Protein-peptide Binding Affinity Research Articles

A deep attention model for wide-genome protein-peptide binding affinity prediction at a sequence level

Peptides are pivotal in numerous biological activities by engaging in up to 40 % of protein-protein interactions in many cellular processes. Due to their exceptional specificity and effectiveness, peptides have emerged as promising candidates for drug design. However, accurately predicting protein-peptide binding affinity remains a challenging. Aiming at the problem, we develop a prediction model PepPAP based on convolutional neural network and multi-head attention, which relies solely on sequence features. These features include physicochemical properties, intrinsic disorder, sequence encoding, and especially interface propensity which is extracted from 16,689 non-redundant protein-peptide complexes. Notably, the adopted regression stratification cross-validation scheme proposed in our previous work is beneficial to improve the prediction for the cases with extreme binding affinity values. On three benchmark test datasets: T100, a series of peptides targeting to PDZ domain and CXCR4, PepPAP shows excellent performance, outperforming the existing methods and demonstrating its good generalization ability. Furthermore, PepPAP has good results in binary interaction prediction, and the analysis of the feature space distribution visualization highlights PepPAP's effectiveness. To the best of our knowledge, PepPAP is the first sequence-based deep attention model for wide-genome protein-peptide binding affinity prediction, and holds the potential to offer valuable insights for the peptide-based drug design.

TDC-2: Multimodal Foundation for Therapeutic Science.

Therapeutics Data Commons (tdcommons.ai) is an open science initiative with unified datasets, AI models, and benchmarks to support research across therapeutic modalities and drug discovery and development stages. The Commons 2.0 (TDC-2) is a comprehensive overhaul of Therapeutic Data Commons to catalyze research in multimodal models for drug discovery by unifying single-cell biology of diseases, biochemistry of molecules, and effects of drugs through multimodal datasets, AI-powered API endpoints, new multimodal tasks and model frameworks, and comprehensive benchmarks. TDC-2 introduces over 1,000 multimodal datasets spanning approximately 85 million cells, pre-calculated embeddings from 5 state-of-the-art single-cell models, and a biomedical knowledge graph. TDC-2 drastically expands the coverage of ML tasks across therapeutic pipelines and 10+ new modalities, spanning but not limited to single-cell gene expression data, clinical trial data, peptide sequence data, peptidomimetics protein-peptide interaction data regarding newly discovered ligands derived from AS-MS spectroscopy, novel 3D structural data for proteins, and cell-type-specific protein-protein interaction networks at single-cell resolution. TDC-2 introduces multimodal data access under an API-first design using the model-view-controller paradigm. TDC-2 introduces 7 novel ML tasks with fine-grained biological contexts: contextualized drug-target identification, single-cell chemical/genetic perturbation response prediction, protein-peptide binding affinity prediction task, and clinical trial outcome prediction task, which introduce antigen-processing-pathway-specific, cell-type-specific, peptide-specific, and patient-specific biological contexts. TDC-2 also releases benchmarks evaluating 15+ state-of-the-art models across 5+ new learning tasks evaluating models on diverse biological contexts and sampling approaches. Among these, TDC-2 provides the first benchmark for context-specific learning. TDC-2, to our knowledge, is also the first to introduce a protein-peptide binding interaction benchmark.

MM-GBSA and QM/MM simulation-based in silico approaches for the inhibition of Acinetobacter baumannii class D OXA-24 β-lactamase using antimicrobial peptides melittin and RP-1

The antimicrobial peptides (AMPs) have been regarded as the next generation antibiotics. This study aimed to explore the AMP inhibitors of β-lactamase enzyme employing computational biology methods. Protein-peptide docking of Acinetobacter baumannii OXA-24 class D β-lactamase with AMPs (melittin and RP-1) were performed using HawkDock webserver. The docked complexes were subjected to energy property analysis through MM/GBSA, and binding affinity (ΔG (kcal/mol)) and stability (dissociation constant, KD (M)) prediction using PRODIGY. Both the AMPs, melittin and RP-1, were well docked with A. baumannii OXA-24. The top ranked OXA-24-melittin and OXA-24-RP-1 complexes were detected on the basis of the HawkDock scores (-2974.08 and -2825.83, respectively), thereafter by rescoring with MM/GBSA-based binding free energy (BFE) of -33.85 and -29.29 kcal/mol, respectively. The PRODIGY-based respective BFE (-8.0 and -10.0 kcal/mol) and KD (1.4 × 10-6 and 5 × 10-8 M) of the complexes revealed excellent protein-peptide binding affinity and complex stability. The iMODS-based molecular dynamic simulation authenticated the stability and molecular motion flexibility of the protein-peptide complexes. The quantum mechanics and molecular mechanics energy estimated using density-functional theory with CP2K software, for the electrostatic interaction of OXA-24-RP-1 (-218862.95 kcal/mol), was more favourable than the H-bonded interaction of OXA-24-melittin (-200620.21 kcal/mol), nevertheless both the peptides were found effective to inhibit OXA-24. Both the AMPs had toxicity profiles within the acceptable limits as predicted through pkCSM. Current findings indicate the potential supplement or replacement of conventionally used antibiotics with melittin and RP-1 to treating or averting A. baumannii infection by escaping the β-lactamases-mediated resistance.

