Prediction of protein secondary structure from the amino acid sequence is a classical bioinformatics problem. Common methods use feed forward neural networks or SVMs combined with a sliding window, as these models does not naturally handle sequential data. Recurrent neural networks are an generalization of the feed forward neural network that naturally handle sequential data. We use a bidirectional recurrent neural network with long short term memory cells for prediction of secondary structure and evaluate using the CB513 dataset. On the secondary structure 8-class problem we report better performance (0.674) than state of the art (0.664). Our model includes feed forward networks between the long short term memory cells, a path that can be further explored.
Motivation: Protein secondary structure prediction can provide important information for protein 3D structure prediction and protein functions. Deep learning, which has been successfully applied to various research fields such as image classification and voice recognition, provides a new opportunity to significantly improve the secondary structure prediction accuracy. Although several deep-learning methods have been developed for secondary structure prediction, there is room for improvement. MUFold-SS was developed to address these issues. Results: Here, a very deep neural network, the deep in…
In this paper we experiment with using neural network structures to predict a protein's secondary structure (α helix positions) from only its primary structure (amino acid sequence). We implement a fully connected neural network (FCNN) and preform three experiments using that FCNN. Firstly, we do a cross-species comparison of models trained and tested on mouse and human datasets. Secondly, we test the impact of varying the length of protein sequence we input into the model. Thirdly, we compare custom error functions designed to focus on the center of the input window. At the end of paper we pr…
The correct prediction of protein secondary structures is one of the key issues in predicting the correct protein folded shape, which is used for determining gene function. Existing methods make use of amino acids properties as indices to classify protein secondary structures, but are faced with a significant number of misclassifications. The paper presents a technique for the classification of protein secondary structures based on protein "signal-plotting" and the use of the Fourier technique for digital signal processing. New indices are proposed to classify protein secondary structures by a…
Predicting protein secondary structure is a fundamental problem in protein structure prediction. Here we present a new supervised generative stochastic network (GSN) based method to predict local secondary structure with deep hierarchical representations. GSN is a recently proposed deep learning technique (Bengio & Thibodeau-Laufer, 2013) to globally train deep generative model. We present the supervised extension of GSN, which learns a Markov chain to sample from a conditional distribution, and applied it to protein structure prediction. To scale the model to full-sized, high-dimensional data…
Protein secondary structure prediction is an important problem in bioinformatics. Inspired by the recent successes of deep neural networks, in this paper, we propose an end-to-end deep network that predicts protein secondary structures from integrated local and global contextual features. Our deep architecture leverages convolutional neural networks with different kernel sizes to extract multiscale local contextual features. In addition, considering long-range dependencies existing in amino acid sequences, we set up a bidirectional neural network consisting of gated recurrent unit to capture g…
The Automated Protein Structure Analysis (APSA) method, which describes the protein backbone as a smooth line in 3-dimensional space and characterizes it by curvature kappa and torsion tau as a function of arc length s, was applied on 77 proteins to determine all secondary structural units via specific kappa(s) and tau(s) patterns. A total of 533 alpha-helices and 644 beta-strands were recognized by APSA, whereas DSSP gives 536 and 651 units, respectively. Kinks and distortions were quantified and the boundaries (entry and exit) of secondary structures were classified. Similarity between prote…
Toyota Technological Institute at Chicago（芝加哥丰田工业大学）；University of Chicago（芝加哥大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
Protein secondary structure (SS) prediction is important for studying protein structure and function. When only the sequence (profile) information is used as input feature, currently the best predictors can obtain ~80% Q3 accuracy, which has not been improved in the past decade. Here we present DeepCNF (Deep Convolutional Neural Fields) for protein SS prediction. DeepCNF is a Deep Learning extension of Conditional Neural Fields (CNF), which is an integration of Conditional Random Fields (CRF) and shallow neural networks. DeepCNF can model not only complex sequence-structure relationship by a d…
