Deep template-based protein structure prediction

MotivationProtein structure prediction has been greatly improved by deep learning, but most efforts are devoted to template-free modeling. But very few deep learning methods are developed for TBM (template-based modeling), a popular technique for protein structure prediction. TBM has been studied extensively in the past, but its accuracy is not satisfactory when highly similar templates are not available.ResultsThis paper presents a new method NDThreader (New Deep-learning Threader) to address the challenges of TBM. NDThreader first employs DRNF (deep convolutional residual neural fields), which is an integration of deep ResNet (convolutional residue neural networks) and CRF (conditional random fields), to align a query protein to templates without using any distance information. Then NDThreader uses ADMM (alternating direction method of multipliers) and DRNF to further improve sequence-template alignments by making use of predicted distance potential. Finally, NDThreader builds 3D models from a sequence-template alignment by feeding it and sequence coevolution information into a deep ResNet to predict inter-atom distance distribution, which is then fed into PyRosetta for 3D model construction. Our experimental results show that NDThreader greatly outperforms existing methods such as CNFpred, HHpred, DeepThreader and CEthreader. NDThreader was blindly tested in CASP14 as a part of RaptorX server, which obtained the best average GDT score among all CASP14 servers on the 58 TBM targets.     

Machine learning in protein structure prediction

Prediction of protein structure from sequence has been intensely studied for many decades, owing to the problem's importance and its uniquely well-defined physical and computational bases. While progress has historically ebbed and flowed, the past two years saw dramatic advances driven by the increasing “neuralization” of structure prediction pipelines, whereby computations previously based on energy models and sampling procedures are replaced by neural networks. The extraction of physical contacts from the evolutionary record; the distillation of sequence–structure patterns from known structures; the incorporation of templates from homologs in the Protein Databank; and the refinement of coarsely predicted structures into finely resolved ones have all been reformulated using neural networks. Cumulatively, this transformation has resulted in algorithms that can now predict single protein domains with a median accuracy of 2.1 Å, setting the stage for a foundational reconfiguration of the role of biomolecular modeling within the life sciences.     

Deep learning based prediction of species-specific protein S-glutathionylation sites

As a widespread and reversible post-translational modification of proteins, S-glutathionylation specifically generates the mixed disulfides between cysteine residues and glutathione, which regulates various biological processes including oxidative stress, nitrosative stress and signal transduction. The identification of proteins and specific sites that undergo S-glutathionylation is crucial for understanding the underlying mechanisms and regulatory effects of S-glutathionylation. Experimental identification of S-glutathionylation sites is laborious and time-consuming, whereas computational predictions are more attractive due to their high speed and convenience. Here, we developed a novel computational framework DeepGSH (http://deepgsh.cancerbio.info/) for species-specific S-glutathionylation sites prediction, based on deep learning and particle swarm optimization algorithms. 5-fold cross validation indicated that DeepGSH was able to achieve an AUC of 0.8393 and 0.8458 for Homo sapiens and Mus musculus. According to critical evaluation and comparison, DeepGSH showed excellent robustness and better performance than existing tools in both species, demonstrating DeepGSH was suitable for S-glutathionylation prediction. The prediction results of DeepGSH might provide guidance for experimental validation of S-glutathionylation sites and helpful information to understand the intrinsic mechanisms.     

Machine learning approach for the prediction of protein secondary structure

PROMIS (protein machine induction system), a program for machine learning, was used to generalize rules that characterize the relationship between primary and secondary structure in globular proteins. These rules can be used to predict an unknown secondary structure from a known primary structure. The symbolic induction method used by PROMIS was specifically designed to produce rules that are meaningful in terms of chemical properties of the residues. The rules found were compared with existing knowledge of protein structure: some features of the rules were already recognized (e.g. amphipathic nature of alpha-helices). Other features are not understood, and are under investigation. The rules produced a prediction accuracy for three states (alpha-helix, beta-strand and coil) of 60% for all proteins, 73% for proteins of known alpha domain type, 62% for proteins of known beta domain type and 59% for proteins of known alpha/beta domain type. We conclude that machine learning is a useful tool in the examination of the large databases generated in molecular biology.     

