A Statistical Model for Predicting Protein Folding Rates from Amino Acid Sequence with Structural Class Information

Importance of Inter-residue Contacts for Understanding Protein Folding and Unfolding Rates, Remote Homology, and Drug Design

Inter-residue interactions in protein structures provide valuable insights into protein folding and stability. Understanding these interactions can be helpful in many crucial applications, including rational design of therapeutic small molecules and biologics, locating functional protein sites, and predicting protein-protein and protein-ligand interactions. The process of developing machine learning models incorporating inter-residue interactions has been improved recently. This review highlights the theoretical models incorporating inter-residue interactions in predicting folding and unfolding rates of proteins. Utilizing contact maps to depict inter-residue interactions aids researchers in developing computer models for detecting remote homologs and interface residues within protein-protein complexes which, in turn, enhances our knowledge of the relationship between sequence and structure of proteins. Further, the application of contact maps derived from inter-residue interactions is highlighted in the field of drug discovery. Overall, this review presents an extensive assessment of the significant models that use inter-residue interactions to investigate folding rates, unfolding rates, remote homology, and drug development, providing potential future advancements in constructing efficient computational models in structural biology.

HormoNet: a deep learning approach for hormone-drug interaction prediction

Several experimental evidences have shown that the human endogenous hormones can interact with drugs in many ways and affect drug efficacy. The hormone drug interactions (HDI) are essential for drug treatment and precision medicine; therefore, it is essential to understand the hormone-drug associations. Here, we present HormoNet to predict the HDI pairs and their risk level by integrating features derived from hormone and drug target proteins. To the best of our knowledge, this is one of the first attempts to employ deep learning approach for prediction of HDI prediction. Amino acid composition and pseudo amino acid composition were applied to represent target information using 30 physicochemical and conformational properties of the proteins. To handle the imbalance problem in the data, we applied synthetic minority over-sampling technique technique. Additionally, we constructed novel datasets for HDI prediction and the risk level of their interaction. HormoNet achieved high performance on our constructed hormone-drug benchmark datasets. The results provide insights into the understanding of the relationship between hormone and a drug, and indicate the potential benefit of reducing risk levels of interactions in designing more effective therapies for patients in drug treatments. Our benchmark datasets and the source codes for HormoNet are available in: https://github.com/EmamiNeda/HormoNet .

Protein structuromics: A new method for protein structure-function crosstalk in glioma

Glioma is a type of tumor that starts in the glial cells of the brain or spine. Since the 1800s, when the disease was first named, its survival rates have always been unsatisfactory. Despite great advances in molecular biology and traditional treatment methods, many questions regarding cancer occurrence and the underlying mechanism remain to be answered. In this study, we assessed the protein structural features of 20 oncogenes and 20 anti-oncogenes via protein structure and dynamic analysis methods and 3D structural and systematic analyses of the structure-function relationships of proteins. All of these results directly indicate that unfavorable group proteins show more complex structures than favorable group proteins. As the tumor cell microenvironment changes, the balance of oncogene-related and anti-oncogene-related proteins is disrupted, and most of the structures of the two groups of proteins will be disrupted. However, more unfavorable group proteins will maintain and refold to achieve their correct shape faster and perform their functions more quickly than favorable group proteins, and the former thus support cancer development. We hope that these analyses will help promote mechanistic research and the development of new treatments for glioma.

Appraisal of blob-Based Approaches in the Prediction of Protein Folding Times

A series of reports published in the last 3 years has illustrated that a blob-based model (BBM) can predict the folding time of proteins from their primary amino acid (aa) sequence based on three simple rules established to characterize the long-range backbone dynamics (LRBD) of racemic polypeptides. The sole use of LRBD to predict protein folding times with the BBM represents a radical departure from all other prediction methods currently applied to determine protein folding times, which rely instead on parameters such as the structure content, folding kinetics, chain length, amino acid properties, or contact topography of proteins. Furthermore, the built-in modularity of the BBM enables the parametrization and inclusion of new phenomena affecting the LRBD of polypeptides, while its conceptual simplicity makes it an interesting new mathematical tool for studying protein folding. However, its novelty implies that its relationship with many other methods used to predict protein folding times has not been well researched. Consequently, the purpose of this report is to uncover the physical phenomena encountered during protein folding that are best described by the BBM through the identification of parameters that have been recognized over the years as being strong predictors for protein folding, such as protein size, topology, structural class, and folding kinetics. This was accomplished by determining the parameters most strongly correlated with the folding times predicted by the BBM. While the BBM in its present form appears to be a good indicator of the folding times of the vast majority of the 195 proteins considered so far, this report finds that it excels for moderately large proteins that are primarily composed of locally formed structural motifs such as α-helices or for proteins that fold in multiple steps. Altogether, these observations based on the use of the BBM support the notion that proteins fold the way they do because the LRBD of polypeptides is mostly driven by the local interactions experienced between aa's within reach of one another.

