 unique beta hairpin substructures contained within the Protein Data Bank (PDB) to uncover key characteristics correlated with stable beta hairpin structure, including amino acid biases and enriched interstrand contacts. We find that position specific amino acid preferences, while seen throughout the beta hairpin structure, are most evident within the turn region, where they depend on subtle turn dynamics associated with turn length and secondary structure. We also establish a set of broad design principles, such as the inclusion of aspartic acid residues at a specific position and the careful consideration of desired secondary structure when selecting residues for the turn region, that can be applied to the generation of libraries encoding proteins or peptides containing beta hairpin structures.

DISTEVAL: a web server for evaluating predicted protein distances

Background

Protein inter-residue contact and distance prediction are two key intermediate steps essential to accurate protein structure prediction. Distance prediction comes in two forms: real-valued distances and ‘binned’ distograms, which are a more finely grained variant of the binary contact prediction problem. The latter has been introduced as a new challenge in the 14th Critical Assessment of Techniques for Protein Structure Prediction (CASP14) 2020 experiment. Despite the recent proliferation of methods for predicting distances, few methods exist for evaluating these predictions. Currently only numerical metrics, which evaluate the entire prediction at once, are used. These give no insight into the structural details of a prediction. For this reason, new methods and tools are needed.

Results

We have developed a web server for evaluating predicted inter-residue distances. Our server, DISTEVAL, accepts predicted contacts, distances, and a true structure as optional inputs to generate informative heatmaps, chord diagrams, and 3D models. All of these outputs facilitate visual and qualitative assessment. The server also evaluates predictions using other metrics such as mean absolute error, root mean squared error, and contact precision.

Conclusions

The visualizations generated by DISTEVAL complement each other and collectively serve as a powerful tool for both quantitative and qualitative assessments of predicted contacts and distances, even in the absence of a true 3D structure.

Genetic interaction mapping informs integrative structure determination of protein complexes

Determining structures of protein complexes is crucial for understanding cellular functions. Here, we describe an integrative structure determination approach that relies on in vivo measurements of genetic interactions. We construct phenotypic profiles for point mutations crossed against gene deletions or exposed to environmental perturbations, followed by converting similarities between two profiles into an upper bound on the distance between the mutated residues. We determine the structure of the yeast histone H3-H4 complex based on ~500,000 genetic interactions of 350 mutants. We then apply the method to subunits Rpb1-Rpb2 of yeast RNA polymerase II and subunits RpoB-RpoC of bacterial RNA polymerase. The accuracy is comparable to that based on chemical cross-links; using restraints from both genetic interactions and cross-links further improves model accuracy and precision. The approach provides an efficient means to augment integrative structure determination with in vivo observations.

Predicting changes in protein thermodynamic stability upon point mutation with deep 3D convolutional neural networks

Predicting mutation-induced changes in protein thermodynamic stability (ΔΔG) is of great interest in protein engineering, variant interpretation, and protein biophysics. We introduce ThermoNet, a deep, 3D-convolutional neural network (3D-CNN) designed for structure-based prediction of ΔΔGs upon point mutation. To leverage the image-processing power inherent in CNNs, we treat protein structures as if they were multi-channel 3D images. In particular, the inputs to ThermoNet are uniformly constructed as multi-channel voxel grids based on biophysical properties derived from raw atom coordinates. We train and evaluate ThermoNet with a curated data set that accounts for protein homology and is balanced with direct and reverse mutations; this provides a framework for addressing biases that have likely influenced many previous ΔΔG prediction methods. ThermoNet demonstrates performance comparable to the best available methods on the widely used Ssym test set. In addition, ThermoNet accurately predicts the effects of both stabilizing and destabilizing mutations, while most other methods exhibit a strong bias towards predicting destabilization. We further show that homology between Ssym and widely used training sets like S2648 and VariBench has likely led to overestimated performance in previous studies. Finally, we demonstrate the practical utility of ThermoNet in predicting the ΔΔGs for two clinically relevant proteins, p53 and myoglobin, and for pathogenic and benign missense variants from ClinVar. Overall, our results suggest that 3D-CNNs can model the complex, non-linear interactions perturbed by mutations, directly from biophysical properties of atoms.

The role of local versus nonlocal physicochemical restraints in determining protein native structure

The tertiary structure of a native protein is dictated by the interplay of local secondary structure propensities, hydrogen bonding, and tertiary interactions. It is argued that the space of known protein topologies covers all single domain folds and results from the compactness of the native structure and excluded volume. Protein compactness combined with the chirality of the protein's side chains also yields native-like Ramachandran plots. It is the many-body, tertiary interactions among residues that collectively select for the global structure that a particular protein sequence adopts. This explains why the recent advances in deep-learning approaches that predict protein side-chain contacts, the distance matrix between residues, and sequence alignments are successful. They succeed because they implicitly learned the many-body interactions among protein residues.

CGLFold: a contact-assisted de novo protein structure prediction using global exploration and loop perturbation sampling algorithm

MOTIVATION

Regions that connect secondary structure elements in a protein are known as loops, whose slight change will produce dramatic effect on the entire topology. This study investigates whether the accuracy of protein structure prediction can be improved using a loop-specific sampling strategy.

