Comparison study on statistical features of predicted secondary structures for protein structural class prediction: From content to position

Comparative Study on Feature Selection in Protein Structure and Function Prediction

Many effective methods extract and fuse different protein features to study the relationship between protein sequence, structure, and function, but different methods have preferences in solving the research of protein structure and function, which requires selecting valuable and contributing features to design more effective prediction methods. This work mainly focused on the feature selection methods in the study of protein structure and function, and systematically compared and analyzed the efficiency of different feature selection methods in the prediction of protein structures, protein disorders, protein molecular chaperones, and protein solubility. The results show that the feature selection method based on nonlinear SVM performs best in protein structure prediction, protein solubility prediction, protein molecular chaperone prediction, and protein solubility prediction. After selection, the accuracy of features is improved by 13.16% ~71%, especially the Kmer features and PSSM features of proteins.

Characterization and Structural Prediction of Proteins in SARS-CoV-2 Bangladeshi Variant Through Bioinformatics

The renowned respiratory disease induced by the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) has become a global epidemic in just less than a year by the first half of 2020. The subsequent efficient human-to-human transmission of this virus eventually affected millions of people worldwide. The most devastating thing is that the infection rate is continuously uprising and resulting in significant mortality especially among the older age population and those with health co-morbidities. This enveloped, positive-sense RNA virus is chiefly responsible for the infection of the upper respiratory system. The virulence of the SARS-CoV-2 is mostly regulated by its proteins such as entry to the host cell through fusion mechanism, fusion of infected cells with neighboring uninfected cells to spread virus, inhibition of host gene expression, cellular differentiation, apoptosis, mitochondrial biogenesis, etc. But very little is known about the protein structures and functionalities. Therefore, the main purpose of this study is to learn more about these proteins through bioinformatics approaches. In this study, ORF10, ORF7b, ORF7a, ORF6, membrane glycoprotein, and envelope protein have been selected from a Bangladeshi Corona-virus strain G039392 and a number of bioinformatics tools (MEGA-X-V10.1.7, PONDR, ProtScale, ProtParam, SCRIBER, NetSurfP v2.0, IntFOLD, UCSF Chimera, and PyMol) and strategies were implemented for multiple sequence alignment and phylogeny analysis with 9 different variants, predicting hydropathicity, amino acid compositions, protein-binding propensity, protein disorders, and 2D and 3D protein modeling. Selected proteins were characterized as highly flexible, structurally and electrostatically extremely stable, ordered, biologically active, hydrophobic, and closely related to proteins of different variants. This detailed information regarding the characterization and structure of proteins of SARS-CoV-2 Bangladeshi variant was performed for the first time ever to unveil the deep mechanism behind the virulence features. And this robust appraisal also paves the future way for molecular docking, vaccine development targeting these characterized proteins.

Prediction of High-Risk Types of Human Papillomaviruses Using Reduced Amino Acid Modes

A human papillomavirus type plays an important role in the early diagnosis of cervical cancer. Most of the prediction methods use protein sequence and structure information, but the reduced amino acid modes have not been used until now. In this paper, we introduced the modes of reduced amino acids to predict high-risk HPV. We first reduced 20 amino acids into several nonoverlapping groups and calculated their structure and physicochemical modes for high-risk HPV prediction, which was tested and compared with the existing methods on 68 samples of known HPV types. The experiment result indicates that the proposed method achieved better performance with an accuracy of 96.49%, indicating that the reduced amino acid modes might be used to improve the prediction of high-risk HPV types.

Using Recursive Feature Selection with Random Forest to Improve Protein Structural Class Prediction for Low-Similarity Sequences

Many combinations of protein features are used to improve protein structural class prediction, but the information redundancy is often ignored. In order to select the important features with strong classification ability, we proposed a recursive feature selection with random forest to improve protein structural class prediction. We evaluated the proposed method with four experiments and compared it with the available competing prediction methods. The results indicate that the proposed feature selection method effectively improves the efficiency of protein structural class prediction. Only less than 5% features are used, but the prediction accuracy is improved by 4.6-13.3%. We further compared different protein features and found that the predicted secondary structural features achieve the best performance. This understanding can be used to design more powerful prediction methods for the protein structural class.

