A novel method for protein-protein interaction site prediction using phylogenetic substitution models

EGRET: edge aggregated graph attention networks and transfer learning improve protein-protein interaction site prediction

MOTIVATION

Protein-protein interactions (PPIs) are central to most biological processes. However, reliable identification of PPI sites using conventional experimental methods is slow and expensive. Therefore, great efforts are being put into computational methods to identify PPI sites.

RESULTS

We present Edge Aggregated GRaph Attention NETwork (EGRET), a highly accurate deep learning-based method for PPI site prediction, where we have used an edge aggregated graph attention network to effectively leverage the structural information. We, for the first time, have used transfer learning in PPI site prediction. Our proposed edge aggregated network, together with transfer learning, has achieved notable improvement over the best alternate methods. Furthermore, we systematically investigated EGRET's network behavior to provide insights about the causes of its decisions.

AVAILABILITY

EGRET is freely available as an open source project at https://github.com/Sazan-Mahbub/EGRET.

CONTACT

shams_bayzid@cse.buet.ac.bd.

A Deep Learning and XGBoost-Based Method for Predicting Protein-Protein Interaction Sites

Knowledge about protein-protein interactions is beneficial in understanding cellular mechanisms. Protein-protein interactions are usually determined according to their protein-protein interaction sites. Due to the limitations of current techniques, it is still a challenging task to detect protein-protein interaction sites. In this article, we presented a method based on deep learning and XGBoost (called DeepPPISP-XGB) for predicting protein-protein interaction sites. The deep learning model served as a feature extractor to remove redundant information from protein sequences. The Extreme Gradient Boosting algorithm was used to construct a classifier for predicting protein-protein interaction sites. The DeepPPISP-XGB achieved the following results: area under the receiver operating characteristic curve of 0.681, a recall of 0.624, and area under the precision-recall curve of 0.339, being competitive with the state-of-the-art methods. We also validated the positive role of global features in predicting protein-protein interaction sites.

LZerD Protein-Protein Docking Webserver Enhanced With de novo Structure Prediction

Protein-protein docking is a useful tool for modeling the structures of protein complexes that have yet to be experimentally determined. Understanding the structures of protein complexes is a key component for formulating hypotheses in biophysics regarding the functional mechanisms of complexes. Protein-protein docking is an established technique for cases where the structures of the subunits have been determined. While the number of known structures deposited in the Protein Data Bank is increasing, there are still many cases where the structures of individual proteins that users want to dock are not determined yet. Here, we have integrated the AttentiveDist method for protein structure prediction into our LZerD webserver for protein-protein docking, which enables users to simply submit protein sequences and obtain full-complex atomic models, without having to supply any structure themselves. We have further extended the LZerD docking interface with a symmetrical homodimer mode. The LZerD server is available at https://lzerd.kiharalab.org/.

LZerD webserver for pairwise and multiple protein-protein docking

Protein complexes are involved in many important processes in living cells. To understand the mechanisms of these processes, it is necessary to solve the 3D structures of the protein complexes. When protein complex structures have not yet been determined by experiment, protein-protein docking tools can be used to computationally model the structures of these complexes. Here, we present a webserver which provides access to LZerD and Multi-LZerD protein docking tools. The protocol provided by the server have performed consistently among the top in the CAPRI blind evaluation. LZerD docks pairs of structures, while Multi-LZerD can dock three or more structures simultaneously. LZerD uses a soft protein surface representation with 3D Zernike descriptors and explores the binding pose space using geometric hashing. Multi-LZerD performs multi-chain docking by combining pairwise solutions by LZerD. Both methods output full-atom docked models of the input proteins. Users can also input distance constraints between interacting or non-interacting residues as well as residues that locate at the interface or far from the interface. The webserver is equipped with a user-friendly panel that visualizes the distribution and structures of binding poses of top scoring models. The LZerD webserver is available at https://lzerd.kiharalab.org.