RESULTS

A novel de novo protein structure prediction method that combines global exploration and loop perturbation is proposed in this study. In the global exploration phase, the fragment recombination and assembly are used to explore the massive conformational space and generate native-like topology. In the loop perturbation phase, a loop-specific local perturbation model is designed to improve the accuracy of the conformation and is solved by differential evolution algorithm. These two phases enable a cooperation between global exploration and local exploitation. The filtered contact information is used to construct the conformation selection model for guiding the sampling. The proposed CGLFold is tested on 145 benchmark proteins, 14 free modeling (FM) targets of CASP13 and 29 FM targets of CASP12. The experimental results show that the loop-specific local perturbation can increase the structure diversity and success rate of conformational update and gradually improve conformation accuracy. CGLFold obtains template modeling score ≥ 0.5 models on 95 standard test proteins, 7 FM targets of CASP13 and 9 FM targets of CASP12.

AVAILABILITY AND IMPLEMENTATION

The source code and executable versions are freely available at https://github.com/iobio-zjut/CGLFold.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

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.

CATHER: a novel threading algorithm with predicted contacts

MOTIVATION

Threading is one of the most effective methods for protein structure prediction. In recent years, the increasing accuracy in protein contact map prediction opens a new avenue to improve the performance of threading algorithms. Several preliminary studies suggest that with predicted contacts, the performance of threading algorithms can be improved greatly. There is still much room to explore to make better use of predicted contacts.

RESULTS

We have developed a new contact-assisted threading algorithm named CATHER using both conventional sequential profiles and contact map predicted by a deep learning-based algorithm. Benchmark tests on an independent test set and the CASP12 targets demonstrated that CATHER made significant improvement over other methods which only use either sequential profile or predicted contact map. Our method was ranked at the Top 10 among all 39 participated server groups on the 32 free modeling targets in the blind tests of the CASP13 experiment. These data suggest that it is promising to push forward the threading algorithms by using predicted contacts.

AVAILABILITY AND IMPLEMENTATION

http://yanglab.nankai.edu.cn/CATHER/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

DeepMSA: constructing deep multiple sequence alignment to improve contact prediction and fold-recognition for distant-homology proteins

MOTIVATION

The success of genome sequencing techniques has resulted in rapid explosion of protein sequences. Collections of multiple homologous sequences can provide critical information to the modeling of structure and function of unknown proteins. There are however no standard and efficient pipeline available for sensitive multiple sequence alignment (MSA) collection. This is particularly challenging when large whole-genome and metagenome databases are involved.

RESULTS

We developed DeepMSA, a new open-source method for sensitive MSA construction, which has homologous sequences and alignments created from multi-sources of whole-genome and metagenome databases through complementary hidden Markov model algorithms. The practical usefulness of the pipeline was examined in three large-scale benchmark experiments based on 614 non-redundant proteins. First, DeepMSA was utilized to generate MSAs for residue-level contact prediction by six coevolution and deep learning-based programs, which resulted in an accuracy increase in long-range contacts by up to 24.4% compared to the default programs. Next, multiple threading programs are performed for homologous structure identification, where the average TM-score of the template alignments has over 7.5% increases with the use of the new DeepMSA profiles. Finally, DeepMSA was used for secondary structure prediction and resulted in statistically significant improvements in the Q3 accuracy. It is noted that all these improvements were achieved without re-training the parameters and neural-network models, demonstrating the robustness and general usefulness of the DeepMSA in protein structural bioinformatics applications, especially for targets without homologous templates in the PDB library.

AVAILABILITY AND IMPLEMENTATION

https://zhanglab.ccmb.med.umich.edu/DeepMSA/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

A fully open-source framework for deep learning protein real-valued distances

As deep learning algorithms drive the progress in protein structure prediction, a lot remains to be studied at this merging superhighway of deep learning and protein structure prediction. Recent findings show that inter-residue distance prediction, a more granular version of the well-known contact prediction problem, is a key to predicting accurate models. However, deep learning methods that predict these distances are still in the early stages of their development. To advance these methods and develop other novel methods, a need exists for a small and representative dataset packaged for faster development and testing. In this work, we introduce protein distance net (PDNET), a framework that consists of one such representative dataset along with the scripts for training and testing deep learning methods. The framework also includes all the scripts that were used to curate the dataset, and generate the input features and distance maps. Deep learning models can also be trained and tested in a web browser using free platforms such as Google Colab. We discuss how PDNET can be used to predict contacts, distance intervals, and real-valued distances.

State-of-the-art web services for de novo protein structure prediction

Residue coevolution estimations coupled to machine learning methods are revolutionizing the ability of protein structure prediction approaches to model proteins that lack clear homologous templates in the Protein Data Bank (PDB). This has been patent in the last round of the Critical Assessment of Structure Prediction (CASP), which presented several very good models for the hardest targets. Unfortunately, literature reporting on these advances often lacks digests tailored to lay end users; moreover, some of the top-ranking predictors do not provide webservers that can be used by nonexperts. How can then end users benefit from these advances and correctly interpret the predicted models? Here we review the web resources that biologists can use today to take advantage of these state-of-the-art methods in their research, including not only the best de novo modeling servers but also datasets of models precomputed by experts for structurally uncharacterized protein families. We highlight their features, advantages and pitfalls for predicting structures of proteins without clear templates. We present a broad number of applications that span from driving forward biochemical investigations that lack experimental structures to actually assisting experimental structure determination in X-ray diffraction, cryo-EM and other forms of integrative modeling. We also discuss issues that must be considered by users yet still require further developments, such as global and residue-wise model quality estimates and sources of residue coevolution other than monomeric tertiary structure.

