i-Patch: interprotein contact prediction using local network information

Integrating unsupervised language model with multi-view multiple sequence alignments for high-accuracy inter-chain contact prediction

Accurate identification of inter-chain contacts in the protein complex is critical to determine the corresponding 3D structures and understand the biological functions. We proposed a new deep learning method, ICCPred, to deduce the inter-chain contacts from the amino acid sequences of the protein complex. This pipeline was built on the designed deep residual network architecture, integrating the pre-trained language model with three multiple sequence alignments (MSAs) from different biological views. Experimental results on 709 non-redundant benchmarking protein complexes showed that the proposed ICCPred significantly increased inter-chain contact prediction accuracy compared to the state-of-the-art approaches. Detailed data analyses showed that the significant advantage of ICCPred lies in the utilization of pre-trained transformer language models which can effectively extract the complementary co-evolution diversity from three MSAs. Meanwhile, the designed deep residual network enhances the correlation between the co-evolution diversity and the patterns of inter-chain contacts. These results demonstrated a new avenue for high-accuracy deep-learning inter-chain contact prediction that is applicable to large-scale protein-protein interaction annotations from sequence alone.

In silico Approaches for the Design and Optimization of Interfering Peptides Against Protein-Protein Interactions

Large contact surfaces of protein-protein interactions (PPIs) remain to be an ongoing issue in the discovery and design of small molecule modulators. Peptides are intrinsically capable of exploring larger surfaces, stable, and bioavailable, and therefore bear a high therapeutic value in the treatment of various diseases, including cancer, infectious diseases, and neurodegenerative diseases. Given these promising properties, a long way has been covered in the field of targeting PPIs via peptide design strategies. In silico tools have recently become an inevitable approach for the design and optimization of these interfering peptides. Various algorithms have been developed to scrutinize the PPI interfaces. Moreover, different databases and software tools have been created to predict the peptide structures and their interactions with target protein complexes. High-throughput screening of large peptide libraries against PPIs; "hotspot" identification; structure-based and off-structure approaches of peptide design; 3D peptide modeling; peptide optimization strategies like cyclization; and peptide binding energy evaluation are among the capabilities of in silico tools. In the present study, the most recent advances in the field of in silico approaches for the design of interfering peptides against PPIs will be reviewed. The future perspective of the field and its advantages and limitations will also be pinpointed.

A compact vocabulary of paratope-epitope interactions enables predictability of antibody-antigen binding

Antibody-antigen binding relies on the specific interaction of amino acids at the paratope-epitope interface. The predictability of antibody-antigen binding is a prerequisite for de novo antibody and (neo-)epitope design. A fundamental premise for the predictability of antibody-antigen binding is the existence of paratope-epitope interaction motifs that are universally shared among antibody-antigen structures. In a dataset of non-redundant antibody-antigen structures, we identify structural interaction motifs, which together compose a commonly shared structure-based vocabulary of paratope-epitope interactions. We show that this vocabulary enables the machine learnability of antibody-antigen binding on the paratope-epitope level using generative machine learning. The vocabulary (1) is compact, less than 10⁴ motifs; (2) distinct from non-immune protein-protein interactions; and (3) mediates specific oligo- and polyreactive interactions between paratope-epitope pairs. Our work leverages combined structure- and sequence-based learning to demonstrate that machine-learning-driven predictive paratope and epitope engineering is feasible.