Computational structure modeling for diverse categories of macromolecular interactions

Computational protein-protein docking is one of the most intensively studied topics in structural bioinformatics. The field has made substantial progress through over three decades of development. The development began with methods for rigid-body docking of two proteins, which have now been extended in different directions to cover the various macromolecular interactions observed in a cell. Here, we overview the recent developments of the variations of docking methods, including multiple protein docking, peptide-protein docking, and disordered protein docking methods.

A Two-Layer SVM Ensemble-Classifier to Predict Interface Residue Pairs of Protein Trimers

Study of interface residue pairs is important for understanding the interactions between monomers inside a trimer protein–protein complex. We developed a two-layer support vector machine (SVM) ensemble-classifier that considers physicochemical and geometric properties of amino acids and the influence of surrounding amino acids. Different descriptors and different combinations may give different prediction results. We propose feature combination engineering based on correlation coefficients and F-values. The accuracy of our method is 65.38% in independent test set, indicating biological significance. Our predictions are consistent with the experimental results. It shows the effectiveness and reliability of our method to predict interface residue pairs of protein trimers.

Predicting Protein-Protein Interfaces that Bind Intrinsically Disordered Protein Regions

A long-standing goal in biology is the complete annotation of function and structure on all protein-protein interactions, a large fraction of which is mediated by intrinsically disordered protein regions (IDRs). However, knowledge derived from experimental structures of such protein complexes is disproportionately small due, in part, to challenges in studying interactions of IDRs. Here, we introduce IDRBind, a computational method that by combining gradient boosted trees and conditional random field models predicts binding sites of IDRs with performance approaching state-of-the-art globular interface predictions, making it suitable for proteome-wide applications. Although designed and trained with a focus on molecular recognition features, which are long interaction-mediating-elements in IDRs, IDRBind also predicts the binding sites of short peptides more accurately than existing specialized predictors. Consistent with IDRBind's specificity, a comparison of protein interface categories uncovered uniform trends in multiple physicochemical properties, positioning molecular recognition feature interfaces between peptide and globular interfaces.

Modeling disordered protein interactions from biophysical principles

Disordered protein-protein interactions (PPIs), those involving a folded protein and an intrinsically disordered protein (IDP), are prevalent in the cell, including important signaling and regulatory pathways. IDPs do not adopt a single dominant structure in isolation but often become ordered upon binding. To aid understanding of the molecular mechanisms of disordered PPIs, it is crucial to obtain the tertiary structure of the PPIs. However, experimental methods have difficulty in solving disordered PPIs and existing protein-protein and protein-peptide docking methods are not able to model them. Here we present a novel computational method, IDP-LZerD, which models the conformation of a disordered PPI by considering the biophysical binding mechanism of an IDP to a structured protein, whereby a local segment of the IDP initiates the interaction and subsequently the remaining IDP regions explore and coalesce around the initial binding site. On a dataset of 22 disordered PPIs with IDPs up to 69 amino acids, successful predictions were made for 21 bound and 18 unbound receptors. The successful modeling provides additional support for biophysical principles. Moreover, the new technique significantly expands the capability of protein structure modeling and provides crucial insights into the molecular mechanisms of disordered PPIs.

Human and server docking prediction for CAPRI round 30-35 using LZerD with combined scoring functions

We report the performance of protein-protein docking predictions by our group for recent rounds of the Critical Assessment of Prediction of Interactions (CAPRI), a community-wide assessment of state-of-the-art docking methods. Our prediction procedure uses a protein-protein docking program named LZerD developed in our group. LZerD represents a protein surface with 3D Zernike descriptors (3DZD), which are based on a mathematical series expansion of a 3D function. The appropriate soft representation of protein surface with 3DZD makes the method more tolerant to conformational change of proteins upon docking, which adds an advantage for unbound docking. Docking was guided by interface residue prediction performed with BindML and cons-PPISP as well as literature information when available. The generated docking models were ranked by a combination of scoring functions, including PRESCO, which evaluates the native-likeness of residues' spatial environments in structure models. First, we discuss the overall performance of our group in the CAPRI prediction rounds and investigate the reasons for unsuccessful cases. Then, we examine the performance of several knowledge-based scoring functions and their combinations for ranking docking models. It was found that the quality of a pool of docking models generated by LZerD, that is whether or not the pool includes near-native models, can be predicted by the correlation of multiple scores. Although the current analysis used docking models generated by LZerD, findings on scoring functions are expected to be universally applicable to other docking methods. Proteins 2017; 85:513-527. © 2016 Wiley Periodicals, Inc.