Refined template selection and combination algorithm significantly improves template-based modeling accuracy

In contrast to ab-initio protein modeling methodologies, comparative modeling is considered as the most popular and reliable algorithm to model protein structure. However, the selection of the best set of templates is still a major challenge. An effective template-ranking algorithm is developed to efficiently select only the reliable hits for predicting the protein structures. The algorithm employs the pairwise as well as multiple sequence alignments of template hits to rank and select the best possible set of templates. It captures several key sequences and structural information of template hits and converts into scores to effectively rank them. This selected set of templates is used to model a target. Modeling accuracy of the algorithm is tested and evaluated on TBM-HA domain containing CASP8, CASP9 and CASP10 targets. On an average, this template ranking and selection algorithm improves GDT-TS, GDT-HA and TM_Score by 3.531, 4.814 and 0.022, respectively. Further, it has been shown that the inclusion of structurally similar templates with ample conformational diversity is crucial for the modeling algorithm to maximally as well as reliably span the target sequence and construct its near-native model. The optimal model sampling also holds the key to predict the best possible target structure.

ResPRE: high-accuracy protein contact prediction by coupling precision matrix with deep residual neural networks

MOTIVATION

Contact-map of a protein sequence dictates the global topology of structural fold. Accurate prediction of the contact-map is thus essential to protein 3D structure prediction, which is particularly useful for the protein sequences that do not have close homology templates in the Protein Data Bank.

RESULTS

We developed a new method, ResPRE, to predict residue-level protein contacts using inverse covariance matrix (or precision matrix) of multiple sequence alignments (MSAs) through deep residual convolutional neural network training. The approach was tested on a set of 158 non-homologous proteins collected from the CASP experiments and achieved an average accuracy of 50.6% in the top-L long-range contact prediction with L being the sequence length, which is 11.7% higher than the best of other state-of-the-art approaches ranging from coevolution coupling analysis to deep neural network training. Detailed data analyses show that the major advantage of ResPRE lies at the utilization of precision matrix that helps rule out transitional noises of contact-maps compared with the previously used covariance matrix. Meanwhile, the residual network with parallel shortcut layer connections increases the learning ability of deep neural network training. It was also found that appropriate collection of MSAs can further improve the accuracy of final contact-map predictions. The standalone package and online server of ResPRE are made freely available, which should bring important impact on protein structure and function modeling studies in particular for the distant- and non-homology protein targets.

AVAILABILITY AND IMPLEMENTATION

https://zhanglab.ccmb.med.umich.edu/ResPRE and https://github.com/leeyang/ResPRE.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Improved inference of intermolecular contacts through protein-protein interaction prediction using coevolutionary analysis

MOTIVATION

Predicting residue-residue contacts between interacting proteins is an important problem in bioinformatics. The growing wealth of sequence data can be used to infer these contacts through correlated mutation analysis on multiple sequence alignments of interacting homologs of the proteins of interest. This requires correct identification of pairs of interacting proteins for many species, in order to avoid introducing noise (i.e. non-interacting sequences) in the analysis that will decrease predictive performance.

RESULTS

We have designed Ouroboros, a novel algorithm to reduce such noise in intermolecular contact prediction. Our method iterates between weighting proteins according to how likely they are to interact based on the correlated mutations signal, and predicting correlated mutations based on the weighted sequence alignment. We show that this approach accurately discriminates between protein interaction versus non-interaction and simultaneously improves the prediction of intermolecular contact residues compared to a naive application of correlated mutation analysis. This requires no training labels concerning interactions or contacts. Furthermore, the method relaxes the assumption of one-to-one interaction of previous approaches, allowing for the study of many-to-many interactions.

AVAILABILITY AND IMPLEMENTATION

Source code and test data are available at www.bif.wur.nl/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

The PSIPRED Protein Analysis Workbench: 20 years on

The PSIPRED Workbench is a web server offering a range of predictive methods to the bioscience community for 20 years. Here, we present the work we have completed to update the PSIPRED Protein Analysis Workbench and make it ready for the next 20 years. The main focus of our recent website upgrade work has been the acceleration of analyses in the face of increasing protein sequence database size. We additionally discuss any new software, the new hardware infrastructure, our webservices and web site. Lastly we survey updates to some of the key predictive algorithms available through our website.

Reducing Ensembles of Protein Tertiary Structures Generated De Novo via Clustering

Controlling the quality of tertiary structures computed for a protein molecule remains a central challenge in de-novo protein structure prediction. The rule of thumb is to generate as many structures as can be afforded, effectively acknowledging that having more structures increases the likelihood that some will reside near the sought biologically-active structure. A major drawback with this approach is that computing a large number of structures imposes time and space costs. In this paper, we propose a novel clustering-based approach which we demonstrate to significantly reduce an ensemble of generated structures without sacrificing quality. Evaluations are related on both benchmark and CASP target proteins. Structure ensembles subjected to the proposed approach and the source code of the proposed approach are publicly-available at the links provided in Section 1.

Emergence of New Disease: How Can Artificial Intelligence Help?

Emergence of new disease remains a critical parameter in human health and society. Advances in artificial intelligence (AI) allow for rapid processing and analysis of massive and complex data. In this forum article, the recent applications across disease prediction and drug development in relation to the COVID-19 pandemic are reviewed.

Will Cryo-Electron Microscopy Shift the Current Paradigm in Protein Structure Prediction?

