基于支持向量机和贝叶斯方法的蛋白质四级结构分类研究

用支持向量机和贝叶斯两种方法对蛋白质四级结构进行分类研究。结果表明,基于支持向量机的分类结果最好,其l0CV检验的总分类精度、正样本正确预测率、Matthes相关系数和假阳性率分别为74．2％、84．6％、0．474、38．9％;基于贝叶斯的分类结果没有支持向量机的分类结果好,但其l0CV检验的假阳性率最低(15．9％)．这些结果说明同源寡聚蛋白质一级序列包含四级结构信息,同时特征向量的确表示了埋藏在缔合亚基作用部位接触表面的基本信息。     

Structure based function prediction of proteins using fragment library frequency vectors

The function of the protein is primarily dictated by its structure. Therefore it is far more logical to find the functional clues of the protein in its overall 3-dimensional fold or its global structure. In this paper, we have developed a novel Support Vector Machines (SVM) based prediction model for functional classification and prediction of proteins using features extracted from its global structure based on fragment libraries. Fragment libraries have been previously used for abintio modelling of proteins and protein structure comparisons. The query protein structure is broken down into a collection of short contiguous backbone fragments and this collection is discretized using a library of fragments. The input feature vector is frequency vector that counts the number of each library fragment in the collection of fragments by all-to-all fragment comparisons. SVM models were trained and optimised for obtaining the best 10-fold Cross validation accuracy for classification. As an example, this method was applied for prediction and classification of Cell Adhesion molecules (CAMs). Thirty-four different fragment libraries with sizes ranging from 4 to 400 and fragment lengths ranging from 4 to 12 were used for obtaining the best prediction model. The best 10-fold CV accuracy of 95.25% was obtained for library of 400 fragments of length 10. An accuracy of 87.5% was obtained on an unseen test dataset consisting of 20 CAMs and 20 NonCAMs. This shows that protein structure can be accurately and uniquely described using 400 representative fragments of length 10.     

基于氨基酸和二肽组成的蛋白质四级结构分类研究

基于最近邻居算法,从蛋白质一级序列出发,利用蛋白质序列氨基酸组成、二肤组成以及混合组成方法对蛋白质单聚体、二聚体、三聚体、四聚体、五聚体、六聚体和八聚体进行分类研究。结果表明:采用二肽组成编码方法的预洲效果最好,Jackknife检验和独立测试集检验的总体预测精度分别达到90.83%和95.48%,比相同数据集上基于伪氨基酸组成和组分耦合预测的方法提高了12和15个百分点;特别是对于五聚体蛋白,预测精度分别提高了90和50个百分点;说明二肽组成对于蛋白质四级结构分类研究是一种非常有效的特征提取方法。     

RSARF: prediction of residue solvent accessibility from protein sequence using random forest method

Prediction of protein structure from its amino acid sequence is still a challenging problem. The complete physicochemical understanding of protein folding is essential for the accurate structure prediction. Knowledge of residue solvent accessibility gives useful insights into protein structure prediction and function prediction. In this work, we propose a random forest method, RSARF, to predict residue accessible surface area from protein sequence information. The training and testing was performed using 120 proteins containing 22006 residues. For each residue, buried and exposed state was computed using five thresholds (0%, 5%, 10%, 25%, and 50%). The prediction accuracy for 0%, 5%, 10%, 25%, and 50% thresholds are 72.9%, 78.25%, 78.12%, 77.57% and 72.07% respectively. Further, comparison of RSARF with other methods using a benchmark dataset containing 20 proteins shows that our approach is useful for prediction of residue solvent accessibility from protein sequence without using structural information. The RSARF program, datasets and supplementary data are available at http://caps.ncbs.res.in/download/pugal/RSARF/.     

The recognition of multi-class protein folds by adding average chemical shifts of secondary structure elements

The recognition of protein folds is an important step in the prediction of protein structure and function. Recently, an increasing number of researchers have sought to improve the methods for protein fold recognition. Following the construction of a dataset consisting of 27 protein fold classes by Ding and Dubchak in 2001, prediction algorithms, parameters and the construction of new datasets have improved for the prediction of protein folds. In this study, we reorganized a dataset consisting of 76-fold classes constructed by Liu et al. and used the values of the increment of diversity, average chemical shifts of secondary structure elements and secondary structure motifs as feature parameters in the recognition of multi-class protein folds. With the combined feature vector as the input parameter for the Random Forests algorithm and ensemble classification strategy, we propose a novel method to identify the 76 protein fold classes. The overall accuracy of the test dataset using an independent test was 66.69%; when the training and test sets were combined, with 5-fold cross-validation, the overall accuracy was 73.43%. This method was further used to predict the test dataset and the corresponding structural classification of the first 27-protein fold class dataset, resulting in overall accuracies of 79.66% and 93.40%, respectively. Moreover, when the training set and test sets were combined, the accuracy using 5-fold cross-validation was 81.21%. Additionally, this approach resulted in improved prediction results using the 27-protein fold class dataset constructed by Ding and Dubchak.     