Predicting Binding Affinity Between MHC-I Receptor and Peptides Based on Molecular Docking and Protein-peptide Interaction Interface Characteristics

Background:Predicting protein-peptide binding affinity is one of the leading research subjects in peptide drug design and repositioning. In previous studies, models constructed by researchers just used features of peptide structures. These features had limited information and could not describe the proteinpeptide interaction mode. This made models and predicted results lack interpretability in pharmacy and biology, which led to the protein-peptide interaction mode not being reflected. Therefore, it was of little significance for the design of peptide drugs.Objective:Considering the protein-peptide interaction mode, we extracted protein-peptide interaction interface characteristics and built machine learning models to improve the performance and enhance the interpretability of models.Methods:Taking MHC-I protein and its binding peptides as the research object, protein-peptide complexes were obtained by molecular docking, and 94 protein-peptide interaction interface characteristics were calculated. Then ten important features were selected using recursive feature elimination to construct SVR, RF, and MLP models to predict protein-peptide binding affinity.Results:The MAE of the SVR, RF and MLP models constructed using protein-peptide interaction interface characteristics are 0.2279, 0.2939 and 0.2041, their MSE are 0.1289, 0.1308 and 0.0780, and their R2 reached 0.8711, 0.8692 and 0.9220, respectively.Conclusion:The model constructed using protein-peptide interaction interface characteristics showed better prediction results. The key features for predicting protein-peptide binding affinity are the bSASA of negatively charged species, hydrogen bond acceptor, hydrophobic group, planarity, and aromatic ring.

Multiplex measurement of protein-peptide dissociation constants using dialysis and mass spectrometry.

We propose a high-throughput method for quantitatively measuring hundreds of protein-peptide binding affinities in parallel. In this assay, a solution of protein is dialyzed into a buffer containing a pool of potential binding peptides, such that upon equilibration the relative abundance of a peptide species is mathematically related to that peptide's dissociation constant, Kd . We use isobaric multiplexed quantitative proteomics to simultaneously determine the relative abundance, and hence the Kd and its associated error, for an entire peptide library. We apply this technique, which we call PEDAL (parallel equilibrium dialysis for affinity learning), to determine accurate Kd 's between a PDZ domain and hundreds of peptides, spanning an affinity range of multiple orders of magnitude in a single experiment. PEDAL is a convenient, fast, and low-cost method for measuring large numbers of protein-peptide affinities in parallel, providing a rare combination of true in-solution binding equilibria with the ability to multiplex.

PPI-Affinity: A Web Tool for the Prediction and Optimization of Protein-Peptide and Protein-Protein Binding Affinity.

Virtual screening of protein–protein and protein–peptide interactions is a challenging task that directly impacts the processes of hit identification and hit-to-lead optimization in drug design projects involving peptide-based pharmaceuticals. Although several screening tools designed to predict the binding affinity of protein–protein complexes have been proposed, methods specifically developed to predict protein–peptide binding affinity are comparatively scarce. Frequently, predictors trained to score the affinity of small molecules are used for peptides indistinctively, despite the larger complexity and heterogeneity of interactions rendered by peptide binders. To address this issue, we introduce PPI-Affinity, a tool that leverages support vector machine (SVM) predictors of binding affinity to screen datasets of protein–protein and protein–peptide complexes, as well as to generate and rank mutants of a given structure. The performance of the SVM models was assessed on four benchmark datasets, which include protein–protein and protein–peptide binding affinity data. In addition, we evaluated our model on a set of mutants of EPI-X4, an endogenous peptide inhibitor of the chemokine receptor CXCR4, and on complexes of the serine proteases HTRA1 and HTRA3 with peptides. PPI-Affinity is freely accessible at https://protdcal.zmb.uni-due.de/PPIAffinity.

Self-derived peptides from the SARS-CoV-2 spike glycoprotein disrupting shaping and stability of the homotrimer unit

The structural spike (S) protein from the SARS-CoV-2 β-coronavirus is shown to make different pre- and post-fusion conformations within its homotrimer unit. To support the ongoing novel vaccine design and development strategies, we report the structure-based design approach to develop self-derived S peptides. A dataset of crucial regions from the S protein were transformed into linear motifs that could act as the blockers or stabilizers for the S protein homotrimer unit. Among these distinct S peptides, the pep02 (537-QQFGRDIAD-545) and pep07 (821-RDLICAQKFNGLTVLPPLLTDE-842) were found making stable folded binding with the S protein (550–750 and 950–1050 regions). Upon inserting SARS-CoV-2 S variants in the peptide destabilized the complexed S protein structure, resulting an allosteric effect in different functional regions of the protein. Particularly, the molecular dynamics revealed that A544D mutation in the pep02 peptide induced instability for the complexed S protein, whereas the N943K variant from pep09 exhibited an opposite behavior. An increased protein-peptide binding affinity and the stable structural folding were observed in mutated systems, compared to that of the wild type systems. The presence of mutation has induced an “up” active conformation of the spike (RBD) domain, responsible for interacting the host cell receptor. Among the lower affinity peptide datasets (e.g., pep01), the S1 and S2 subunit in the protein formed an “open” conformation, whereas with higher affinity peptides (e.g., pep07) these domains gained a “closed” conformation. These findings propose that our designed self-derived S peptides could replace a single S protein monomer, blocking the homotrimer formation or inducing stability.