Protein dynamics is undoubtedly a pervasive ingredient in all biological functions. However, structural biology has been strongly driven by a static-centered view of protein architecture. We argue that the recent advances of cryo-electron microscopy (EM) have the potential to more broadly explore the conformational landscapes of protein complexes and therefore will enhance our ability to predict the diverse conformations of tertiary and quaternary protein structures that are functionally relevant in physiological conditions.

Neural networks for protein structure and function prediction and dynamic analysis

Hardware and software advancements along with the accumulation of large amounts of data in recent years have together spurred a remarkable growth in the application of neural networks to various scientific fields. Machine learning based on neural networks with multiple (hidden) layers is becoming an extremely powerful approach for analyzing data. With the accumulation of large amounts of protein data such as structural and functional assay data, the effects of such approaches within the field of protein informatics are increasing. Here, we introduce our recent studies based on applications of neural networks for protein structure and function prediction and dynamic analysis involving: (i) inter-residue contact prediction based on a multiple sequence alignment (MSA) of amino acid sequences, (ii) prediction of protein-compound interaction using assay data, and (iii) detection of protein allostery from trajectories of molecular dynamic (MD) simulation.

In Silico Modeling of Crimean Congo Hemorrhagic Fever Virus Glycoprotein-N and Screening of Anti Viral Hits by Virtual Screening

Crimean-Congo hemorrhagic fever (CCHF) is a widespread zoonotic viral disease, caused by a tick-born virus Crimean-Congo hemorrhagic fever virus (CCHFV). This disease is endemic in Middle East, Asia, Africa and South-Eastern Europe with the mortality rate of 5–30%. CCHFV genome is composed of three segments: large, medium and small segments. M segment encodes a polyprotein (glycoprotein) so called glycoprotein N (Gn) which is considered as a potential druggable target for the effective therapy of CCHF. The complete structure of Gn is still not characterized. The aim of the current study is to predict the complete three-dimensional (3D-) structure of CCHFV Gn protein via threading-based modeling and investigate the residues crucial for binding with CCHFV envelop. The developed model displayed excellent stereo-chemical and geometrical properties. Subsequently structure based virtual screening (SBVS) was applied to discover novel inhibitors of Gn protein. A library of > 1300 anti-virals was selected from PubChem database and directed to the predicted binding site of Gn. The SBVS results led to the identification of thirty-seven compounds that inhibit the protein in computational analysis. Those 37 hits were subject to pharmacokinetic profiling which demonstrated that 30/37 compound possess safer pharmacokinetic properties. Thus, by specifically targeting Gn, less toxic and more potent inhibitors of CCHFV were identified in silico.

Epistatic contributions promote the unification of incompatible models of neutral molecular evolution

Mathematical models of evolution help us understand mechanisms driving protein-sequence change. Previous models recapitulate a disjoint subset of statistical features of natural sequences. We present a neutral evolution model that unifies features including extreme variance of the molecular clock’s tick rate and the observation of an evolutionary Stokes shift, an irreversible effect of mutations in the fitness landscape during sequence evolution. We show that interactions between amino acid sites, which inform our fitness metric, are required to observe these features. These interactions are inferred by using direct coupling analysis, which has been successfully utilized to predict protein structures, dynamics, and complexes from coevolutionary information. We anticipate our model will have applications in phylogenetics, ancestral reconstruction of sequences, and protein design.

We introduce a model of amino acid sequence evolution that accounts for the statistical behavior of real sequences induced by epistatic interactions. We base the model dynamics on parameters derived from multiple sequence alignments analyzed by using direct coupling analysis methodology. Known statistical properties such as overdispersion, heterotachy, and gamma-distributed rate-across-sites are shown to be emergent properties of this model while being consistent with neutral evolution theory, thereby unifying observations from previously disjointed evolutionary models of sequences. The relationship between site restriction and heterotachy is characterized by tracking the effective alphabet dynamics of sites. We also observe an evolutionary Stokes shift in the fitness of sequences that have undergone evolution under our simulation. By analyzing the structural information of some proteins, we corroborate that the strongest Stokes shifts derive from sites that physically interact in networks near biochemically important regions. Perspectives on the implementation of our model in the context of the molecular clock are discussed.

Evaluating the significance of contact maps in low-homology protein modeling using contact-assisted threading

The development of improved threading algorithms for remote homology modeling is a critical step forward in template-based protein structure prediction. We have recently demonstrated the utility of contact information to boost protein threading by developing a new contact-assisted threading method. However, the nature and extent to which the quality of a predicted contact map impacts the performance of contact-assisted threading remains elusive. Here, we systematically analyze and explore this interdependence by employing our newly-developed contact-assisted threading method over a large-scale benchmark dataset using predicted contact maps from four complementary methods including direct coupling analysis (mfDCA), sparse inverse covariance estimation (PSICOV), classical neural network-based meta approach (MetaPSICOV), and state-of-the-art ultra-deep learning model (RaptorX). Experimental results demonstrate that contact-assisted threading using high-quality contacts having the Matthews Correlation Coefficient (MCC) ≥ 0.5 improves threading performance in nearly 30% cases, while low-quality contacts with MCC <0.35 degrades the performance for 50% cases. This holds true even in CASP13 dataset, where threading using high-quality contacts (MCC ≥ 0.5) significantly improves the performance of 22 instances out of 29. Collectively, our study uncovers the mutual association between the quality of predicted contacts and its possible utility in boosting threading performance for improving low-homology protein modeling.