Fast learning optimized prediction methodology (FLOPRED) for protein secondary structure prediction

Computational methods are rapidly gaining importance in the field of structural biology, mostly due to the explosive progress in genome sequencing projects and the large disparity between the number of sequences and the number of structures. There has been an exponential growth in the number of available protein sequences and a slower growth in the number of structures. There is therefore an urgent need to develop computational methods to predict structures and identify their functions from the sequence. Developing methods that will satisfy these needs both efficiently and accurately is of paramount importance for advances in many biomedical fields, including drug development and discovery of biomarkers. A novel method called fast learning optimized prediction methodology (FLOPRED) is proposed for predicting protein secondary structure, using knowledge-based potentials combined with structure information from the CATH database. A neural network-based extreme learning machine (ELM) and advanced particle swarm optimization (PSO) are used with this data that yield better and faster convergence to produce more accurate results. Protein secondary structures are predicted reliably, more efficiently and more accurately using FLOPRED. These techniques yield superior classification of secondary structure elements, with a training accuracy ranging between 83?% and 87?% over a widerange of hidden neurons and a cross-validated testing accuracy ranging between 81?% and 84?% and a segment overlap (SOV) score of 78?% that are obtained with different sets of proteins. These results are comparable to other recently published studies, but are obtained with greater efficiencies, in terms of time and cost.     

An evolutionary method for learning HMM structure: prediction of protein secondary structure

Background  

The prediction of the secondary structure of proteins is one of the most studied problems in bioinformatics. Despite their success in many problems of biological sequence analysis, Hidden Markov Models (HMMs) have not been used much for this problem, as the complexity of the task makes manual design of HMMs difficult. Therefore, we have developed a method for evolving the structure of HMMs automatically, using Genetic Algorithms (GAs).     

A global machine learning based scoring function for protein structure prediction

We present a knowledge‐based function to score protein decoys based on their similarity to native structure. A set of features is constructed to describe the structure and sequence of the entire protein chain. Furthermore, a qualitative relationship is established between the calculated features and the underlying electromagnetic interaction that dominates this scale. The features we use are associated with residue–residue distances, residue–solvent distances, pairwise knowledge‐based potentials and a four‐body potential. In addition, we introduce a new target to be predicted, the fitness score, which measures the similarity of a model to the native structure. This new approach enables us to obtain information both from decoys and from native structures. It is also devoid of previous problems associated with knowledge‐based potentials. These features were obtained for a large set of native and decoy structures and a back‐propagating neural network was trained to predict the fitness score. Overall this new scoring potential proved to be superior to the knowledge‐based scoring functions used as its inputs. In particular, in the latest CASP (CASP10) experiment our method was ranked third for all targets, and second for freely modeled hard targets among about 200 groups for top model prediction. Ours was the only method ranked in the top three for all targets and for hard targets. This shows that initial results from the novel approach are able to capture details that were missed by a broad spectrum of protein structure prediction approaches. Source codes and executable from this work are freely available at http://mathmed.org /#Software and http://mamiris.com/ . Proteins 2014; 82:752–759. © 2013 Wiley Periodicals, Inc.     

Servers for protein structure prediction

The 1990s cultivated a generation of protein structure human predictors. As a result of structural genomics and genome sequencing projects, and significant improvements in the performance of protein structure prediction methods, a generation of automated servers has evolved in the past few years. Servers for close and distant homology modeling are now routinely used by many biologists, and have already been applied to the experimental structure determination process itself, and to the interpretation and annotation of genome sequences. Because dozens of servers are currently available, it is hard for a biologist to know which server(s) to use; however, the state of the art of these methods is now assessed through the LiveBench and CAFASP experiments. Meta-servers--servers that use the results of other autonomous servers to produce a consensus prediction--have proven to be the best performers, and are already challenging all but a handful of expert human predictors. The difference in performance of the top ten autonomous (non-meta) servers is small and hard to assess using relatively small test sets. Recent experiments suggest that servers will soon free humans from most of the burden of protein structure prediction.     