Integrating computational methods and experimental data for understanding the recognition mechanism and binding affinity of protein-protein complexes

Protein-protein interactions perform several functions inside the cell. Understanding the recognition mechanism and binding affinity of protein-protein complexes is a challenging problem in experimental and computational biology. In this review, we focus on two aspects (i) understanding the recognition mechanism and (ii) predicting the binding affinity. The first part deals with computational techniques for identifying the binding site residues and the contribution of important interactions for understanding the recognition mechanism of protein-protein complexes in comparison with experimental observations. The second part is devoted to the methods developed for discriminating high and low affinity complexes, and predicting the binding affinity of protein-protein complexes using three-dimensional structural information and just from the amino acid sequence. The overall view enhances our understanding of the integration of experimental data and computational methods, recognition mechanism of protein-protein complexes and the binding affinity.

BindML/BindML+: Detecting Protein-Protein Interaction Interface Propensity from Amino Acid Substitution Patterns

Prediction of protein-protein interaction sites in a protein structure provides important information for elucidating the mechanism of protein function and can also be useful in guiding a modeling or design procedures of protein complex structures. Since prediction methods essentially assess the propensity of amino acids that are likely to be part of a protein docking interface, they can help in designing protein-protein interactions. Here, we introduce BindML and BindML+ protein-protein interaction sites prediction methods. BindML predicts protein-protein interaction sites by identifying mutation patterns found in known protein-protein complexes using phylogenetic substitution models. BindML+ is an extension of BindML for distinguishing permanent and transient types of protein-protein interaction sites. We developed an interactive web-server that provides a convenient interface to assist in structural visualization of protein-protein interactions site predictions. The input data for the web-server are a tertiary structure of interest. BindML and BindML+ are available at http://kiharalab.org/bindml/ and http://kiharalab.org/bindml/plus/ .

Computational Approaches for Predicting Binding Partners, Interface Residues, and Binding Affinity of Protein-Protein Complexes

Studying protein-protein interactions leads to a better understanding of the underlying principles of several biological pathways. Cost and labor-intensive experimental techniques suggest the need for computational methods to complement them. Several such state-of-the-art methods have been reported for analyzing diverse aspects such as predicting binding partners, interface residues, and binding affinity for protein-protein complexes with reliable performance. However, there are specific drawbacks for different methods that indicate the need for their improvement. This review highlights various available computational algorithms for analyzing diverse aspects of protein-protein interactions and endorses the necessity for developing new robust methods for gaining deep insights about protein-protein interactions.

Cryo-EM Data Are Superior to Contact and Interface Information in Integrative Modeling

Protein-protein interactions carry out a large variety of essential cellular processes. Cryo-electron microscopy (cryo-EM) is a powerful technique for the modeling of protein-protein interactions at a wide range of resolutions, and recent developments have caused a revolution in the field. At low resolution, cryo-EM maps can drive integrative modeling of the interaction, assembling existing structures into the map. Other experimental techniques can provide information on the interface or on the contacts between the monomers in the complex. This inevitably raises the question regarding which type of data is best suited to drive integrative modeling approaches. Systematic comparison of the prediction accuracy and specificity of the different integrative modeling paradigms is unavailable to date. Here, we compare EM-driven, interface-driven, and contact-driven integrative modeling paradigms. Models were generated for the protein docking benchmark using the ATTRACT docking engine and evaluated using the CAPRI two-star criterion. At 20 Å resolution, EM-driven modeling achieved a success rate of 100%, outperforming the other paradigms even with perfect interface and contact information. Therefore, even very low resolution cryo-EM data is superior in predicting heterodimeric and heterotrimeric protein assemblies. Our study demonstrates that a force field is not necessary, cryo-EM data alone is sufficient to accurately guide the monomers into place. The resulting rigid models successfully identify regions of conformational change, opening up perspectives for targeted flexible remodeling.