A two-stage approach towards protein secondary structure classification

Protein secondary structure (PSS) describes the local folded structures which get formed inside a polypeptide due to interactions among atoms of the backbone. Generally, globular proteins are divided into four classes, namely all-α, all-β, α + β, and α/β. As nearly 90% of proteins fall into the said four classes, these are mostly considered for the purpose of computational classification of proteins. Classification of PSS is important for different biological functions that include protein fold recognition, tertiary structure prediction, prediction of DNA-binding sites, and reduction of the conformation search space among others. In this paper, we have proposed a machine learning-based model for secondary structure classification of proteins into four classes: all-α, all-β, α + β, and α/β. In doing so, we have considered both sequence-based and structure-based features. At first, mutual information (MI), a filter-based feature selection method, is used to remove the redundant features, and then these selected features are used to train three different classifiers-random forest, K-nearest neighbor (KNN), and multi-layer perceptron (MLP). After that, some standard classifier combination approaches are applied to integrate the decision made by the said classifiers and it has been found that weighted product rule performs the best among all. The overall accuracies obtained using the proposed model on the four standard datasets, namely 640, 1189, 25pdb, and fc699 are 86.89%, 92.93%, 91.38%, and 94.87% respectively. The proposed model outperforms some state-of-the-art methods considered here for comparison. Significantly high classification accuracy produced by our proposed model on four datasets is attributed to the development of a comprehensive feature set (by eliminating redundant features through feature selection technique) which is then passed through an ensemble consists of three different classifiers. Assigning different weights to the outcome of different classifiers thus proved to be useful in designing the model for predicting the secondary structure of proteins based on its sequence-based and structure-based features. Graphical abstract.

Prediction of protein structural classes by different feature expressions based on 2-D wavelet denoising and fusion

BACKGROUND

Protein structural class predicting is a heavily researched subject in bioinformatics that plays a vital role in protein functional analysis, protein folding recognition, rational drug design and other related fields. However, when traditional feature expression methods are adopted, the features usually contain considerable redundant information, which leads to a very low recognition rate of protein structural classes.

RESULTS

We constructed a prediction model based on wavelet denoising using different feature expression methods. A new fusion idea, first fuse and then denoise, is proposed in this article. Two types of pseudo amino acid compositions are utilized to distill feature vectors. Then, a two-dimensional (2-D) wavelet denoising algorithm is used to remove the redundant information from two extracted feature vectors. The two feature vectors based on parallel 2-D wavelet denoising are fused, which is known as PWD-FU-PseAAC. The related source codes are available at https://github.com/Xiaoheng-Wang12/Wang-xiaoheng/tree/master.

CONCLUSIONS

Experimental verification of three low-similarity datasets suggests that the proposed model achieves notably good results as regarding the prediction of protein structural classes.

Physicochemical Position-Dependent Properties in the Protein Secondary Structures

BACKGROUND

Establishing theories for designing arbitrary protein structures is complicated and depends on understanding the principles for protein folding, which is affected by applied features. Computer algorithms can reach high precision and stability in computationally designed enzymes and binders by applying informative features obtained from natural structures.

METHODS

In this study, a position-specific analysis of secondary structures (α-helix, β-strand, and tight turn) was performed to reveal novel features for protein structure prediction and protein design.

RESULTS

Our results showed that the secondary structures in the N-termini region tend to be more compact than C-termini. Decoying periodicity in length and distribution of amino acids in α-helices is deciphered using the curve-fitting methods. Compared with α-helix, β-strands do not show distinct periodicities in length. Also, significant differences in position-dependent distribution of physicochemical properties are shown in secondary structures.

CONCLUSION

Position-specific propensities in our study underline valuable parameters that could be used by researchers in the field of structural biology, particularly protein design through site-directed mutagenesis.