We present analysis of a novel tool for protein secondary structure prediction using the recently-investigated Neural Machine Translation framework. The tool provides a fast and accurate folding prediction based on primary structure with subsecond prediction time even for batched inputs. We hypothesize that Neural Machine Translation can improve upon current predictive accuracy by better encoding complex relationships between nearby but non-adjacent amino acids. We overview our modifications to the framework in order to improve accuracy on protein sequences. We report 65.9% Q3 accuracy and ana…
A new method for the Automated Protein Structure Analysis (APSA) is derived, which simplifies the protein backbone to a smooth curve in 3-dimensional space. For the purpose of obtaining this smooth line each amino acid is represented by its C$_α$ atom, which serves as suitable anchor point for a cubic spline fit. The backbone line is characterized by arc length $s$, curvature $κ(s)$, and torsion $τ(s)$. The $κ(s)$ and $τ(s)$ diagrams of the protein backbone suppress, because of the level of coarse graining applied, details of the bond framework of the backbone, however reveal accurately all se…
Shenzhen Research Institute of Big Data（深圳大数据研究院）；Tencent AI Lab（腾讯人工智能实验室）；CryoEM Center, SUSTech（南方科技大学冷冻电镜中心）
Protein secondary structure prediction (PSSP) is essential for protein function analysis. However, for low homologous proteins, the PSSP suffers from insufficient input features. In this paper, we explicitly import external self-supervised knowledge for low homologous PSSP under the guidance of residue-wise profile fusion. In practice, we firstly demonstrate the superiority of profile over Position-Specific Scoring Matrix (PSSM) for low homologous PSSP. Based on this observation, we introduce the novel self-supervised BERT features as the pseudo profile, which implicitly involves the residue d…
Columbia University（哥伦比亚大学）；Ben Gurion University（本古里安大学）
We tackle the problem of protein secondary structure prediction using a common task framework. This lead to the introduction of multiple ideas for neural architectures based on state of the art building blocks, used in this task for the first time. We take a principled machine learning approach, which provides genuine, unbiased performance measures, correcting longstanding errors in the application domain. We focus on the Q8 resolution of secondary structure, an active area for continuously improving methods. We use an ensemble of strong predictors to achieve accuracy of 70.7% (on the CB513 te…
The protein backbone is described as a smooth curved and twisted line in three-dimensional (3D) space and characterized by its curvature $κ(s)$ and torsion $τ(s)$ both expressed as a function of arc length s. It is shown that the function $τ(s)$ is sufficient to analyze the contributions of all amino acids to the conformation of the protein backbone. The characteristic peak and trough patterns of the $τ(s)$ diagrams can be translated into a 16-letter code, which provides a rapid identification of helices, strands, and turns, specifies entry and exit points of secondary structural units, and de…
We propose an optimized parameter set for protein secondary structure prediction using three layer feed forward back propagation neural network. The methodology uses four parameters viz. encoding scheme, window size, number of neurons in the hidden layer and type of learning algorithm. The input layer of the network consists of neurons changing from 3 to 19, corresponding to different window sizes. The hidden layer chooses a natural number from 1 to 20 as the number of neurons. The output layer consists of three neurons, each corresponding to known secondary structural classes viz. alpha helix…
The Automated Protein Structure Analysis (APSA) method is used for the classification of supersecondary structures. Basis for the classification is the encoding of three-dimensional (3D) residue conformations into a 16-letter code (3D-1D projection). It is shown that the letter code of the protein makes it possible to reconstruct its overall shape without ambiguity (1D-3D translation). Accordingly, the letter code is used for the development of classification rules that distinguish supersecondary structures by the properties of their turns and the orientation of the flanking helix or strand st…
In sequence-based predictions, conventionally an input sequence is represented by a multiple sequence alignment (MSA) or a representation derived from MSA, such as a position-specific scoring matrix. Recently, inspired by the development in natural language processing, several applications of sequence embedding have been observed. Here, we review different approaches of protein sequence embeddings and their applications including protein contact prediction, secondary structure, prediction, and function prediction.