Coevolutionary Analysis of Protein Subfamilies by Sequence Reweighting

Extracting structural information from sequence co-variation has become a common computational biology practice in the recent years, mainly due to the availability of large sequence alignments of protein families. However, identifying features that are specific to sub-classes and not shared by all members of the family using sequence-based approaches has remained an elusive problem. We here present a coevolutionary-based method to differentially analyze subfamily specific structural features by a continuous sequence reweighting (SR) approach. We introduce the underlying principles and test its predictive capabilities on the Response Regulator family, whose subfamilies have been previously shown to display distinct, specific homo-dimerization patterns. Our results show that this reweighting scheme is effective in assigning structural features known a priori to subfamilies, even when sequence data is relatively scarce. Furthermore, sequence reweighting allows assessing if individual structural contacts pertain to specific subfamilies and it thus paves the way for the identification specificity-determining contacts from sequence variation data.

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.

Improved protein structure prediction using potentials from deep learning

Protein structure prediction can be used to determine the three-dimensional shape of a protein from its amino acid sequence¹. This problem is of fundamental importance as the structure of a protein largely determines its function²; however, protein structures can be difficult to determine experimentally. Considerable progress has recently been made by leveraging genetic information. It is possible to infer which amino acid residues are in contact by analysing covariation in homologous sequences, which aids in the prediction of protein structures³. Here we show that we can train a neural network to make accurate predictions of the distances between pairs of residues, which convey more information about the structure than contact predictions. Using this information, we construct a potential of mean force⁴ that can accurately describe the shape of a protein. We find that the resulting potential can be optimized by a simple gradient descent algorithm to generate structures without complex sampling procedures. The resulting system, named AlphaFold, achieves high accuracy, even for sequences with fewer homologous sequences. In the recent Critical Assessment of Protein Structure Prediction⁵ (CASP13)-a blind assessment of the state of the field-AlphaFold created high-accuracy structures (with template modelling (TM) scores⁶ of 0.7 or higher) for 24 out of 43 free modelling domains, whereas the next best method, which used sampling and contact information, achieved such accuracy for only 14 out of 43 domains. AlphaFold represents a considerable advance in protein-structure prediction. We expect this increased accuracy to enable insights into the function and malfunction of proteins, especially in cases for which no structures for homologous proteins have been experimentally determined⁷.

DeepECA: an end-to-end learning framework for protein contact prediction from a multiple sequence alignment

Background

Recently developed methods of protein contact prediction, a crucially important step for protein structure prediction, depend heavily on deep neural networks (DNNs) and multiple sequence alignments (MSAs) of target proteins. Protein sequences are accumulating to an increasing degree such that abundant sequences to construct an MSA of a target protein are readily obtainable. Nevertheless, many cases present different ends of the number of sequences that can be included in an MSA used for contact prediction. The abundant sequences might degrade prediction results, but opportunities remain for a limited number of sequences to construct an MSA. To resolve these persistent issues, we strove to develop a novel framework using DNNs in an end-to-end manner for contact prediction.

Results

We developed neural network models to improve precision of both deep and shallow MSAs. Results show that higher prediction accuracy was achieved by assigning weights to sequences in a deep MSA. Moreover, for shallow MSAs, adding a few sequential features was useful to increase the prediction accuracy of long-range contacts in our model. Based on these models, we expanded our model to a multi-task model to achieve higher accuracy by incorporating predictions of secondary structures and solvent-accessible surface areas. Moreover, we demonstrated that ensemble averaging of our models can raise accuracy. Using past CASP target protein domains, we tested our models and demonstrated that our final model is superior to or equivalent to existing meta-predictors.

Conclusions

The end-to-end learning framework we built can use information derived from either deep or shallow MSAs for contact prediction. Recently, an increasing number of protein sequences have become accessible, including metagenomic sequences, which might degrade contact prediction results. Under such circumstances, our model can provide a means to reduce noise automatically. According to results of tertiary structure prediction based on contacts and secondary structures predicted by our model, more accurate three-dimensional models of a target protein are obtainable than those from existing ECA methods, starting from its MSA. DeepECA is available from https://github.com/tomiilab/DeepECA.

Deep learning for mining protein data

The recent emergence of deep learning to characterize complex patterns of protein big data reveals its potential to address the classic challenges in the field of protein data mining. Much research has revealed the promise of deep learning as a powerful tool to transform protein big data into valuable knowledge, leading to scientific discoveries and practical solutions. In this review, we summarize recent publications on deep learning predictive approaches in the field of mining protein data. The application architectures of these methods include multilayer perceptrons, stacked autoencoders, deep belief networks, two- or three-dimensional convolutional neural networks, recurrent neural networks, graph neural networks, and complex neural networks and are described from five perspectives: residue-level prediction, sequence-level prediction, three-dimensional structural analysis, interaction prediction, and mass spectrometry data mining. The advantages and deficiencies of these architectures are presented in relation to various tasks in protein data mining. Additionally, some practical issues and their future directions are discussed, such as robust deep learning for protein noisy data, architecture optimization for specific tasks, efficient deep learning for limited protein data, multimodal deep learning for heterogeneous protein data, and interpretable deep learning for protein understanding. This review provides comprehensive perspectives on general deep learning techniques for protein data analysis.