M-TASSER: an algorithm for protein quaternary structure prediction

In a cell, it has been estimated that each protein on average interacts with roughly 10 others, resulting in tens of thousands of proteins known or suspected to have interaction partners; of these, only a tiny fraction have solved protein structures. To partially address this problem, we have developed M-TASSER, a hierarchical method to predict protein quaternary structure from sequence that involves template identification by multimeric threading, followed by multimer model assembly and refinement. The final models are selected by structure clustering. M-TASSER has been tested on a benchmark set comprising 241 dimers having templates with weak sequence similarity and 246 without multimeric templates in the dimer library. Of the total of 207 targets predicted to interact as dimers, 165 (80%) were correctly assigned as interacting with a true positive rate of 68% and a false positive rate of 17%. The initial best template structures have an average root mean-square deviation to native of 5.3, 6.7, and 7.4 Å for the monomer, interface, and dimer structures. The final model shows on average a root mean-square deviation improvement of 1.3, 1.3, and 1.5 Å over the initial template structure for the monomer, interface, and dimer structures, with refinement evident for 87% of the cases. Thus, we have developed a promising approach to predict full-length quaternary structure for proteins that have weak sequence similarity to proteins of solved quaternary structure.     

Predicting homo-oligomers and hetero-oligomers by pseudo-amino acid composition: an approach from discrete wavelet transformation

Many proteins exist in vivo as oligomers with different quaternary structural attributes rather than as individual chains. These proteins are the structural components of various biological functions, including cooperative effects, allosteric mechanisms and ion-channel gating. With the dramatic increase in the number of protein sequences submitted to the public databank, it is important for both basic research and drug discovery research to acquire the knowledge about possible quaternary structural attributes of their interested proteins in a timely manner. A high-throughput method (DWT_SVM), fusing discrete wavelet transform (DWT) and support vector machine (SVM) classifier algorithm with various physicochemical features, has been developed to predict protein quaternary structure. The accuracy in distinguishing candidate proteins as homo-oligomer or hetero-oligomer using the dataset R₂₇₂₀ was 85.95% and 85.49% respectively by jackknife, showing that DWT_SVM is guide promising in predicting protein quaternary structures. The online service is available at http://bioinfo.ncu.edu.cn/Services.aspx. Protein sequences in FASTA format can be directly fed to the system OligoPred. The processed results will be presented in a diagram that includes the information of feature extraction and the classification error rate.     

Prediction of protein domain with mRMR feature selection and analysis

The domains are the structural and functional units of proteins. With the avalanche of protein sequences generated in the postgenomic age, it is highly desired to develop effective methods for predicting the protein domains according to the sequences information alone, so as to facilitate the structure prediction of proteins and speed up their functional annotation. However, although many efforts have been made in this regard, prediction of protein domains from the sequence information still remains a challenging and elusive problem. Here, a new method was developed by combing the techniques of RF (random forest), mRMR (maximum relevance minimum redundancy), and IFS (incremental feature selection), as well as by incorporating the features of physicochemical and biochemical properties, sequence conservation, residual disorder, secondary structure, and solvent accessibility. The overall success rate achieved by the new method on an independent dataset was around 73%, which was about 28-40% higher than those by the existing method on the same benchmark dataset. Furthermore, it was revealed by an in-depth analysis that the features of evolution, codon diversity, electrostatic charge, and disorder played more important roles than the others in predicting protein domains, quite consistent with experimental observations. It is anticipated that the new method may become a high-throughput tool in annotating protein domains, or may, at the very least, play a complementary role to the existing domain prediction methods, and that the findings about the key features with high impacts to the domain prediction might provide useful insights or clues for further experimental investigations in this area. Finally, it has not escaped our notice that the current approach can also be utilized to study protein signal peptides, B-cell epitopes, HIV protease cleavage sites, among many other important topics in protein science and biomedicine.     