TMH Stab-pred: Predicting the stability of α-helical membrane proteins using sequence and structural features

The folding and stability of transmembrane proteins (TMPs) are governed by the insertion of secondary structural elements into the cell membrane followed by their assembly. Understanding the important features that dictate the stability of TMPs is important for elucidating their functions. In this work, we related sequence and structure-based parameters with free energy (ΔG0) of α-helical membrane proteins. Our results showed that the free energy transfer of hydrophobic peptides, relative contact order, total interaction energy, number of hydrogen bonds and lipid accessibility of transmembrane regions are important for stability. Further, we have developed multiple-regression models to predict the stability of α-helical membrane proteins using these features and our method can predict the stability with a correlation and mean absolute error (MAE) of 0.89 and 1.21 kcal/mol, respectively, on jack-knife test. The method was validated with a blind test set of three recently reported experimental ΔG0, which could predict the stability within an average MAE of 0.51 kcal/mol. Further, we developed a webserver for predicting the stability and it is freely available at (https://web.iitm.ac.in/bioinfo2/TMHS/). The importance of selected parameters and limitations are discussed.

Structure-function crosstalk in liver cancer research: Protein structuromics

Liver cancer can be primary (starting in the liver) or secondary (cancer that has spread from elsewhere to the liver, known as liver metastasis). Liver metastasis is more common than primary liver cancer. Despite great advances in molecular biology methods and treatments, liver cancer is still associated with a poor survival rate and a high death rate, and there is no cure. Many questions remain regarding the mechanisms of liver cancer occurrence and development as well as tumor reoccurrence after treatment. In this study, we assessed the protein structural features of 20 oncogenes and 20 anti-oncogenes via protein structure and dynamic analysis methods and 3D structural and systematic analyses of the structure-function relationships of proteins. Our aim was to provide new insights that may inform research on the development and treatment of liver cancer.

Synergetic Effects of Alanine and Glycine in Blob-Based Methods for Predicting Protein Folding Times

The polypeptide PGlyAlaGlu was prepared with 20 mol % glycine (Gly), 36 mol % d,l-alanine (Ala), and 44 mol % d,l-glutamic acid (Glu) and labeled with the dye 1-pyrenemethylamine to yield a series of Py-PGlyAlaGlu samples. The fluorescence decays of the Py-PGlyAlaGlu samples were analyzed according to the fluorescence blob model (FBM) to obtain the number N_blob^exp of amino acids (aa's) encompassed inside the subvolume V_blob of the polypeptide probed by an excited pyrene. An N_blob^exp value of 29 (±2) was retrieved for Py-PGlyAlaGlu, which was much larger than for any of the copolypeptide PGlyGlu or PAlaGlu prepared with either Gly and Glu or Ala and Glu, respectively. The continuous increase in N_blob^exp with decreasing side chain size (SCS) from 10 aa's for PGlu to 16 aa's for PAlaGlu and 22 aa's for PGlyGlu was used earlier to define the reach of an aa and determine the groups of aa's that could interact with each other along a polypeptide backbone according to their SCS. These groups of aa's, referred to as blobs, led to the implementation of blob-based models (BBM) to predict the folding time τ_F^theo,BBM of 145 proteins, which was found to match their experimental folding time τ_F^exp with a relatively high 0.71 correlation coefficient. Nevertheless, the much higher N_blob^exp value found for Py-PGlyAlaGlu compared to all other pyrene-labeled polypeptides studied to date indicates that the reach of aa's along a polypeptide sequence is affected not only by SCS but also by synergetic effects between different aa's. Following this new insight, a revised BBM was implemented to predict τ_F^theo,BBM for 195 proteins assuming the existence or absence of synergies to control the interactions between aa's along a polypeptide sequence. Similarly good correlation coefficients of 0.71 and 0.74 were obtained for a direct 1:1 comparison of τ_F^exp and τ_F^theo,BBM for the 195 proteins without and with synergies, respectively. This result suggests that synergetic effects between different aa's have little effect on τ_F^theo,BBM predicted from BBM underlying the robustness of this methodology.

Enhanced Protein Structural Class Prediction Using Effective Feature Modeling and Ensemble of Classifiers

Protein Secondary Structural Class (PSSC) information is important in investigating further challenges of protein sequences like protein fold recognition, protein tertiary structure prediction, and analysis of protein functions for drug discovery. Identification of PSSC using biological methods is time-consuming and cost-intensive. Several computational models have been developed to predict the structural class; however, they lack in generalization of the model. Hence, predicting PSSC based on protein sequences is still proving to be an uphill task. In this article, we proposed an effective, novel and generalized prediction model consisting of a feature modeling and an ensemble of classifiers. The proposed feature modeling extracts discriminating information (features) by leveraging three techniques: (i) Embedding - features are extracted on the basis of spatial residue arrangements of the sequences using word embedding approaches; (ii) SkipXGram Bi-gram - various sets of skipped bi-gram features are extracted from the sequences; and (iii) General Statistical (GS) based features are extracted which covers the global information of structural sequences. The combined effective sets of features are trained and classified using an ensemble of three classifiers: Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting Machines (GBM). The proposed model when assessed on five benchmark datasets (high and low sequence similarity), viz. z277, z498, 25PDB, 1189, and FC699, reported an overall accuracy of 93.55, 97.58, 81.82, 81.11, and 93.93 percent respectively. The proposed model is further validated on a large-scale updated low similarity ( ≤ 25%) dataset, where it achieved an overall accuracy of 81.11 percent. The proposed generalized model is robust and consistently outperformed several state-of-the-art models on all the five benchmark datasets.