Induced fit with replica exchange improves protein complex structure prediction

Despite the progress in prediction of protein complexes over the last decade, recent blind protein complex structure prediction challenges revealed limited success rates (less than 20% models with DockQ score > 0.4) on targets that exhibit significant conformational change upon binding. To overcome limitations in capturing backbone motions, we developed a new, aggressive sampling method that incorporates temperature replica exchange Monte Carlo (T-REMC) and conformational sampling techniques within docking protocols in Rosetta. Our method, ReplicaDock 2.0, mimics induced-fit mechanism of protein binding to sample backbone motions across putative interface residues on-the-fly, thereby recapitulating binding-partner induced conformational changes. Furthermore, ReplicaDock 2.0 clocks in at 150-500 CPU hours per target (protein-size dependent); a runtime that is significantly faster than Molecular Dynamics based approaches. For a benchmark set of 88 proteins with moderate to high flexibility (unbound-to-bound iRMSD over 1.2 Å), ReplicaDock 2.0 successfully docks 61% of moderately flexible complexes and 35% of highly flexible complexes. Additionally, we demonstrate that by biasing backbone sampling particularly towards residues comprising flexible loops or hinge domains, highly flexible targets can be predicted to under 2 Å accuracy. This indicates that additional gains are possible when mobile protein segments are known.     

Improving protein tertiary structure prediction by deep learning and distance prediction in CASP14

Substantial progresses in protein structure prediction have been made by utilizing deep-learning and residue-residue distance prediction since CASP13. Inspired by the advances, we improve our CASP14 MULTICOM protein structure prediction system by incorporating three new components: (a) a new deep learning-based protein inter-residue distance predictor to improve template-free (ab initio) tertiary structure prediction, (b) an enhanced template-based tertiary structure prediction method, and (c) distance-based model quality assessment methods empowered by deep learning. In the 2020 CASP14 experiment, MULTICOM predictor was ranked seventh out of 146 predictors in tertiary structure prediction and ranked third out of 136 predictors in inter-domain structure prediction. The results demonstrate that the template-free modeling based on deep learning and residue-residue distance prediction can predict the correct topology for almost all template-based modeling targets and a majority of hard targets (template-free targets or targets whose templates cannot be recognized), which is a significant improvement over the CASP13 MULTICOM predictor. Moreover, the template-free modeling performs better than the template-based modeling on not only hard targets but also the targets that have homologous templates. The performance of the template-free modeling largely depends on the accuracy of distance prediction closely related to the quality of multiple sequence alignments. The structural model quality assessment works well on targets for which enough good models can be predicted, but it may perform poorly when only a few good models are predicted for a hard target and the distribution of model quality scores is highly skewed. MULTICOM is available at https://github.com/jianlin-cheng/MULTICOM_Human_CASP14/tree/CASP14_DeepRank3 and https://github.com/multicom-toolbox/multicom/tree/multicom_v2.0 .     

Machine learning techniques for protein function prediction

Proteins play important roles in living organisms, and their function is directly linked with their structure. Due to the growing gap between the number of proteins being discovered and their functional characterization (in particular as a result of experimental limitations), reliable prediction of protein function through computational means has become crucial. This paper reviews the machine learning techniques used in the literature, following their evolution from simple algorithms such as logistic regression to more advanced methods like support vector machines and modern deep neural networks. Hyperparameter optimization methods adopted to boost prediction performance are presented. In parallel, the metamorphosis in the features used by these algorithms from classical physicochemical properties and amino acid composition, up to text-derived features from biomedical literature and learned feature representations using autoencoders, together with feature selection and dimensionality reduction techniques, are also reviewed. The success stories in the application of these techniques to both general and specific protein function prediction are discussed.     

Biophysical constraints for protein structure prediction

Though highly desirable, neither a single experimental technique nor a computational approach can be sufficient enough to rationalize a protein structure. The incorporation of biophysical constraints, which can be rationalized based on conventional biophysical measurements, might lead to considerable improvement of the simulation procedures. In this regard, our analysis of 180 proteins in different conformational states allows prediction of the overall protein dimension based on the chain length, i.e., the protein molecular weight, with an accuracy of 10%.     