Prediction of protein structure class by coupling improved genetic algorithm and support vector machine

Structural class characterizes the overall folding type of a protein or its domain. Most of the existing methods for determining the structural class of a protein are based on a group of features that only possesses a kind of discriminative information for the prediction of protein structure class. However, different types of discriminative information associated with primary sequence have been completely missed, which undoubtedly has reduced the success rate of prediction. We present a novel method for the prediction of protein structure class by coupling the improved genetic algorithm (GA) with the support vector machine (SVM). This improved GA was applied to the selection of an optimized feature subset and the optimization of SVM parameters. Jackknife tests on the working datasets indicated that the prediction accuracies for the different classes were in the range of 97.8–100% with an overall accuracy of 99.5%. The results indicate that the approach has a high potential to become a useful tool in bioinformatics.     

Predicting membrane protein types with bragging learner

The membrane protein type is an important feature in characterizing the overall topological folding type of a protein or its domains therein. Many investigators have put their efforts to the prediction of membrane protein type. Here, we propose a new approach, the bootstrap aggregating method or bragging learner, to address this problem based on the protein amino acid composition. As a demonstration, the benchmark dataset constructed by K.C. Chou and D.W. Elrod was used to test the new method. The overall success rate thus obtained by jackknife cross-validation was over 84%, indicating that the bragging learner as presented in this paper holds a quite high potential in predicting the attributes of proteins, or at least can play a complementary role to many existing algorithms in this area. It is anticipated that the prediction quality can be further enhanced if the pseudo amino acid composition can be effectively incorporated into the current predictor. An online membrane protein type prediction web server developed in our lab is available at http://chemdata.shu.edu.cn/protein/protein.jsp.     

Prediction of protein subcellular location using a combined feature of sequence

To understand the structure and function of a protein, an important task is to know where it occurs in the cell. Thus, a computational method for properly predicting the subcellular location of proteins would be significant in interpreting the original data produced by the large-scale genome sequencing projects. The present work tries to explore an effective method for extracting features from protein primary sequence and find a novel measurement of similarity among proteins for classifying a protein to its proper subcellular location. We considered four locations in eukaryotic cells and three locations in prokaryotic cells, which have been investigated by several groups in the past. A combined feature of primary sequence defined as a 430D (dimensional) vector was utilized to represent a protein, including 20 amino acid compositions, 400 dipeptide compositions and 10 physicochemical properties. To evaluate the prediction performance of this encoding scheme, a jackknife test based on nearest neighbor algorithm was employed. The prediction accuracies for cytoplasmic, extracellular, mitochondrial, and nuclear proteins in the former dataset were 86.3%, 89.2%, 73.5% and 89.4%, respectively, and the total prediction accuracy reached 86.3%. As for the prediction accuracies of cytoplasmic, extracellular, and periplasmic proteins in the latter dataset, the prediction accuracies were 97.4%, 86.0%, and 79.7, respectively, and the total prediction accuracy of 92.5% was achieved. The results indicate that this method outperforms some existing approaches based on amino acid composition or amino acid composition and dipeptide composition.     

PIER: protein interface recognition for structural proteomics

Recent advances in structural proteomics call for development of fast and reliable automatic methods for prediction of functional surfaces of proteins with known three-dimensional structure, including binding sites for known and unknown protein partners as well as oligomerization interfaces. Despite significant progress the problem is still far from being solved. Most existing methods rely, at least partially, on evolutionary information from multiple sequence alignments projected on protein surface. The common drawback of such methods is their limited applicability to the proteins with a sparse set of sequential homologs, as well as inability to detect interfaces in evolutionary variable regions. In this study, the authors developed an improved method for predicting interfaces from a single protein structure, which is based on local statistical properties of the protein surface derived at the level of atomic groups. The proposed Protein IntErface Recognition (PIER) method achieved the overall precision of 60% at the recall threshold of 50% at the residue level on a diverse benchmark of 490 homodimeric, 62 heterodimeric, and 196 transient interfaces (compared with 25% precision at 50% recall expected from random residue function assignment). For 70% of proteins in the benchmark, the binding patch residues were successfully detected with precision exceeding 50% at 50% recall. The calculation only took seconds for an average 300-residue protein. The authors demonstrated that adding the evolutionary conservation signal only marginally influenced the overall prediction performance on the benchmark; moreover, for certain classes of proteins, using this signal actually resulted in a deteriorated prediction. Thorough benchmarking using other datasets from literature showed that PIER yielded improved performance as compared with several alignment-free or alignment-dependent predictions. The accuracy, efficiency, and dependence on structure alone make PIER a suitable tool for automated high-throughput annotation of protein structures emerging from structural proteomics projects.     