National Key Laboratory for Novel Software Technology（计算机软件新技术国家重点实验室）；Department of Computer Science and Technology（计算机科学与技术学院）；Nanjing University（南京大学）
The tertiary structure of protein, as well as the local secondary structure organization are fully determined by the angles of the peptidic bound. The backbone dihedral angles not only determine the global fold of the protein, but also the details of the local chain organization. Although a wealth of structural information is available in different databases and numerous structural biology softwares have been developed, rapid conformational characterization remains challenging. We present here RamaDA, a program able to give a synthetic description of the conformation of a protein. The RamaDA p…
Data Science Institute（数据科学研究所）；Department of Biomedical Engineering（生物医学工程系）；University of Virginia（弗吉尼亚大学）
The biological function of a protein stems from its 3-dimensional structure, which is thermodynamically determined by the energetics of interatomic forces between its amino acid building blocks (the order of amino acids, known as the sequence, defines a protein). Given the costs (time, money, human resources) of determining protein structures via experimental means such as X-ray crystallography, can we better describe and compare protein 3D structures in a robust and efficient manner, so as to gain meaningful biological insights? We begin by considering a relatively simple problem, limiting ou…
Current protein forcefields like the ones seen in CHARMM or Xplor-NIH have many terms that include bonded and non-bonded terms. Yet the forcefields do not take into account the use of hydrogen bonds which are important for secondary structure creation and stabilization of proteins. SCOPE is an open-source program that generates proteins from rotamer space. It then creates a forcefield that uses only non-bonded and hydrogen bond energy terms to create a profile for a given protein. The profiles can then be used in an artificial neural network to create a linear model that is funneled to the tru…
Background: It is necessary and essential to discovery protein function from the novel primary sequences. Wet lab experimental procedures are not only time-consuming, but also costly, so predicting protein structure and function reliably based only on amino acid sequence has significant value. TATA-binding protein (TBP) is a kind of DNA binding protein, which plays a key role in the transcription regulation. Our study proposed an automatic approach for identifying TATA-binding proteins efficiently, accurately, and conveniently. This method would guide for the special protein identification wit…
Background: Disulfide bonds are crucial to protein structural formation. Developing an effective method topredict disulfide bonding formation is important for protein structural modeling and functional study. Mostcurrent methods still have shortcomings, including low accuracy and strict requirements for the selection ofdiscriminative features. Results: In this study, we introduced a nested structure of Bidirectional Long-short Term Memory(BLSTM)neural network called Global-Local-BLSTM (GL-BLSTM) for disulfide bonding state prediction. Based on thepatterns of disulfide bond formation, a BLSTM n…
Secondary structure elements of many protein families exhibit differential conservation on their opposing faces. Amphipathic helices and beta-sheets by definition possess this property, and play crucial functional roles. This type of evolutionary trajectory of a protein family is usually critical to the functions of the protein family, as well as in creating functions within subfamilies. That is, differential conservation maintains properties of a protein structure related to its orientation, and that are important in packing, recognition, and catalysis. Here I define and formulate a new conce…
Statistical analysis of evolutionary-related protein sequences provides insights about their structure, function, and history. We show that Restricted Boltzmann Machines (RBM), designed to learn complex high-dimensional data and their statistical features, can efficiently model protein families from sequence information. We here apply RBM to twenty protein families, and present detailed results for two short protein domains, Kunitz and WW, one long chaperone protein, Hsp70, and synthetic lattice proteins for benchmarking. The features inferred by the RBM are biologically interpretable: they ar…
Ranaspumin-2 (Rsn-2) is a surfactant protein found in the foam nests of the túngara frog. Previous experimental work has led to a proposed model of adsorption which involves an unusual clam shell-like `unhinging' of the protein at an interface. Interestingly, there is no concomitant denaturation of the secondary structural elements of Rsn-2 with the large scale transformation of its tertiary structure. In this work we use both experiment and simulation to better understand the driving forces underpinning this unusual process. We develop a modified Gō-model approach where we have included expli…
Absent experimental evidence, a robust methodology to predict the likelihood of N-glycosylation in human proteins is essential for guiding experimental work. Based on the distribution of amino acids in the neighborhood of the NxS/T sequon (N-site); the structural attributes of the N-site that include Accessible Surface Area, secondary structural elements, main-chain phi-psi, turn types; the relative location of the N-site in the primary sequence; and the nature of the glycan bound, the ridge regression estimated linear probability model is used to predict this likelihood. This model yields a K…
We show that in the generic situation where a biological network, e.g. a protein interaction network, is in fact a subnetwork embedded in a larger "bulk" network, the presence of the bulk causes not just extrinsic noise but also memory effects. This means that the dynamics of the subnetwork will depend not only on its present state, but also its past. We use projection techniques to get explicit expressions for the memory functions that encode such memory effects, for generic protein interaction networks involving binary and unary reactions such as complex formation and phosphorylation, respec…
In higher eukaryotes, alternative splicing is usually regulated by protein factors, which bind to the pre-mRNA and affect the recognition of splicing signals. There is recent evidence that the secondary structure of the pre-mRNA may also play an important role in this process, either by facilitating or by hindering the interaction with factors and small nuclear ribonucleoproteins (snRNPs) that regulate splicing. Moreover, the secondary structure could play a fundamental role in the splicing of yeast species, which lack many of the regulatory splicing factors present in metazoans. This review d…