Study on the Influence of mRNA, the Genetic Language, on Protein Folding Rates

Many works have reported that protein folding rates are influenced by the characteristics of amino acid sequences and protein structures. However, few reports on the problem of whether the corresponding mRNA sequences are related to the protein folding rates can be found. An mRNA sequence is regarded as a kind of genetic language, and its vocabulary and phraseology must provide influential information regarding the protein folding rate. In the present work, linear regressions on the parameters of the vocabulary and phraseology of mRNA sequences and the corresponding protein folding rates were analyzed. The results indicated that D₂ (the adjacent base-related information redundancy) values and the GC content values of the corresponding mRNA sequences exhibit significant negative relations with the protein folding rates, but D₁ (the single base information redundancy) values exhibit significant positive relations with the protein folding rates. In addition, the results show that the relationships between the parameters of the genetic language and the corresponding protein folding rates are obviously different for different protein groups. Some useful parameters that are related to protein folding rates were found. The results indicate that when predicting protein folding rates, the information from protein structures and their amino acid sequences is insufficient, and some information for regulating the protein folding rates must be derived from the mRNA sequences.

AptaNet as a deep learning approach for aptamer-protein interaction prediction

Aptamers are short oligonucleotides (DNA/RNA) or peptide molecules that can selectively bind to their specific targets with high specificity and affinity. As a powerful new class of amino acid ligands, aptamers have high potentials in biosensing, therapeutic, and diagnostic fields. Here, we present AptaNet-a new deep neural network-to predict the aptamer-protein interaction pairs by integrating features derived from both aptamers and the target proteins. Aptamers were encoded by using two different strategies, including k-mer and reverse complement k-mer frequency. Amino acid composition (AAC) and pseudo amino acid composition (PseAAC) were applied to represent target information using 24 physicochemical and conformational properties of the proteins. To handle the imbalance problem in the data, we applied a neighborhood cleaning algorithm. The predictor was constructed based on a deep neural network, and optimal features were selected using the random forest algorithm. As a result, 99.79% accuracy was achieved for the training dataset, and 91.38% accuracy was obtained for the testing dataset. AptaNet achieved high performance on our constructed aptamer-protein benchmark dataset. The results indicate that AptaNet can help identify novel aptamer-protein interacting pairs and build more-efficient insights into the relationship between aptamers and proteins. Our benchmark dataset and the source codes for AptaNet are available in: https://github.com/nedaemami/AptaNet .

The Influences of Palindromes in mRNA on Protein Folding Rates

BACKGROUND

It is currently believed that protein folding rates are influenced by protein structure, environment and temperature, amino acid sequence and so on. We have been working for long to determine whether and in what ways mRNA affects the protein folding rate. A large number of palindromes aroused our attention in our previous research. Whether these palindromes do have important influences on protein folding rates and what's the mechanism? Very few related studies are focused on these problems.

OBJECTIVE

In this article, our motivation is to find out if palindromes have important influences on protein folding rates and what's the mechanism.

METHODS

In this article, the parameters of the palindromes were defined and calculated, the linear regression analysis between the values of each parameter and the experimental protein folding rates were done. Furthermore, to compare the results of different kinds of proteins, proteins were classified into the two-state proteins and the multi-state proteins. For the two kinds of proteins, the above linear regression analysis were performed respectively.

RESULTS

Protein folding rates were negatively correlated to the palindrome frequencies for all proteins. An extremely significant negative linear correlation appeared in the relationship between palindrome densities and protein folding rates. And the repeatedly used bases by different palindromes simultaneously have an important effect on the relationship between palindrome density and protein folding rate.

CONCLUSION

The palindromes have important influences on protein folding rates, and the repeatedly used bases in different palindromes simultaneously play a key role in influencing the protein folding rates.

An Effective Cumulative Torsion Angles Model for Prediction of Protein Folding Rates

BACKGROUND

Protein folding rate is mainly determined by the size of the conformational space to search, which in turn is dictated by factors such as size, structure and amino-acid sequence in a protein. It is important to integrate these factors effectively to form a more precisely description of conformation space. But there is no general paradigm to answer this question except some intuitions and empirical rules. Therefore, at the present stage, predictions of the folding rate can be improved through finding new factors, and some insights are given to the above question.

OBJECTIVE

Its purpose is to propose a new parameter that can describe the size of the conformational space to improve the prediction accuracy of protein folding rate.