Revisiting the "satisfaction of spatial restraints" approach of MODELLER for protein homology modeling

The most frequently used approach for protein structure prediction is currently homology modeling. The 3D model building phase of this methodology is critical for obtaining an accurate and biologically useful prediction. The most widely employed tool to perform this task is MODELLER. This program implements the “modeling by satisfaction of spatial restraints” strategy and its core algorithm has not been altered significantly since the early 1990s. In this work, we have explored the idea of modifying MODELLER with two effective, yet computationally light strategies to improve its 3D modeling performance. Firstly, we have investigated how the level of accuracy in the estimation of structural variability between a target protein and its templates in the form of σ values profoundly influences 3D modeling. We show that the σ values produced by MODELLER are on average weakly correlated to the true level of structural divergence between target-template pairs and that increasing this correlation greatly improves the program’s predictions, especially in multiple-template modeling. Secondly, we have inquired into how the incorporation of statistical potential terms (such as the DOPE potential) in the MODELLER’s objective function impacts positively 3D modeling quality by providing a small but consistent improvement in metrics such as GDT-HA and lDDT and a large increase in stereochemical quality. Python modules to harness this second strategy are freely available at https://github.com/pymodproject/altmod. In summary, we show that there is a large room for improving MODELLER in terms of 3D modeling quality and we propose strategies that could be pursued in order to further increase its performance.

Proteins are fundamental biological molecules that carry out countless activities in living beings. Since the function of proteins is dictated by their three-dimensional atomic structures, acquiring structural details of proteins provides deep insights into their function. Currently, the most frequently used computational approach for protein structure prediction is template-based modeling. In this approach, a target protein is modeled using the experimentally-derived structural information of a template protein assumed to have a similar structure to the target. MODELLER is the most frequently used program for template-based 3D model building. Despite its success, its predictions are not always accurate enough to be useful in Biomedical Research. Here, we show that it is possible to greatly increase the performance of MODELLER by modifying two aspects of its algorithm. First, we demonstrate that providing the program with accurate estimations of local target-template structural divergence greatly increases the quality of its predictions. Additionally, we show that modifying MODELLER’s scoring function with statistical potential energetic terms also helps to improve modeling quality. This work will be useful in future research, since it reports practical strategies to improve the performance of this core tool in Structural Bioinformatics.

StructureDistiller: Structural relevance scoring identifies the most informative entries of a contact map

Protein folding and structure prediction are two sides of the same coin. Contact maps and the related techniques of constraint-based structure reconstruction can be considered as unifying aspects of both processes. We present the Structural Relevance (SR) score which quantifies the information content of individual contacts and residues in the context of the whole native structure. The physical process of protein folding is commonly characterized with spatial and temporal resolution: some residues are Early Folding while others are Highly Stable with respect to unfolding events. We employ the proposed SR score to demonstrate that folding initiation and structure stabilization are subprocesses realized by distinct sets of residues. The example of cytochrome c is used to demonstrate how StructureDistiller identifies the most important contacts needed for correct protein folding. This shows that entries of a contact map are not equally relevant for structural integrity. The proposed StructureDistiller algorithm identifies contacts with the highest information content; these entries convey unique constraints not captured by other contacts. Identification of the most informative contacts effectively doubles resilience toward contacts which are not observed in the native contact map. Furthermore, this knowledge increases reconstruction fidelity on sparse contact maps significantly by 0.4 Å.

Evaluating Protein Transfer Learning with TAPE

Machine learning applied to protein sequences is an increasingly popular area of research. Semi-supervised learning for proteins has emerged as an important paradigm due to the high cost of acquiring supervised protein labels, but the current literature is fragmented when it comes to datasets and standardized evaluation techniques. To facilitate progress in this field, we introduce the Tasks Assessing Protein Embeddings (TAPE), a set of five biologically relevant semi-supervised learning tasks spread across different domains of protein biology. We curate tasks into specific training, validation, and test splits to ensure that each task tests biologically relevant generalization that transfers to real-life scenarios. We benchmark a range of approaches to semi-supervised protein representation learning, which span recent work as well as canonical sequence learning techniques. We find that self-supervised pretraining is helpful for almost all models on all tasks, more than doubling performance in some cases. Despite this increase, in several cases features learned by self-supervised pretraining still lag behind features extracted by state-of-the-art non-neural techniques. This gap in performance suggests a huge opportunity for innovative architecture design and improved modeling paradigms that better capture the signal in biological sequences. TAPE will help the machine learning community focus effort on scientifically relevant problems. Toward this end, all data and code used to run these experiments are available at https://github.com/songlab-cal/tape.

Assessing the accuracy of contact predictions in CASP13

The accuracy of sequence-based tertiary contact predictions was assessed in a blind prediction experiment at the CASP13 meeting. After 4 years of significant improvements in prediction accuracy, another dramatic advance has taken place since CASP12 was held 2 years ago. The precision of predicting the top L/5 contacts in the free modeling category, where L is the corresponding length of the protein in residues, has exceeded 70%. As a comparison, the best-performing group at CASP12 with a 47% precision would have finished below the top 1/3 of the CASP13 groups. Extensively trained deep neural network approaches dominate the top performing algorithms, which appear to efficiently integrate information on coevolving residues and interacting fragments or possibly utilize memories of sequence similarities and sometimes can deliver accurate results even in the absence of virtually any target specific evolutionary information. If the current performance is evaluated by F-score on L contacts, it stands around 24% right now, which, despite the tremendous impact and advance in improving its utility for structure modeling, also suggests that there is much room left for further improvement.