Comparative modeling for protein structure prediction

With the progression of structural genomics projects, comparative modeling remains an increasingly important method of choice. It helps to bridge the gap between the available sequence and structure information by providing reliable and accurate protein models. Comparative modeling based on more than 30% sequence identity is now approaching its natural template-based limits and further improvements require the development of effective refinement techniques capable of driving models toward native structure. For difficult targets, for which the most significant progress in recent years has been observed, optimal template selection and alignment accuracy are still the major problems.     

Performance evaluation measures for protein complex prediction

Protein complexes play a dominant role in cellular organization and function. Prediction of protein complexes from the network of physical interactions between proteins (PPI networks) has thus become one of the important research areas. Recently, many computational approaches have been developed to identify these complexes. Various performance assessment measures have been proposed for evaluating the efficiency of these methods. However, there are many inconsistencies in the definitions and usage of the measures across the literature. To address this issue, we have gathered and presented the most important performance evaluation measures and developed a tool, named CompEvaluator, to critically assess the protein complex prediction methods. The tool and documentation are publicly available at https://sourceforge.net/projects/compevaluator/files/.     

面向蛋白质功能位点识别的机器学习平台构建

有关蛋白质功能的研究是解析生命奥秘的基础,机器学习技术在该领域已有广泛应用。利用支持向量机(support vectormachine,SVM)方法,构建一个预测蛋白质功能位点的通用平台。该平台先提取非同源蛋白质序列,再对这些序列进行特征编码(包括序列的基本信息、物化特征、结构信息及序列保守性特征等),以编码好的样本作为训练数据,利用SVM进行训练,得到敏感性、特异性、Matthew相关系数、准确率及ROC曲线等评价指标,反复测试,得到评价指标最优的SVM模型后,便可以用来预测蛋白质序列上的功能位点。该平台除了应用在预测蛋白质功能位点之外,还可以应用于疾病相关单核苷酸多态性(SNP)预测分析、预测蛋白质结构域分析、生物分子间的相互作用等。     

Analysis of distance-based protein structure prediction by deep learning in CASP13

This paper reports the CASP13 results of distance-based contact prediction, threading, and folding methods implemented in three RaptorX servers, which are built upon the powerful deep convolutional residual neural network (ResNet) method initiated by us for contact prediction in CASP12. On the 32 CASP13 FM (free-modeling) targets with a median multiple sequence alignment (MSA) depth of 36, RaptorX yielded the best contact prediction among 46 groups and almost the best 3D structure modeling among all server groups without time-consuming conformation sampling. In particular, RaptorX achieved top L/5, L/2, and L long-range contact precision of 70%, 58%, and 45%, respectively, and predicted correct folds (TMscore > 0.5) for 18 of 32 targets. Further, RaptorX predicted correct folds for all FM targets with >300 residues (T0950-D1, T0969-D1, and T1000-D2) and generated the best 3D models for T0950-D1 and T0969-D1 among all groups. This CASP13 test confirms our previous findings: (a) predicted distance is more useful than contacts for both template-based and free modeling; and (b) structure modeling may be improved by integrating template and coevolutionary information via deep learning. This paper will discuss progress we have made since CASP12, the strength and weakness of our methods, and why deep learning performed much better in CASP13.     

Deep learning architectures for prediction of nucleosome positioning from sequences data

Background

Nucleosomes are DNA-histone complex, each wrapping about 150 pairs of double-stranded DNA. Their function is fundamental for one of the primary functions of Chromatin i.e. packing the DNA into the nucleus of the Eukaryote cells. Several biological studies have shown that the nucleosome positioning influences the regulation of cell type-specific gene activities. Moreover, computational studies have shown evidence of sequence specificity concerning the DNA fragment wrapped into nucleosomes, clearly underlined by the organization of particular DNA substrings. As the main consequence, the identification of nucleosomes on a genomic scale has been successfully performed by computational methods using a sequence features representation.

Results

In this work, we propose a deep learning model for nucleosome identification. Our model stacks convolutional layers and Long Short-term Memories to automatically extract features from short- and long-range dependencies in a sequence. Using this model we are able to avoid the feature extraction and selection steps while improving the classification performances.

Conclusions

Results computed on eleven data sets of five different organisms, from Yeast to Human, show the superiority of the proposed method with respect to the state of the art recently presented in the literature.