Simplified geometric representations of protein structures identify complementary interaction interfaces

Protein-protein interactions are critical to protein function, but three-dimensional (3D) arrangements of interacting proteins have proven hard to predict, even given the identities and 3D structures of the interacting partners. Specifically, identifying the relevant pairwise interaction surfaces remains difficult, often relying on shape complementarity with molecular docking while accounting for molecular motions to optimize rigid 3D translations and rotations. However, such approaches can be computationally expensive, and faster, less accurate approximations may prove useful for large-scale prediction and assembly of 3D structures of multi-protein complexes. We asked if a reduced representation of protein geometry retains enough information about molecular properties to predict pairwise protein interaction interfaces that are tolerant of limited structural rearrangements. Here, we describe a reduced representation of 3D protein accessible surfaces on which molecular properties such as charge, hydrophobicity, and evolutionary rate can be easily mapped, implemented in the MorphProt package. Pairs of surfaces are compared to rapidly assess partner-specific potential surface complementarity. On two available benchmarks of 185 overall known protein complexes, we observe predictions comparable to other structure-based tools at correctly identifying protein interaction surfaces. Furthermore, we examined the effect of molecular motion through normal mode simulation on a benchmark receptor-ligand pair and observed no marked loss of predictive accuracy for distortions of up to 6 Å Cα-RMSD. Thus, a shape reduction of protein surfaces retains considerable information about surface complementarity, offers enhanced speed of comparison relative to more complex geometric representations, and exhibits tolerance to conformational changes.

Decoding Protein-protein Interactions: An Overview

Drug discovery has focused on the paradigm "one drug, one target" for a long time. However, small molecules can act at multiple macromolecular targets, which serves as the basis for drug repurposing. In an effort to expand the target space, and given advances in X-ray crystallography, protein-protein interactions have become an emerging focus area of drug discovery enterprises. Proteins interact with other biomolecules and it is this intricate network of interactions that determines the behavior of the system and its biological processes. In this review, we briefly discuss networks in disease, followed by computational methods for protein-protein complex prediction. Computational methodologies and techniques employed towards objectives such as protein-protein docking, protein-protein interactions, and interface predictions are described extensively. Docking aims at producing a complex between proteins, while interface predictions identify a subset of residues on one protein that could interact with a partner, and protein-protein interaction sites address whether two proteins interact. In addition, approaches to predict hot spots and binding sites are presented along with a representative example of our internal project on the chemokine CXC receptor 3 B-isoform and predictive modeling with IP10 and PF4.

Using B cell receptor lineage structures to predict affinity

We are frequently faced with a large collection of antibodies, and want to select those with highest affinity for their cognate antigen. When developing a first-line therapeutic for a novel pathogen, for instance, we might look for such antibodies in patients that have recovered. There exist effective experimental methods of accomplishing this, such as cell sorting and baiting; however they are time consuming and expensive. Next generation sequencing of B cell receptor (BCR) repertoires offers an additional source of sequences that could be tapped if we had a reliable method of selecting those coding for the best antibodies. In this paper we introduce a method that uses evolutionary information from the family of related sequences that share a naive ancestor to predict the affinity of each resulting antibody for its antigen. When combined with information on the identity of the antigen, this method should provide a source of effective new antibodies. We also introduce a method for a related task: given an antibody of interest and its inferred ancestral lineage, which branches in the tree are likely to harbor key affinity-increasing mutations? We evaluate the performance of these methods on a wide variety of simulated samples, as well as two real data samples. These methods are implemented as part of continuing development of the partis BCR inference package, available at https://github.com/psathyrella/partis. Comments Please post comments or questions on this paper as new issues at https://git.io/Jvxkn.

Pattern to Knowledge: Deep Knowledge-Directed Machine Learning for Residue-Residue Interaction Prediction

Residue-residue close contact (R2R-C) data procured from three-dimensional protein-protein interaction (PPI) experiments is currently used for predicting residue-residue interaction (R2R-I) in PPI. However, due to complex physiochemical environments, R2R-I incidences, facilitated by multiple factors, are usually entangled in the source environment and masked in the acquired data. Here we present a novel method, P2K (Pattern to Knowledge), to disentangle R2R-I patterns and render much succinct discriminative information expressed in different specific R2R-I statistical/functional spaces. Since such knowledge is not visible in the data acquired, we refer to it as deep knowledge. Leveraging the deep knowledge discovered to construct machine learning models for sequence-based R2R-I prediction, without trial-and-error combination of the features over external knowledge of sequences, our R2R-I predictor was validated for its effectiveness under stringent leave-one-complex-out-alone cross-validation in a benchmark dataset, and was surprisingly demonstrated to perform better than an existing sequence-based R2R-I predictor by 28% (p: 1.9E-08). P2K is accessible via our web server on https://p2k.uwaterloo.ca .