RaaMLab: A MATLAB toolbox that generates amino acid groups and reduced amino acid modes

Amino acid (AA) classification and its different biophysical and chemical characteristics have been widely applied to analyze and predict the structural, functional, expression and interaction profiles of proteins and peptides. We present RaaMLab, a free and open-source MATLAB toolbox, to facilitate studies on proteins and peptides, to generate AA groups and to extract the structural and physicochemical features of reduced AAs (RedAA). This toolbox offers 4 kinds of databases, including the physicochemical properties of AAs and their groupings, 49 AA classification methods and 5 types of biophysicochemical features of RedAAs. These factors can be easily computed based on user-defined alphabet size and AA properties of AA groupings. RaaMLab is an open source freely available at https://github.com/bioinfo0706/RaaMLab. This website also contains a tutorial, extensive documentation and examples.

One novel representation of DNA sequence based on the global and local position information

One novel representation of DNA sequence combining the global and local position information of the original sequence has been proposed to distinguish the different species. First, for the sufficient exploitation of global information, one graphical representation of DNA sequence has been formulated according to the curve of Fermat spiral. Then, for the consideration of local characteristics of DNA sequence, attaching each point in the curve of Fermat spiral with the related mass has been applied based on the relationships of neighboring four nucleotides. In this paper, the normalized moments of inertia of the curve of Fermat spiral which composed by the points with mass has been calculated as the numerical description of the corresponding DNA sequence on the first exons of beta-global genes. Choosing the Euclidean distance as the measurement of the numerical descriptions, the similarity between species has shown the performance of proposed method.

Detrended cross-correlation coefficient: Application to predict apoptosis protein subcellular localization

Apoptosis, or programed cell death, plays a central role in the development and homeostasis of an organism. Obtaining information on subcellular location of apoptosis proteins is very helpful for understanding the apoptosis mechanism. The prediction of subcellular localization of an apoptosis protein is still a challenging task, and existing methods mainly based on protein primary sequences. In this paper, we introduce a new position-specific scoring matrix (PSSM)-based method by using detrended cross-correlation (DCCA) coefficient of non-overlapping windows. Then a 190-dimensional (190D) feature vector is constructed on two widely used datasets: CL317 and ZD98, and support vector machine is adopted as classifier. To evaluate the proposed method, objective and rigorous jackknife cross-validation tests are performed on the two datasets. The results show that our approach offers a novel and reliable PSSM-based tool for prediction of apoptosis protein subcellular localization.

An estimator for local analysis of genome based on the minimal absent word

This study presents an alternative alignment-free relative feature analysis method based on the minimal absent word, which has potential advantages over the local alignment method in local analysis. Smooth-local-analysis-curve and similarity-distribution are constructed for a fast, efficient, and visual comparison. Moreover, when the multi-sequence-comparison is needed, the local-analysis-curves can illustrate some interesting zones.

Prediction of Protein Structural Classes for Low-Similarity Sequences Based on Consensus Sequence and Segmented PSSM

Prediction of protein structural classes for low-similarity sequences is useful for understanding fold patterns, regulation, functions, and interactions of proteins. It is well known that feature extraction is significant to prediction of protein structural class and it mainly uses protein primary sequence, predicted secondary structure sequence, and position-specific scoring matrix (PSSM). Currently, prediction solely based on the PSSM has played a key role in improving the prediction accuracy. In this paper, we propose a novel method called CSP-SegPseP-SegACP by fusing consensus sequence (CS), segmented PsePSSM, and segmented autocovariance transformation (ACT) based on PSSM. Three widely used low-similarity datasets (1189, 25PDB, and 640) are adopted in this paper. Then a 700-dimensional (700D) feature vector is constructed and the dimension is decreased to 224D by using principal component analysis (PCA). To verify the performance of our method, rigorous jackknife cross-validation tests are performed on 1189, 25PDB, and 640 datasets. Comparison of our results with the existing PSSM-based methods demonstrates that our method achieves the favorable and competitive performance. This will offer an important complementary to other PSSM-based methods for prediction of protein structural classes for low-similarity sequences.