Enhanced performance in prediction of protein active sites with THEMATICS and support vector machines

Theoretical microscopic titration curves (THEMATICS) is a computational method for the identification of active sites in proteins through deviations in computed titration behavior of ionizable residues. While the sensitivity to catalytic sites is high, the previously reported sensitivity to catalytic residues was not as high, about 50%. Here THEMATICS is combined with support vector machines (SVM) to improve sensitivity for catalytic residue prediction from protein 3D structure alone. For a test set of 64 proteins taken from the Catalytic Site Atlas (CSA), the average recall rate for annotated catalytic residues is 61%; good precision is maintained selecting only 4% of all residues. The average false positive rate, using the CSA annotations is only 3.2%, far lower than other 3D-structure-based methods. THEMATICS-SVM returns higher precision, lower false positive rate, and better overall performance, compared with other 3D-structure-based methods. Comparison is also made with the latest machine learning methods that are based on both sequence alignments and 3D structures. For annotated sets of well-characterized enzymes, THEMATICS-SVM performance compares very favorably with methods that utilize sequence homology. However, since THEMATICS depends only on the 3D structure of the query protein, no decline in performance is expected when applied to novel folds, proteins with few sequence homologues, or even orphan sequences. An extension of the method to predict non-ionizable catalytic residues is also presented. THEMATICS-SVM predicts a local network of ionizable residues with strong interactions between protonation events; this appears to be a special feature of enzyme active sites.     

Predicting disulfide connectivity from protein sequence using multiple sequence feature vectors and secondary structure

MOTIVATION: Disulfide bonds are primary covalent crosslinks between two cysteine residues in proteins that play critical roles in stabilizing the protein structures and are commonly found in extracy-toplasmatic or secreted proteins. In protein folding prediction, the localization of disulfide bonds can greatly reduce the search in conformational space. Therefore, there is a great need to develop computational methods capable of accurately predicting disulfide connectivity patterns in proteins that could have potentially important applications. RESULTS: We have developed a novel method to predict disulfide connectivity patterns from protein primary sequence, using a support vector regression (SVR) approach based on multiple sequence feature vectors and predicted secondary structure by the PSIPRED program. The results indicate that our method could achieve a prediction accuracy of 74.4% and 77.9%, respectively, when averaged on proteins with two to five disulfide bridges using 4-fold cross-validation, measured on the protein and cysteine pair on a well-defined non-homologous dataset. We assessed the effects of different sequence encoding schemes on the prediction performance of disulfide connectivity. It has been shown that the sequence encoding scheme based on multiple sequence feature vectors coupled with predicted secondary structure can significantly improve the prediction accuracy, thus enabling our method to outperform most of other currently available predictors. Our work provides a complementary approach to the current algorithms that should be useful in computationally assigning disulfide connectivity patterns and helps in the annotation of protein sequences generated by large-scale whole-genome projects. AVAILABILITY: The prediction web server and Supplementary Material are accessible at http://foo.maths.uq.edu.au/~huber/disulfide     

Prediction of complex super-secondary structure βαβ motifs based on combined features

Prediction of a complex super-secondary structure is a key step in the study of tertiary structures of proteins. The strand-loop-helix-loop-strand (βαβ) motif is an important complex super-secondary structure in proteins. Many functional sites and active sites often occur in polypeptides of βαβ motifs. Therefore, the accurate prediction of βαβ motifs is very important to recognizing protein tertiary structure and the study of protein function. In this study, the βαβ motif dataset was first constructed using the DSSP package. A statistical analysis was then performed on βαβ motifs and non-βαβ motifs. The target motif was selected, and the length of the loop-α-loop varies from 10 to 26 amino acids. The ideal fixed-length pattern comprised 32 amino acids. A Support Vector Machine algorithm was developed for predicting βαβ motifs by using the sequence information, the predicted structure and function information to express the sequence feature. The overall predictive accuracy of 5-fold cross-validation and independent test was 81.7% and 76.7%, respectively. The Matthew’s correlation coefficient of the 5-fold cross-validation and independent test are 0.63 and 0.53, respectively. Results demonstrate that the proposed method is an effective approach for predicting βαβ motifs and can be used for structure and function studies of proteins.     