METHODS

Based on the optimal set of amino acids in a protein, an effective cumulative backbone torsion angles (CBTAeff) was proposed to describe the size of the conformational space. Linear regression model was used to predict protein folding rate with CBTAeff as a parameter. The degree of correlation was described by the coefficient of determination and the mean absolute error MAE between the predicted folding rates and experimental observations.

RESULTS

It achieved a high correlation (with the coefficient of determination of 0.70 and MAE of 1.88) between the logarithm of folding rates and the (CBTAeff)0.5 with experimental over 112 twoand multi-state folding proteins.

CONCLUSION

The remarkable performance of our simplistic model demonstrates that CBTA based on optimal set was the major determinants of the conformation space of natural proteins.

AggreRATE-Pred: a mathematical model for the prediction of change in aggregation rate upon point mutation

MOTIVATION

Protein aggregation is a major unsolved problem in biochemistry with implications for several human diseases, biotechnology and biomaterial sciences. A majority of sequence-structural properties known for their mechanistic roles in protein aggregation do not correlate well with the aggregation kinetics. This limits the practical utility of predictive algorithms.

RESULTS

We analyzed experimental data on 183 unique single point mutations that lead to change in aggregation rates for 23 polypeptides and proteins. Our initial mathematical model obtained a correlation coefficient of 0.43 between predicted and experimental change in aggregation rate upon mutation (P-value <0.0001). However, when the dataset was classified based on protein length and conformation at the mutation sites, the average correlation coefficient almost doubled to 0.82 (range: 0.74-0.87; P-value <0.0001). We observed that distinct sequence and structure-based properties determine protein aggregation kinetics in each class. In conclusion, the protein aggregation kinetics are impacted by local factors and not by global ones, such as overall three-dimensional protein fold, or mechanistic factors such as the presence of aggregation-prone regions.

AVAILABILITY AND IMPLEMENTATION

The web server is available at http://www.iitm.ac.in/bioinfo/aggrerate-pred/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

MPTherm-pred: Analysis and Prediction of Thermal Stability Changes upon Mutations in Transmembrane Proteins

The stability of membrane proteins differs from globular proteins due to the presence of nonpolar membrane-spanning regions. Using a dataset of 929 membrane protein mutations whose effects on thermal stability (ΔT_m) were experimentally determined, we found that the average ΔT_m due to 190 stabilizing and 232 destabilizing mutations occurring in membrane-spanning regions are 2.43(3.1) °C and -5.48(5.5) °C, respectively. The ΔT_m values for mutations occurring in solvent-exposed regions are 2.56(2.82) and - 6.8(7.2) °C. We have systematically analyzed the factors influencing the stability of mutants and observed that changes in hydrophobicity, number of contacts between Cα atoms and frequency of aliphatic residues are important determinants of the stability change induced by mutations occurring in membrane-spanning regions. We have developed structure- and sequence-based machine learning predictors of ΔT_m due to mutations specifically for membrane proteins. They showed a correlation and mean absolute error (MAE) of 0.72 and 2.85 °C, respectively, between experimental and predicted ΔT_m for mutations in membrane-spanning regions on 10-fold group-wise cross-validation. The average correlation and MAE for mutations in aqueous regions are 0.73 and 3.7 °C, respectively. These MAE values are about 50% lower than standard deviations from the mean ΔT_m values. The reliability of the method was affirmed on a test set of mutations occurring in evolutionary independent protein sequences. The developed MPTherm-pred server for predicting thermal stability changes upon mutations in membrane proteins is available at https://web.iitm.ac.in/bioinfo2/mpthermpred/. Our results provide insights into factors influencing the stability of membrane proteins and can aid in designing mutants that are more resistant to thermal stress.

Mutations in transmembrane proteins: diseases, evolutionary insights, prediction and comparison with globular proteins

Membrane proteins are unique in that they interact with lipid bilayers, making them indispensable for transporting molecules and relaying signals between and across cells. Due to the significance of the protein's functions, mutations often have profound effects on the fitness of the host. This is apparent both from experimental studies, which implicated numerous missense variants in diseases, as well as from evolutionary signals that allow elucidating the physicochemical constraints that intermembrane and aqueous environments bring. In this review, we report on the current state of knowledge acquired on missense variants (referred to as to single amino acid variants) affecting membrane proteins as well as the insights that can be extrapolated from data already available. This includes an overview of the annotations for membrane protein variants that have been collated within databases dedicated to the topic, bioinformatics approaches that leverage evolutionary information in order to shed light on previously uncharacterized membrane protein structures or interaction interfaces, tools for predicting the effects of mutations tailored specifically towards the characteristics of membrane proteins as well as two clinically relevant case studies explaining the implications of mutated membrane proteins in cancer and cardiomyopathy.