Exploring the potential of 3D Zernike descriptors and SVM for protein-protein interface prediction

Background

The correct determination of protein–protein interaction interfaces is important for understanding disease mechanisms and for rational drug design. To date, several computational methods for the prediction of protein interfaces have been developed, but the interface prediction problem is still not fully understood. Experimental evidence suggests that the location of binding sites is imprinted in the protein structure, but there are major differences among the interfaces of the various protein types: the characterising properties can vary a lot depending on the interaction type and function. The selection of an optimal set of features characterising the protein interface and the development of an effective method to represent and capture the complex protein recognition patterns are of paramount importance for this task.

Results

In this work we investigate the potential of a novel local surface descriptor based on 3D Zernike moments for the interface prediction task. Descriptors invariant to roto-translations are extracted from circular patches of the protein surface enriched with physico-chemical properties from the HQI8 amino acid index set, and are used as samples for a binary classification problem. Support Vector Machines are used as a classifier to distinguish interface local surface patches from non-interface ones. The proposed method was validated on 16 classes of proteins extracted from the Protein–Protein Docking Benchmark 5.0 and compared to other state-of-the-art protein interface predictors (SPPIDER, PrISE and NPS-HomPPI).

Conclusions

The 3D Zernike descriptors are able to capture the similarity among patterns of physico-chemical and biochemical properties mapped on the protein surface arising from the various spatial arrangements of the underlying residues, and their usage can be easily extended to other sets of amino acid properties. The results suggest that the choice of a proper set of features characterising the protein interface is crucial for the interface prediction task, and that optimality strongly depends on the class of proteins whose interface we want to characterise. We postulate that different protein classes should be treated separately and that it is necessary to identify an optimal set of features for each protein class.

Electronic supplementary material

The online version of this article (10.1186/s12859-018-2043-3) contains supplementary material, which is available to authorized users.

Progress and challenges in predicting protein interfaces

The majority of biological processes are mediated via protein-protein interactions. Determination of residues participating in such interactions improves our understanding of molecular mechanisms and facilitates the development of therapeutics. Experimental approaches to identifying interacting residues, such as mutagenesis, are costly and time-consuming and thus, computational methods for this purpose could streamline conventional pipelines. Here we review the field of computational protein interface prediction. We make a distinction between methods which address proteins in general and those targeted at antibodies, owing to the radically different binding mechanism of antibodies. We organize the multitude of currently available methods hierarchically based on required input and prediction principles to provide an overview of the field.

T-Cell Receptor Binding Affects the Dynamics of the Peptide/MHC-I Complex

The recognition of peptide/MHC by T-cell receptors is one of the most important interactions in the adaptive immune system. A large number of computational studies have investigated the structural dynamics of this interaction. However, to date only limited attention has been paid to differences between the dynamics of peptide/MHC with the T-cell receptor bound and unbound. Here we present the first large-scale molecular dynamics simulation study of this type investigating HLA-B*08:01 in complex with the Epstein-Barr virus peptide FLRGRAYGL and all possible single-point mutations (n = 172). All of the simulations were performed with and without the LC 13 T-cell receptor for a simulation time of 100 ns, yielding 344 simulations and a total simulation time of 34 400 ns. Our study is 2 orders of magnitude larger than the average T-cell receptor/peptide/MHC molecular dynamics simulation study. This data set provides reliable insights into alterations of the peptide/MHC-I dynamics caused by the presence of the T-cell receptor. We found that simulations in the presence of T-cell receptors have more hydrogen bonds between the peptide and MHC, altered flexibility patterns in the MHC helices and the peptide, a lower MHC groove width range, and altered solvent-accessible surface areas. This indicates that without a T-cell receptor the MHC binding groove can open and close, while the presence of the T-cell receptor inhibits these breathing-like motions.