The prediction of protein structural class using averaged chemical shifts

Knowledge of protein structural class can provide important information about its folding patterns. Many approaches have been developed for the prediction of protein structural classes. However, the information used by these approaches is primarily based on amino acid sequences. In this study, a novel method is presented to predict protein structural classes by use of chemical shift (CS) information derived from nuclear magnetic resonance spectra. Firstly, 399 non-homologue (about 15% identity) proteins were constructed to investigate the distribution of averaged CS values of six nuclei ((13)CO, (13)Cα, (13)Cβ, (1)HN, (1)Hα and (15)N) in three protein structural classes. Subsequently, support vector machine was proposed to predict three protein structural classes by using averaged CS information of six nuclei. Overall accuracy of jackknife cross-validation achieves 87.0%. Finally, the feature selection technique is applied to exclude redundant information and find out an optimized feature set. Results show that the overall accuracy increased to 88.0% by using the averaged CSs of (13)CO, (1)Hα and (15)N. The proposed approach outperformed other state-of-the-art methods in terms of predictive accuracy in particular for low-similarity protein data. We expect that our proposed approach will be an excellent alternative to traditional methods for protein structural class prediction.     

Prediction of flexible/rigid regions from protein sequences using k-spaced amino acid pairs

Background

Traditionally, it is believed that the native structure of a protein corresponds to a global minimum of its free energy. However, with the growing number of known tertiary (3D) protein structures, researchers have discovered that some proteins can alter their structures in response to a change in their surroundings or with the help of other proteins or ligands. Such structural shifts play a crucial role with respect to the protein function. To this end, we propose a machine learning method for the prediction of the flexible/rigid regions of proteins (referred to as FlexRP); the method is based on a novel sequence representation and feature selection. Knowledge of the flexible/rigid regions may provide insights into the protein folding process and the 3D structure prediction.

Results

The flexible/rigid regions were defined based on a dataset, which includes protein sequences that have multiple experimental structures, and which was previously used to study the structural conservation of proteins. Sequences drawn from this dataset were represented based on feature sets that were proposed in prior research, such as PSI-BLAST profiles, composition vector and binary sequence encoding, and a newly proposed representation based on frequencies of k-spaced amino acid pairs. These representations were processed by feature selection to reduce the dimensionality. Several machine learning methods for the prediction of flexible/rigid regions and two recently proposed methods for the prediction of conformational changes and unstructured regions were compared with the proposed method. The FlexRP method, which applies Logistic Regression and collocation-based representation with 95 features, obtained 79.5% accuracy. The two runner-up methods, which apply the same sequence representation and Support Vector Machines (SVM) and Naïve Bayes classifiers, obtained 79.2% and 78.4% accuracy, respectively. The remaining considered methods are characterized by accuracies below 70%. Finally, the Naïve Bayes method is shown to provide the highest sensitivity for the prediction of flexible regions, while FlexRP and SVM give the highest sensitivity for rigid regions.

Conclusion

A new sequence representation that uses k-spaced amino acid pairs is shown to be the most efficient in the prediction of the flexible/rigid regions of protein sequences. The proposed FlexRP method provides the highest prediction accuracy of about 80%. The experimental tests show that the FlexRP and SVM methods achieved high overall accuracy and the highest sensitivity for rigid regions, while the best quality of the predictions for flexible regions is achieved by the Naïve Bayes method.     

Prediction of Protein Secondary Structure Using Feature Selection and Analysis Approach

The prediction of the secondary structure of a protein from its amino acid sequence is an important step towards the prediction of its three-dimensional structure. However, the accuracy of ab initio secondary structure prediction from sequence is about 80 % currently, which is still far from satisfactory. In this study, we proposed a novel method that uses binomial distribution to optimize tetrapeptide structural words and increment of diversity with quadratic discriminant to perform prediction for protein three-state secondary structure. A benchmark dataset including 2,640 proteins with sequence identity of less than 25 % was used to train and test the proposed method. The results indicate that overall accuracy of 87.8 % was achieved in secondary structure prediction by using ten-fold cross-validation. Moreover, the accuracy of predicted secondary structures ranges from 84 to 89 % at the level of residue. These results suggest that the feature selection technique can detect the optimized tetrapeptide structural words which affect the accuracy of predicted secondary structures.