Solution of Levinthal's Paradox and a Physical Theory of Protein Folding Times

“How do proteins fold?” Researchers have been studying different aspects of this question for more than 50 years. The most conceptual aspect of the problem is how protein can find the global free energy minimum in a biologically reasonable time, without exhaustive enumeration of all possible conformations, the so-called “Levinthal’s paradox.” Less conceptual but still critical are aspects about factors defining folding times of particular proteins and about perspectives of machine learning for their prediction. We will discuss in this review the key ideas and discoveries leading to the current understanding of folding kinetics, including the solution of Levinthal’s paradox, as well as the current state of the art in the prediction of protein folding times.

Seq2Feature: a comprehensive web-based feature extraction tool

Motivation

Machine learning techniques require various descriptors from protein and nucleic acid sequences to understand/predict their structure and function as well as distinguishing between disease and neutral mutations. Hence, availability of a feature extraction tool is necessary to bridge the gap.

Results

We developed a comprehensive web-based tool, Seq2Feature, which computes 252 protein and 41 DNA sequence-based descriptors. These features include physicochemical, energetic and conformational properties of proteins, mutation matrices and contact potentials as well as nucleotide composition, physicochemical and conformational properties of DNA. We propose that Seq2Feature could serve as an effective tool for extracting protein and DNA sequence-based features as applicable inputs to machine learning algorithms.

Availability and implementation

https://www.iitm.ac.in/bioinfo/SBFE/index.html.

Supplementary information

Supplementary data are available at Bioinformatics online.

ProBAPred: Inferring protein–protein binding affinity by incorporating protein sequence and structural features

Protein-protein binding interaction is the most prevalent biological activity that mediates a great variety of biological processes. The increasing availability of experimental data of protein–protein interaction allows a systematic construction of protein–protein interaction networks, significantly contributing to a better understanding of protein functions and their roles in cellular pathways and human diseases. Compared to well-established classification for protein–protein interactions (PPIs), limited work has been conducted for estimating protein–protein binding free energy, which can provide informative real-value regression models for characterizing the protein–protein binding affinity. In this study, we propose a novel ensemble computational framework, termed ProBAPred (Protein–protein Binding Affinity Predictor), for quantitative estimation of protein–protein binding affinity. A large number of sequence and structural features, including physical–chemical properties, binding energy and conformation annotations, were collected and calculated from currently available protein binding complex datasets and the literature. Feature selection based on the WEKA package was performed to identify and characterize the most informative and contributing feature subsets. Experiments on the independent test showed that our ensemble method achieved the lowest Mean Absolute Error (MAE; 1.657[Formula: see text]kcal/mol) and the second highest correlation coefficient ([Formula: see text]), compared with the existing methods. The datasets and source codes of ProBAPred, and the supplementary materials in this study can be downloaded at http://lightning.med.monash.edu/probapred/ for academic use. We anticipate that the developed ProBAPred regression models can facilitate computational characterization and experimental studies of protein–protein binding affinity.

Thermosensitivity of growth is determined by chaperone-mediated proteome reallocation

Maintenance of a properly folded proteome is critical for bacterial survival at notably different growth temperatures. Understanding the molecular basis of thermoadaptation has progressed in two main directions, the sequence and structural basis of protein thermostability and the mechanistic principles of protein quality control assisted by chaperones. Yet we do not fully understand how structural integrity of the entire proteome is maintained under stress and how it affects cellular fitness. To address this challenge, we reconstruct a genome-scale protein-folding network for Escherichia coli and formulate a computational model, FoldME, that provides statistical descriptions of multiscale cellular response consistent with many datasets. FoldME simulations show (i) that the chaperones act as a system when they respond to unfolding stress rather than achieving efficient folding of any single component of the proteome, (ii) how the proteome is globally balanced between chaperones for folding and the complex machinery synthesizing the proteins in response to perturbation, (iii) how this balancing determines growth rate dependence on temperature and is achieved through nonspecific regulation, and (iv) how thermal instability of the individual protein affects the overall functional state of the proteome. Overall, these results expand our view of cellular regulation, from targeted specific control mechanisms to global regulation through a web of nonspecific competing interactions that modulate the optimal reallocation of cellular resources. The methodology developed in this study enables genome-scale integration of environment-dependent protein properties and a proteome-wide study of cellular stress responses.

A novel model-based on FCM-LM algorithm for prediction of protein folding rate

The prediction of protein folding rates is of paramount importance in describing the protein folding mechanism, which has broad applications in fields such as enzyme engineering and protein engineering. Therefore, predicting protein folding rates using the first-order of protein sequence, secondary structure and amino acid properties has become a very active research topic in recent years. This paper presents a new fuzzy cognitive map (FCM) model based on deep learning neural networks which uses data obtained from biological experiments to predict the protein folding rate. FCM extracts the important data features from the protein sequence which then initializes the deep neural networks effectively. It was found that the Levenberg-Marquardt (LM) algorithm for deep neural networks can improve the prediction accuracy of the protein folding rates. The correlation coefficient between the predicted values and those real values obtained from experiments reached 0.94 and 0.9 in two independent numerical tests.