Computational prediction of protein interfaces: A review of data driven methods

Reliably pinpointing which specific amino acid residues form the interface(s) between a protein and its binding partner(s) is critical for understanding the structural and physicochemical determinants of protein recognition and binding affinity, and has wide applications in modeling and validating protein interactions predicted by high-throughput methods, in engineering proteins, and in prioritizing drug targets. Here, we review the basic concepts, principles and recent advances in computational approaches to the analysis and prediction of protein-protein interfaces. We point out caveats for objectively evaluating interface predictors, and discuss various applications of data-driven interface predictors for improving energy model-driven protein-protein docking. Finally, we stress the importance of exploiting binding partner information in reliably predicting interfaces and highlight recent advances in this emerging direction.

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.

Improving predictions of protein-protein interfaces by combining amino acid-specific classifiers based on structural and physicochemical descriptors with their weighted neighbor averages

Protein-protein interactions are involved in nearly all regulatory processes in the cell and are considered one of the most important issues in molecular biology and pharmaceutical sciences but are still not fully understood. Structural and computational biology contributed greatly to the elucidation of the mechanism of protein interactions. In this paper, we present a collection of the physicochemical and structural characteristics that distinguish interface-forming residues (IFR) from free surface residues (FSR). We formulated a linear discriminative analysis (LDA) classifier to assess whether chosen descriptors from the BlueStar STING database (http://www.cbi.cnptia.embrapa.br/SMS/) are suitable for such a task. Receiver operating characteristic (ROC) analysis indicates that the particular physicochemical and structural descriptors used for building the linear classifier perform much better than a random classifier and in fact, successfully outperform some of the previously published procedures, whose performance indicators were recently compared by other research groups. The results presented here show that the selected set of descriptors can be utilized to predict IFRs, even when homologue proteins are missing (particularly important for orphan proteins where no homologue is available for comparative analysis/indication) or, when certain conformational changes accompany interface formation. The development of amino acid type specific classifiers is shown to increase IFR classification performance. Also, we found that the addition of an amino acid conservation attribute did not improve the classification prediction. This result indicates that the increase in predictive power associated with amino acid conservation is exhausted by adequate use of an extensive list of independent physicochemical and structural parameters that, by themselves, fully describe the nano-environment at protein-protein interfaces. The IFR classifier developed in this study is now integrated into the BlueStar STING suite of programs. Consequently, the prediction of protein-protein interfaces for all proteins available in the PDB is possible through STING_interfaces module, accessible at the following website: (http://www.cbi.cnptia.embrapa.br/SMS/predictions/index.html).

Antibody i-Patch prediction of the antibody binding site improves rigid local antibody-antigen docking

Antibodies are a class of proteins indispensable for the vertebrate immune system. The general architecture of all antibodies is very similar, but they contain a hypervariable region which allows millions of antibody variants to exist, each of which can bind to different molecules. This binding malleability means that antibodies are an increasingly important category of biopharmaceuticals and biomarkers. We present Antibody i-Patch, a method that annotates the most likely antibody residues to be in contact with the antigen. We show that our predictions correlate with energetic importance and thus we argue that they may be useful in guiding mutations in the artificial affinity maturation process. Using our predictions as constraints for a rigid-body docking algorithm, we are able to obtain high-quality results in minutes. Our annotation method and re-scoring system for docking achieve their predictive power by using antibody-specific statistics. Antibody i-Patch is available from http://www.stats.ox.ac.uk/research/proteins/resources.