An improved model for whole genome phylogenetic analysis by Fourier transform

DNA sequence similarity comparison is one of the major steps in computational phylogenetic studies. The sequence comparison of closely related DNA sequences and genomes is usually performed by multiple sequence alignments (MSA). While the MSA method is accurate for some types of sequences, it may produce incorrect results when DNA sequences undergone rearrangements as in many bacterial and viral genomes. It is also limited by its computational complexity for comparing large volumes of data. Previously, we proposed an alignment-free method that exploits the full information contents of DNA sequences by Discrete Fourier Transform (DFT), but still with some limitations. Here, we present a significantly improved method for the similarity comparison of DNA sequences by DFT. In this method, we map DNA sequences into 2-dimensional (2D) numerical sequences and then apply DFT to transform the 2D numerical sequences into frequency domain. In the 2D mapping, the nucleotide composition of a DNA sequence is a determinant factor and the 2D mapping reduces the nucleotide composition bias in distance measure, and thus improving the similarity measure of DNA sequences. To compare the DFT power spectra of DNA sequences with different lengths, we propose an improved even scaling algorithm to extend shorter DFT power spectra to the longest length of the underlying sequences. After the DFT power spectra are evenly scaled, the spectra are in the same dimensionality of the Fourier frequency space, then the Euclidean distances of full Fourier power spectra of the DNA sequences are used as the dissimilarity metrics. The improved DFT method, with increased computational performance by 2D numerical representation, can be applicable to any DNA sequences of different length ranges. We assess the accuracy of the improved DFT similarity measure in hierarchical clustering of different DNA sequences including simulated and real datasets. The method yields accurate and reliable phylogenetic trees and demonstrates that the improved DFT dissimilarity measure is an efficient and effective similarity measure of DNA sequences. Due to its high efficiency and accuracy, the proposed DFT similarity measure is successfully applied on phylogenetic analysis for individual genes and large whole bacterial genomes.

Prediction of protein structural features from sequence data based on Shannon entropy and Kolmogorov complexity

While the genome for a given organism stores the information necessary for the organism to function and flourish it is the proteins that are encoded by the genome that perhaps more than anything else characterize the phenotype for that organism. It is therefore not surprising that one of the many approaches to understanding and predicting protein folding and properties has come from genomics and more specifically from multiple sequence alignments. In this work I explore ways in which data derived from sequence alignment data can be used to investigate in a predictive way three different aspects of protein structure: secondary structures, inter-residue contacts and the dynamics of switching between different states of the protein. In particular the use of Kolmogorov complexity has identified a novel pathway towards achieving these goals.

An Improved Protein Structural Classes Prediction Method by Incorporating Both Sequence and Structure Information

Protein structural classes information is beneficial for secondary and tertiary structure prediction, protein folds prediction, and protein function analysis. Thus, predicting protein structural classes is of vital importance. In recent years, several computational methods have been developed for low-sequence-similarity (25%-40%) protein structural classes prediction. However, the reported prediction accuracies are actually not satisfactory. Aiming to further improve the prediction accuracies, we propose three different feature extraction methods and construct a comprehensive feature set that captures both sequence and structure information. By applying a random forest (RF) classifier to the feature set, we further develop a novel method for structural classes prediction. We test the proposed method on three benchmark datasets (25PDB, 640, and 1189) with low sequence similarity, and obtain the overall prediction accuracies of 93.5%, 92.6%, and 93.4%, respectively. Compared with six competing methods, the accuracies we achieved are 3.4%, 6.2%, and 8.7% higher than those achieved by the best-performing methods, showing the superiority of our method. Moreover, due to the limitation of the size of the three benchmark datasets, we further test the proposed method on three updated large-scale datasets with different sequence similarities (40%, 30%, and 25%). The proposed method achieves above 90% accuracies for all the three datasets, consistent with the accuracies on the above three benchmark datasets. Experimental results suggest our method as an effective and promising tool for structural classes prediction. Currently, a webserver that implements the proposed method is available on http://121.192.180.204:8080/RF_PSCP/Index.html.