Federating Structural Models and Data: Outcomes from A Workshop on Archiving Integrative Structures

Structures of biomolecular systems are increasingly computed by integrative modeling. In this approach, a structural model is constructed by combining information from multiple sources, including varied experimental methods and prior models. In 2019, a Workshop was held as a Biophysical Society Satellite Meeting to assess progress and discuss further requirements for archiving integrative structures. The primary goal of the Workshop was to build consensus for addressing the challenges involved in creating common data standards, building methods for federated data exchange, and developing mechanisms for validating integrative structures. The summary of the Workshop and the recommendations that emerged are presented here.

High-accuracy protein structures by combining machine-learning with physics-based refinement

Protein structure prediction has long been available as an alternative to experimental structure determination, especially via homology modeling based on templates from related sequences. Recently, models based on distance restraints from coevolutionary analysis via machine learning to have significantly expanded the ability to predict structures for sequences without templates. One such method, AlphaFold, also performs well on sequences where templates are available but without using such information directly. Here we show that combining machine-learning based models from AlphaFold with state-of-the-art physics-based refinement via molecular dynamics simulations further improves predictions to outperform any other prediction method tested during the latest round of CASP. The resulting models have highly accurate global and local structures, including high accuracy at functionally important interface residues, and they are highly suitable as initial models for crystal structure determination via molecular replacement.

PredMP: a web server for de novo prediction and visualization of membrane proteins

MOTIVATION

PredMP is the first web service, to our knowledge, that aims at de novo prediction of the membrane protein (MP) 3D structure followed by the embedding of the MP into the lipid bilayer for visualization. Our approach is based on a high-throughput Deep Transfer Learning (DTL) method that first predicts MP contacts by learning from non-MPs and then predicts the 3D model of the MP using the predicted contacts as distance restraints. This algorithm is derived from our previous Deep Learning (DL) method originally developed for soluble protein contact prediction, which has been officially ranked No. 1 in CASP12. The DTL framework in our approach overcomes the challenge that there are only a limited number of solved MP structures for training the deep learning model. There are three modules in the PredMP server: (i) The DTL framework followed by the contact-assisted folding protocol has already been implemented in RaptorX-Contact, which serves as the key module for 3D model generation; (ii) The 1D annotation module, implemented in RaptorX-Property, is used to predict the secondary structure and disordered regions; and (iii) the visualization module to display the predicted MPs embedded in the lipid bilayer guided by the predicted transmembrane topology.

RESULTS

Tested on 510 non-redundant MPs, our server predicts correct folds for ∼290 MPs, which significantly outperforms existing methods. Tested on a blind and live benchmark CAMEO from September 2016 to January 2018, PredMP can successfully model all 10 MPs belonging to the hard category.

AVAILABILITY AND IMPLEMENTATION

PredMP is freely accessed on the web at http://www.predmp.com.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

AlphaFold at CASP13

SUMMARY

Computational prediction of protein structure from sequence is broadly viewed as a foundational problem of biochemistry and one of the most difficult challenges in bioinformatics. Once every two years the Critical Assessment of protein Structure Prediction (CASP) experiments are held to assess the state of the art in the field in a blind fashion, by presenting predictor groups with protein sequences whose structures have been solved but have not yet been made publicly available. The first CASP was organized in 1994, and the latest, CASP13, took place last December, when for the first time the industrial laboratory DeepMind entered the competition. DeepMind's entry, AlphaFold, placed first in the Free Modeling (FM) category, which assesses methods on their ability to predict novel protein folds (the Zhang group placed first in the Template-Based Modeling (TBM) category, which assess methods on predicting proteins whose folds are related to ones already in the Protein Data Bank.) DeepMind's success generated significant public interest. Their approach builds on two ideas developed in the academic community during the preceding decade: (i) the use of co-evolutionary analysis to map residue co-variation in protein sequence to physical contact in protein structure, and (ii) the application of deep neural networks to robustly identify patterns in protein sequence and co-evolutionary couplings and convert them into contact maps. In this Letter, we contextualize the significance of DeepMind's entry within the broader history of CASP, relate AlphaFold's methodological advances to prior work, and speculate on the future of this important problem.

Accurate prediction of protein contact maps by coupling residual two-dimensional bidirectional long short-term memory with convolutional neural networks

Motivation

Accurate prediction of a protein contact map depends greatly on capturing as much contextual information as possible from surrounding residues for a target residue pair. Recently, ultra-deep residual convolutional networks were found to be state-of-the-art in the latest Critical Assessment of Structure Prediction techniques (CASP12) for protein contact map prediction by attempting to provide a protein-wide context at each residue pair. Recurrent neural networks have seen great success in recent protein residue classification problems due to their ability to propagate information through long protein sequences, especially Long Short-Term Memory (LSTM) cells. Here, we propose a novel protein contact map prediction method by stacking residual convolutional networks with two-dimensional residual bidirectional recurrent LSTM networks, and using both one-dimensional sequence-based and two-dimensional evolutionary coupling-based information.

Results

We show that the proposed method achieves a robust performance over validation and independent test sets with the Area Under the receiver operating characteristic Curve (AUC) > 0.95 in all tests. When compared to several state-of-the-art methods for independent testing of 228 proteins, the method yields an AUC value of 0.958, whereas the next-best method obtains an AUC of 0.909. More importantly, the improvement is over contacts at all sequence-position separations. Specifically, a 8.95%, 5.65% and 2.84% increase in precision were observed for the top L∕10 predictions over the next best for short, medium and long-range contacts, respectively. This confirms the usefulness of ResNets to congregate the short-range relations and 2D-BRLSTM to propagate the long-range dependencies throughout the entire protein contact map 'image'.