Characterizing informative sequence descriptors and predicting binding affinities of heterodimeric protein complexes

Background

Protein-protein interactions (PPIs) are involved in various biological processes, and underlying mechanism of the interactions plays a crucial role in therapeutics and protein engineering. Most machine learning approaches have been developed for predicting the binding affinity of protein-protein complexes based on structure and functional information. This work aims to predict the binding affinity of heterodimeric protein complexes from sequences only.

Results

This work proposes a support vector machine (SVM) based binding affinity classifier, called SVM-BAC, to classify heterodimeric protein complexes based on the prediction of their binding affinity. SVM-BAC identified 14 of 580 sequence descriptors (physicochemical, energetic and conformational properties of the 20 amino acids) to classify 216 heterodimeric protein complexes into low and high binding affinity. SVM-BAC yielded the training accuracy, sensitivity, specificity, AUC and test accuracy of 85.80%, 0.89, 0.83, 0.86 and 83.33%, respectively, better than existing machine learning algorithms. The 14 features and support vector regression were further used to estimate the binding affinities (Pkd) of 200 heterodimeric protein complexes. Prediction performance of a Jackknife test was the correlation coefficient of 0.34 and mean absolute error of 1.4. We further analyze three informative physicochemical properties according to their contribution to prediction performance. Results reveal that the following properties are effective in predicting the binding affinity of heterodimeric protein complexes: apparent partition energy based on buried molar fractions, relations between chemical structure and biological activity in principal component analysis IV, and normalized frequency of beta turn.

Conclusions

The proposed sequence-based prediction method SVM-BAC uses an optimal feature selection method to identify 14 informative features to classify and predict binding affinity of heterodimeric protein complexes. The characterization analysis revealed that the average numbers of beta turns and hydrogen bonds at protein-protein interfaces in high binding affinity complexes are more than those in low binding affinity complexes.

Machine Learning: How Much Does It Tell about Protein Folding Rates?

The prediction of protein folding rates is a necessary step towards understanding the principles of protein folding. Due to the increasing amount of experimental data, numerous protein folding models and predictors of protein folding rates have been developed in the last decade. The problem has also attracted the attention of scientists from computational fields, which led to the publication of several machine learning-based models to predict the rate of protein folding. Some of them claim to predict the logarithm of protein folding rate with an accuracy greater than 90%. However, there are reasons to believe that such claims are exaggerated due to large fluctuations and overfitting of the estimates. When we confronted three selected published models with new data, we found a much lower predictive power than reported in the original publications. Overly optimistic predictive powers appear from violations of the basic principles of machine-learning. We highlight common misconceptions in the studies claiming excessive predictive power and propose to use learning curves as a safeguard against those mistakes. As an example, we show that the current amount of experimental data is insufficient to build a linear predictor of logarithms of folding rates based on protein amino acid composition.

Exploring preferred amino acid mutations in cancer genes: Applications to identify potential drug targets

Somatic mutations developed with missense, silent, insertions and deletions have varying effects on the resulting protein and are one of the important reasons for cancer development. In this study, we have systematically analysed the effect of these mutations at protein level in 41 different cancer types from COSMIC database on different perspectives: (i) Preference of residues at the mutant positions, (ii) probability of substitutions, (iii) influence of neighbouring residues in driver and passenger mutations, (iv) distribution of driver and passenger mutations around hotspot site in five typical genes and (v) distribution of silent and missense substitutions. We observed that R→H substitution is dominant in drivers followed by R→Q and R→C whereas E→K has the highest preference in passenger mutations. A set of 17 mutations including R→Y, W→A and V→R are specific to driver mutations and 31 preferred substitutions are observed only in passenger mutations. These frequencies of driver mutations vary across different cancer types and are selective to specific tissues. Further, driver missense mutations are mainly surrounded with silent driver mutations whereas the passenger missense mutations are surrounded with silent passenger mutations. This study reveals the variation of mutations at protein level in different cancer types and their preferences in cancer genes and provides new insights for understanding cancer mutations and drug development.

Discrimination of driver and passenger mutations in epidermal growth factor receptor in cancer

Cancer is one of the most life-threatening diseases and mutations in several genes are the vital cause in tumorigenesis. Protein kinases play essential roles in cancer progression and specifically, epidermal growth factor receptor (EGFR) is an important target for cancer therapy. In this work, we have developed a method to classify single amino acid polymorphisms (SAPs) in EGFR into disease-causing (driver) and neutral (passenger) mutations using both sequence and structure based features of the mutation site by machine learning approaches. We compiled a set of 222 features and selected a set of 21 properties utilizing feature selection methods, for maximizing the prediction performance. In a set of 540 mutants, we obtained an overall classification accuracy of 67.8% with 10 fold cross validation using support vector machines. Further, the mutations have been grouped into four sets based on secondary structure and accessible surface area, which enhanced the overall classification accuracy to 80.2%, 81.9%, 77.9% and 75.1% for helix, strand, coil-buried and coil-exposed mutants, respectively. The method was tested with a blind dataset of 60 mutations, which showed an average accuracy of 85.4%. These accuracy levels are superior to other methods available in the literature for EGFR mutants, with an increase of more than 30%. Moreover, we have screened all possible single amino acid polymorphisms (SAPs) in EGFR and suggested the probable driver and passenger mutations, which would help in the development of mutation specific drugs for cancer treatment.