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.

Algorithmic approaches to protein-protein interaction site prediction

Interaction sites on protein surfaces mediate virtually all biological activities, and their identification holds promise for disease treatment and drug design. Novel algorithmic approaches for the prediction of these sites have been produced at a rapid rate, and the field has seen significant advancement over the past decade. However, the most current methods have not yet been reviewed in a systematic and comprehensive fashion. Herein, we describe the intricacies of the biological theory, datasets, and features required for modern protein-protein interaction site (PPIS) prediction, and present an integrative analysis of the state-of-the-art algorithms and their performance. First, the major sources of data used by predictors are reviewed, including training sets, evaluation sets, and methods for their procurement. Then, the features employed and their importance in the biological characterization of PPISs are explored. This is followed by a discussion of the methodologies adopted in contemporary prediction programs, as well as their relative performance on the datasets most recently used for evaluation. In addition, the potential utility that PPIS identification holds for rational drug design, hotspot prediction, and computational molecular docking is described. Finally, an analysis of the most promising areas for future development of the field is presented.

A novel feature extraction scheme for prediction of protein-protein interaction sites

Identifying protein-protein interaction (PPI) sites plays an important and challenging role in some topics of biology. Although many methods have been proposed, this problem is still far away to be solved. Here, a feature selection approach with an 11-sliding window and random forest algorithm is proposed, which is called DX-RF. This method has achieved an accuracy of 88.79%, recall of 82.09%, and precision of 85.76% with top-ranked 34 features on the Hetero test dataset and has 91.6% accuracy, 89.2% precision, 83.54% recall with top-ranked 25 features set on the Homo test dataset. Compared to other methods, the results indicate that the DX-RF method has a strong ability to select relevance features to get a higher performance. Moreover, in order to further understand protein interactions, feature analysis in this study is also performed.

Scoring docking conformations using predicted protein interfaces

Background

Since proteins function by interacting with other molecules, analysis of protein-protein interactions is essential for comprehending biological processes. Whereas understanding of atomic interactions within a complex is especially useful for drug design, limitations of experimental techniques have restricted their practical use. Despite progress in docking predictions, there is still room for improvement. In this study, we contribute to this topic by proposing T-PioDock, a framework for detection of a native-like docked complex 3D structure. T-PioDock supports the identification of near-native conformations from 3D models that docking software produced by scoring those models using binding interfaces predicted by the interface predictor, Template based Protein Interface Prediction (T-PIP).

Results

First, exhaustive evaluation of interface predictors demonstrates that T-PIP, whose predictions are customised to target complexity, is a state-of-the-art method. Second, comparative study between T-PioDock and other state-of-the-art scoring methods establishes T-PioDock as the best performing approach. Moreover, there is good correlation between T-PioDock performance and quality of docking models, which suggests that progress in docking will lead to even better results at recognising near-native conformations.

Conclusion

Accurate identification of near-native conformations remains a challenging task. Although availability of 3D complexes will benefit from template-based methods such as T-PioDock, we have identified specific limitations which need to be addressed. First, docking software are still not able to produce native like models for every target. Second, current interface predictors do not explicitly consider pairwise residue interactions between proteins and their interacting partners which leaves ambiguity when assessing quality of complex conformations.

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.