Availability and implementation

SPOT-Contact server url: http://sparks-lab.org/jack/server/SPOT-Contact/.

Supplementary information

Supplementary data are available at Bioinformatics online.

Ligand-Binding-Site Structure Refinement Using Molecular Dynamics with Restraints Derived from Predicted Binding Site Templates

Accurate modeling of ligand-binding-site structures plays a critical role in structure-based virtual screening. However, the structures of the ligand-binding site in most predicted protein models are generally of low quality and need refinements. In this work, we present a ligand-binding-site structure refinement protocol using molecular dynamics simulation with restraints derived from predicted binding site templates. Our benchmark validation shows great performance for 40 diverse sets of proteins from the Astex list. The ligand-binding sites on modeled protein structures are consistently refined using our method with an average Cα RMSD improvement of 0.90 Å. Comparison of ligand binding modes from ligand docking to initial unrefined and refined structures shows an average of 1.97 Å RMSD improvement in the refined structures. These results demonstrate a promising new method of structure refinement for protein ligand-binding-site structures.

Protein structure prediction assisted with sparse NMR data in CASP13

CASP13 has investigated the impact of sparse NMR data on the accuracy of protein structure prediction. NOESY and ¹⁵ N-¹ H residual dipolar coupling data, typical of that obtained for ¹⁵ N,¹³ C-enriched, perdeuterated proteins up to about 40 kDa, were simulated for 11 CASP13 targets ranging in size from 80 to 326 residues. For several targets, two prediction groups generated models that are more accurate than those produced using baseline methods. Real NMR data collected for a de novo designed protein were also provided to predictors, including one data set in which only backbone resonance assignments were available. Some NMR-assisted prediction groups also did very well with these data. CASP13 also assessed whether incorporation of sparse NMR data improves the accuracy of protein structure prediction relative to nonassisted regular methods. In most cases, incorporation of sparse, noisy NMR data results in models with higher accuracy. The best NMR-assisted models were also compared with the best regular predictions of any CASP13 group for the same target. For six of 13 targets, the most accurate model provided by any NMR-assisted prediction group was more accurate than the most accurate model provided by any regular prediction group; however, for the remaining seven targets, one or more regular prediction method provided a more accurate model than even the best NMR-assisted model. These results suggest a novel approach for protein structure determination, in which advanced prediction methods are first used to generate structural models, and sparse NMR data is then used to validate and/or refine these models.

Assessment of protein assembly prediction in CASP13

We present the assembly category assessment in the 13th edition of the CASP community-wide experiment. For the second time, protein assemblies constitute an independent assessment category. Compared to the last edition we see a clear uptake in participation, more oligomeric targets released, and consistent, albeit modest, improvement of the predictions quality. Looking at the tertiary structure predictions, we observe that ignoring the oligomeric state of the targets hinders modeling success. We also note that some contact prediction groups successfully predicted homomeric interfacial contacts, though it appears that these predictions were not used for assembly modeling. Homology modeling with sizeable human intervention appears to form the basis of the assembly prediction techniques in this round of CASP. Future developments should see more integrated approaches where subunits are modeled in the context of the assemblies they form.

Enhancing Biomolecular Sampling with Reinforcement Learning: A Tree Search Molecular Dynamics Simulation Method

This paper proposes a novel molecular simulation method, called tree search molecular dynamics (TS-MD), to accelerate the sampling of conformational transition pathways, which require considerable computation. In TS-MD, a tree search algorithm, called upper confidence bounds for trees, which is a type of reinforcement learning algorithm, is applied to sample the transition pathway. By learning from the results of the previous simulations, TS-MD efficiently searches conformational space and avoids being trapped in local stable structures. TS-MD exhibits better performance than parallel cascade selection molecular dynamics, which is one of the state-of-the-art methods, for the folding of miniproteins, Chignolin and Trp-cage, in explicit water.

Deeper Profiles and Cascaded Recurrent and Convolutional Neural Networks for state-of-the-art Protein Secondary Structure Prediction

Protein Secondary Structure prediction has been a central topic of research in Bioinformatics for decades. In spite of this, even the most sophisticated ab initio SS predictors are not able to reach the theoretical limit of three-state prediction accuracy (88–90%), while only a few predict more than the 3 traditional Helix, Strand and Coil classes. In this study we present tests on different models trained both on single sequence and evolutionary profile-based inputs and develop a new state-of-the-art system with Porter 5. Porter 5 is composed of ensembles of cascaded Bidirectional Recurrent Neural Networks and Convolutional Neural Networks, incorporates new input encoding techniques and is trained on a large set of protein structures. Porter 5 achieves 84% accuracy (81% SOV) when tested on 3 classes and 73% accuracy (70% SOV) on 8 classes on a large independent set. In our tests Porter 5 is 2% more accurate than its previous version and outperforms or matches the most recent predictors of secondary structure we tested. When Porter 5 is retrained on SCOPe based sets that eliminate homology between training/testing samples we obtain similar results. Porter is available as a web server and standalone program at http://distilldeep.ucd.ie/porter/ alongside all the datasets and alignments.