Abundance of intrinsic structural disorder in the histone H1 subtypes

The intrinsically disordered proteins consist of partially structured regions linked to the unstructured stretches, which consequently form the transient and dynamic conformational ensembles. They undergo disorder to order transition upon binding their partners. Intrinsic disorder is attributed to histones H1, perceived as assemblers of chromatin structure and the regulators of DNA and proteins activity. In this work, the comparison of intrinsic disorder abundance in the histone H1 subtypes was performed both by the analysis of their amino acid composition and by the prediction of disordered stretches, as well as by identifying molecular recognition features (MoRFs) and ANCHOR protein binding regions (APBR) that are responsible for recognition and binding. Both human and model organisms-animals, plants, fungi and protists-have H1 histone subtypes with the properties typical of disordered state. They possess a significantly higher content of hydrophilic and charged amino acid residues, arranged in the long regions, covering over half of the whole amino acid residues in chain. Almost complete disorder corresponds to histone H1 terminal domains, including MoRFs and ANCHOR. Those motifs were also identified in a more ordered histone H1 globular domain. Compared to the control (globular and fibrous) proteins, H1 histones demonstrate the increased folding rate and a higher proportion of low-complexity segments. The results of this work indicate that intrinsic disorder is an inherent structural property of histone H1 subtypes and it is essential for establishing a protein conformation which defines functional outcomes affecting on DNA- and/or partner protein-dependent cell processes.

Prediction of protein folding rates from simplified secondary structure alphabet

Protein folding is a very complicated and highly cooperative dynamic process. However, the folding kinetics is likely to depend more on a few key structural features. Here we find that secondary structures can determine folding rates of only large, multi-state folding proteins and fails to predict those for small, two-state proteins. The importance of secondary structures for protein folding is ordered as: extended β strand > α helix > bend > turn > undefined secondary structure>310 helix > isolated β strand > π helix. Only the first three secondary structures, extended β strand, α helix and bend, can achieve a good correlation with folding rates. This suggests that the rate-limiting step of protein folding would depend upon the formation of regular secondary structures and the buckling of chain. The reduced secondary structure alphabet provides a simplified description for the machine learning applications in protein design.

Gram-positive and Gram-negative subcellular localization using rotation forest and physicochemical-based features

Background

The functioning of a protein relies on its location in the cell. Therefore, predicting protein subcellular localization is an important step towards protein function prediction. Recent studies have shown that relying on Gene Ontology (GO) for feature extraction can improve the prediction performance. However, for newly sequenced proteins, the GO is not available. Therefore, for these cases, the prediction performance of GO based methods degrade significantly.

Results

In this study, we develop a method to effectively employ physicochemical and evolutionary-based information in the protein sequence. To do this, we propose segmentation based feature extraction method to explore potential discriminatory information based on physicochemical properties of the amino acids to tackle Gram-positive and Gram-negative subcellular localization. We explore our proposed feature extraction techniques using 10 attributes that have been experimentally selected among a wide range of physicochemical attributes. Finally by applying the Rotation Forest classification technique to our extracted features, we enhance Gram-positive and Gram-negative subcellular localization accuracies up to 3.4% better than previous studies which used GO for feature extraction.

Conclusion

By proposing segmentation based feature extraction method to explore potential discriminatory information based on physicochemical properties of the amino acids as well as using Rotation Forest classification technique, we are able to enhance the Gram-positive and Gram-negative subcellular localization prediction accuracies, significantly.

Folding RaCe: a robust method for predicting changes in protein folding rates upon point mutations

MOTIVATION

Protein engineering methods are commonly employed to decipher the folding mechanism of proteins and enzymes. However, such experiments are exceedingly time and resource intensive. It would therefore be advantageous to develop a simple computational tool to predict changes in folding rates upon mutations. Such a method should be able to rapidly provide the sequence position and chemical nature to modulate through mutation, to effect a particular change in rate. This can be of importance in protein folding, function or mechanistic studies.

RESULTS

We have developed a robust knowledge-based methodology to predict the changes in folding rates upon mutations formulated from amino and acid properties using multiple linear regression approach. We benchmarked this method against an experimental database of 790 point mutations from 26 two-state proteins. Mutants were first classified according to secondary structure, accessible surface area and position along the primary sequence. Three prime amino acid features eliciting the best relationship with folding rates change were then shortlisted for each class along with an optimized window length. We obtained a self-consistent mean absolute error of 0.36 s(-1) and a mean Pearson correlation coefficient (PCC) of 0.81. Jack-knife test resulted in a MAE of 0.42 s(-1) and a PCC of 0.73. Moreover, our method highlights the importance of outlier(s) detection and studying their implications in the folding mechanism.