Local network patterns in protein-protein interfaces

Protein-protein interfaces hold the key to understanding protein-protein interactions. In this paper we investigated local interaction network patterns beyond pair-wise contact sites by considering interfaces as contact networks among residues. A contact site was defined as any residue on the surface of one protein which was in contact with a residue on the surface of another protein. We labeled the sub-graphs of these contact networks by their amino acid types. The observed distributions of these labeled sub-graphs were compared with the corresponding background distributions and the results suggested that there were preferred chemical patterns of closely packed residues at the interface. These preferred patterns point to biological constraints on physical proximity between those residues on one protein which were involved in binding to residues which were close on the interacting partner. Interaction interfaces were far from random and contain information beyond pairs and triangles. To illustrate the possible application of the local network patterns observed, we introduced a signature method, called iScore, based on these local patterns to assess interface predictions. On our data sets iScore achieved 83.6% specificity with 82% sensitivity.

Mutual information and variants for protein domain-domain contact prediction

BACKGROUND

Predicting protein contacts solely based on sequence information remains a challenging problem, despite the huge amount of sequence data at our disposal. Mutual Information (MI), an information theory measure, has been extensively employed and modified to identify residues within a protein (intra-protein) that are in contact. More recently MI and its variants have also been used in the prediction of contacts between proteins (inter-protein).

METHODS

Here we assess the predictive power of MI and variants for domain-domain contact prediction. We test original MI and these variants, which are called MIp, MIc and ZNMI, on 40 domain-domain test cases containing 10,753 sequences. We also propose and evaluate two new versions of MI that consider triangles of residues and the physiochemical properties of the amino acids, respectively.

RESULTS

We found that all versions of MI are skewed towards predicting surface residues. Since domain-domain contacts are on the surface of each domain, we considered only surface residues when attempting to predict contacts. Our analysis shows that MIc is the best current MI domain-domain contact predictor. At 20% recall MIc achieved a precision of 44.9% when only surface residues were considered. Our triangle and reduced alphabet variants of MI highlight the delicate trade-off between signal and noise in the use of MI for domain-domain contact prediction. We also examine a specific "successful" case study and demonstrate that here, when considering surface residues, even the most accurate domain-domain contact predictor, MIc, performs no better than random.

CONCLUSIONS

All tested variants of MI are skewed towards predicting surface residues. When considering surface residues only, we find MIc to be the best current MI domain-domain contact predictor. Its performance, however, is not as good as a non-MI based contact predictor, i-Patch. Additionally, the intra-protein contact prediction capabilities of MIc outperform its domain-domain contact prediction abilities.

Partner-aware prediction of interacting residues in protein-protein complexes from sequence data

Computational prediction of residues that participate in protein-protein interactions is a difficult task, and state of the art methods have shown only limited success in this arena. One possible problem with these methods is that they try to predict interacting residues without incorporating information about the partner protein, although it is unclear how much partner information could enhance prediction performance. To address this issue, the two following comparisons are of crucial significance: (a) comparison between the predictability of inter-protein residue pairs, i.e., predicting exactly which residue pairs interact with each other given two protein sequences; this can be achieved by either combining conventional single-protein predictions or making predictions using a new model trained directly on the residue pairs, and the performance of these two approaches may be compared: (b) comparison between the predictability of the interacting residues in a single protein (irrespective of the partner residue or protein) from conventional methods and predictions converted from the pair-wise trained model. Using these two streams of training and validation procedures and employing similar two-stage neural networks, we showed that the models trained on pair-wise contacts outperformed the partner-unaware models in predicting both interacting pairs and interacting single-protein residues. Prediction performance decreased with the size of the conformational change upon complex formation; this trend is similar to docking, even though no structural information was used in our prediction. An example application that predicts two partner-specific interfaces of a protein was shown to be effective, highlighting the potential of the proposed approach. Finally, a preliminary attempt was made to score docking decoy poses using prediction of interacting residue pairs; this analysis produced an encouraging result.