Pairwise and multimeric protein-protein docking using the LZerD program suite

Physical interactions between proteins are involved in many important cell functions and are key for understanding the mechanisms of biological processes. Protein-protein docking programs provide a means to computationally construct three-dimensional (3D) models of a protein complex structure from its component protein units. A protein docking program takes two or more individual 3D protein structures, which are either experimentally solved or computationally modeled, and outputs a series of probable complex structures.In this chapter we present the LZerD protein docking suite, which includes programs for pairwise docking, LZerD and PI-LZerD, and multiple protein docking, Multi-LZerD, developed by our group. PI-LZerD takes protein docking interface residues as additional input information. The methods use a combination of shape-based protein surface features as well as physics-based scoring terms to generate protein complex models. The programs are provided as stand-alone programs and can be downloaded from http://kiharalab.org/proteindocking.

Highly precise protein-protein interaction prediction based on consensus between template-based and de novo docking methods

Background

Elucidation of protein-protein interaction (PPI) networks is important for understanding disease mechanisms and for drug discovery. Tertiary-structure-based in silico PPI prediction methods have been developed with two typical approaches: a method based on template matching with known protein structures and a method based on de novo protein docking. However, the template-based method has a narrow applicable range because of its use of template information, and the de novo docking based method does not have good prediction performance. In addition, both of these in silico prediction methods have insufficient precision, and require validation of the predicted PPIs by biological experiments, leading to considerable expenditure; therefore, PPI prediction methods with greater precision are needed.

Results

We have proposed a new structure-based PPI prediction method by combining template-based prediction and de novo docking prediction. When we applied the method to the human apoptosis signaling pathway, we obtained a precision value of 0.333, which is higher than that achieved using conventional methods (0.231 for PRISM, a template-based method, and 0.145 for MEGADOCK, a non-template-based method), while maintaining an F-measure value (0.285) comparable to that obtained using conventional methods (0.296 for PRISM, and 0.220 for MEGADOCK).

Conclusions

Our consensus method successfully predicted a PPI network with greater precision than conventional template/non-template methods, which may thus reduce the cost of validation by laboratory experiments for confirming novel PPIs from predicted PPIs. Therefore, our method may serve as an aid for promoting interactome analysis.

Predicting permanent and transient protein-protein interfaces

Protein-protein interactions (PPIs) are involved in diverse functions in a cell. To optimize functional roles of interactions, proteins interact with a spectrum of binding affinities. Interactions are conventionally classified into permanent and transient, where the former denotes tight binding between proteins that result in strong complexes, whereas the latter compose of relatively weak interactions that can dissociate after binding to regulate functional activity at specific time point. Knowing the type of interactions has significant implications for understanding the nature and function of PPIs. In this study, we constructed amino acid substitution models that capture mutation patterns at permanent and transient type of protein interfaces, which were found to be different with statistical significance. Using the substitution models, we developed a novel computational method that predicts permanent and transient protein binding interfaces (PBIs) in protein surfaces. Without knowledge of the interacting partner, the method uses a single query protein structure and a multiple sequence alignment of the sequence family. Using a large dataset of permanent and transient proteins, we show that our method, BindML+, performs very well in protein interface classification. A very high area under the curve (AUC) value of 0.957 was observed when predicted protein binding sites were classified. Remarkably, near prefect accuracy was achieved with an AUC of 0.991 when actual binding sites were classified. The developed method will be also useful for protein design of permanent and transient PBIs.

PI2PE: A Suite of Web Servers for Predictions Ranging From Protein Structure to Binding Kinetics

PI²PE (http://pipe.sc.fsu.edu) is a suite of four web servers for predicting a variety of folding- and binding-related properties of proteins. These include the solvent accessibility of amino acids upon protein folding, the amino acids forming the interfaces of protein-protein and protein-nucleic acid complexes, and the binding rate constants of these complexes. Three of the servers debuted in 2007, and have garnered ~2,500 unique users and finished over 30,000 jobs. The functionalities of these servers are now enhanced, and a new sever, for predicting the binding rate constants, is added. Together, these web servers form a pipeline from protein sequence to tertiary structure, then to quaternary structure, and finally to binding kinetics.