AVAILABILITY AND IMPLEMENTATION

A web server 'Folding RaCe' has been developed and is available at http://www.iitm.ac.in/bioinfo/proteinfolding/foldingrace.html.

CONTACT

gromiha@iitm.ac.in

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Reduced alphabet for protein folding prediction

What are the key building blocks that would have been needed to construct complex protein folds? This is an important issue for understanding protein folding mechanism and guiding de novo protein design. Twenty naturally occurring amino acids and eight secondary structures consist of a 28-letter alphabet to determine folding kinetics and mechanism. Here we predict folding kinetic rates of proteins from many reduced alphabets. We find that a reduced alphabet of 10 letters achieves good correlation with folding rates, close to the one achieved by full 28-letter alphabet. Many other reduced alphabets are not significantly correlated to folding rates. The finding suggests that not all amino acids and secondary structures are equally important for protein folding. The foldable sequence of a protein could be designed using at least 10 folding units, which can either promote or inhibit protein folding. Reducing alphabet cardinality without losing key folding kinetic information opens the door to potentially faster machine learning and data mining applications in protein structure prediction, sequence alignment and protein design.

General mechanism of two-state protein folding kinetics

We describe here a general model of the kinetic mechanism of protein folding. In the Foldon Funnel Model, proteins fold in units of secondary structures, which form sequentially along the folding pathway, stabilized by tertiary interactions. The model predicts that the free energy landscape has a volcano shape, rather than a simple funnel, that folding is two-state (single-exponential) when secondary structures are intrinsically unstable, and that each structure along the folding path is a transition state for the previous structure. It shows how sequential pathways are consistent with multiple stochastic routes on funnel landscapes, and it gives good agreement with the 9 order of magnitude dependence of folding rates on protein size for a set of 93 proteins, at the same time it is consistent with the near independence of folding equilibrium constant on size. This model gives estimates of folding rates of proteomes, leading to a median folding time in Escherichia coli of about 5 s.

PSNO: predicting cysteine S-nitrosylation sites by incorporating various sequence-derived features into the general form of Chou's PseAAC

S-nitrosylation (SNO) is one of the most universal reversible post-translational modifications involved in many biological processes. Malfunction or dysregulation of SNO leads to a series of severe diseases, such as developmental abnormalities and various diseases. Therefore, the identification of SNO sites (SNOs) provides insights into disease progression and drug development. In this paper, a new bioinformatics tool, named PSNO, is proposed to identify SNOs from protein sequences. Firstly, we explore various promising sequence-derived discriminative features, including the evolutionary profile, the predicted secondary structure and the physicochemical properties. Secondly, rather than simply combining the features, which may bring about information redundancy and unwanted noise, we use the relative entropy selection and incremental feature selection approach to select the optimal feature subsets. Thirdly, we train our model by the technique of the k-nearest neighbor algorithm. Using both informative features and an elaborate feature selection scheme, our method, PSNO, achieves good prediction performance with a mean Mathews correlation coefficient (MCC) value of about 0.5119 on the training dataset using 10-fold cross-validation. These results indicate that PSNO can be used as a competitive predictor among the state-of-the-art SNOs prediction tools. A web-server, named PSNO, which implements the proposed method, is freely available at http://59.73.198.144:8088/PSNO/.

Computational and experimental approaches to reveal the effects of single nucleotide polymorphisms with respect to disease diagnostics

DNA mutations are the cause of many human diseases and they are the reason for natural differences among individuals by affecting the structure, function, interactions, and other properties of DNA and expressed proteins. The ability to predict whether a given mutation is disease-causing or harmless is of great importance for the early detection of patients with a high risk of developing a particular disease and would pave the way for personalized medicine and diagnostics. Here we review existing methods and techniques to study and predict the effects of DNA mutations from three different perspectives: in silico, in vitro and in vivo. It is emphasized that the problem is complicated and successful detection of a pathogenic mutation frequently requires a combination of several methods and a knowledge of the biological phenomena associated with the corresponding macromolecules.

Feature selection and classification of protein-protein complexes based on their binding affinities using machine learning approaches

Protein-protein interactions are intrinsic to virtually every cellular process. Predicting the binding affinity of protein-protein complexes is one of the challenging problems in computational and molecular biology. In this work, we related sequence features of protein-protein complexes with their binding affinities using machine learning approaches. We set up a database of 185 protein-protein complexes for which the interacting pairs are heterodimers and their experimental binding affinities are available. On the other hand, we have developed a set of 610 features from the sequences of protein complexes and utilized Ranker search method, which is the combination of Attribute evaluator and Ranker method for selecting specific features. We have analyzed several machine learning algorithms to discriminate protein-protein complexes into high and low affinity groups based on their Kd values. Our results showed a 10-fold cross-validation accuracy of 76.1% with the combination of nine features using support vector machines. Further, we observed accuracy of 83.3% on an independent test set of 30 complexes. We suggest that our method would serve as an effective tool for identifying the interacting partners in protein-protein interaction networks and human-pathogen interactions based on the strength of interactions.