Machine Learning in Quantitative Protein-peptide Affinity Prediction: Implications for Therapeutic Peptide Design.

Protein-peptide recognition plays an essential role in the orchestration and regulation of cell signaling networks, which is estimated to be responsible for up to 40% of biological interaction events in the human interactome and has recently been recognized as a new and attractive druggable target for drug development and disease intervention. We present a systematic review on the application of machine learning techniques in the quantitative modeling and prediction of protein-peptide binding affinity, particularly focusing on its implications for therapeutic peptide design. We also briefly introduce the physical quantities used to characterize protein-peptide affinity and attempt to extend the content of generalized machine learning methods. Existing issues and future perspective on the statistical modeling and regression prediction of protein- peptide binding affinity are discussed. There is still a long way to go before establishment of general, reliable and efficient machine leaningbased protein-peptide affinity predictors.

Peptide design by optimization on a data-parameterized protein interaction landscape

Many applications in protein engineering require optimizing multiple protein properties simultaneously, such as binding one target but not others or binding a target while maintaining stability. Such multistate design problems require navigating a high-dimensional space to find proteins with desired characteristics. A model that relates protein sequence to functional attributes can guide design to solutions that would be hard to discover via screening. In this work, we measured thousands of protein-peptide binding affinities with the high-throughput interaction assay amped SORTCERY and used the data to parameterize a model of the alpha-helical peptide-binding landscape for three members of the Bcl-2 family of proteins: Bcl-xL, Mcl-1, and Bfl-1. We applied optimization protocols to explore extremes in this landscape to discover peptides with desired interaction profiles. Computational design generated 36 peptides, all of which bound with high affinity and specificity to just one of Bcl-xL, Mcl-1, or Bfl-1, as intended. We designed additional peptides that bound selectively to two out of three of these proteins. The designed peptides were dissimilar to known Bcl-2-binding peptides, and high-resolution crystal structures confirmed that they engaged their targets as expected. Excellent results on this challenging problem demonstrate the power of a landscape modeling approach, and the designed peptides have potential uses as diagnostic tools or cancer therapeutics.

Single Amino Acid Substitution in the Vicinity of a Receptor-Binding Domain Changes Protein-Peptide Binding Affinity.

Using a microarray-based assay, we studied how the substitution of amino acids in the immediate vicinity of the receptor-binding domain on a peptide affects its binding to a protein. Replicates of 802 linear peptides consisting of the variants of WTHPQFAT and LQWHPQAGK, GKFPIPLGKQSG, and NGQFQVWIPGAQK, different by one amino acid, were synthesized on a glass slide with a maskless photolithography. Using a microarray-compatible label-free optical sensor, we measured the binding curves of streptavidin with the synthesized peptides and extracted the streptavidin–peptide affinity constants. We found that (a) the substitution of one residue in the HPQ motif reduces the affinity constant Ka from 108 M–1 by at least 3–4 orders of magnitude, with an exception of HPM; (b) substitution of the immediate flanking residue on the Gln side also causes the affinity to decrease by up to 3–4 orders of magnitude, depending on the substituting residue and the second-neighboring flanking residue; (c) substitution of the flanking residues on the His side has no significant effect on the affinity, possibly due to the strong binding of streptavidin to HPQF and HPQAG motifs. We also found that some of amino acid residues located close to the C-terminus (and the solid surface) improve the yield of peptide synthesis on a glass surface and can be exploited in the fabrication of peptide microarrays.

Development of QSAR-Improved Statistical Potential for the Structure-Based Analysis of ProteinPeptide Binding Affinities.

Proteinpeptide interactions have recently been found to play an essential role in constructing intracellular signaling networks. Understanding the molecular mechanism of such interactions and identification of the interacting partners would be of great value for developing peptide therapeutics against many severe diseases such as cancer. In this study, we describe a structure-based, general-purpose strategy for fast and reliably predicting proteinpeptide binding affinities. This strategy combines unsupervised knowledge-based statistical potential derived from 505 interfacially diverse, non-redundant proteinpeptide complex structures and supervised quantitative structure-activity relationship (QSAR) modeling trained by 250 proteinpeptide interactions with known structure and affinity data. The built partial least squares (PLS) model is confirmed to have high stability and predictive power by using internal 5-fold cross-validation and rigorous Monte Carlo cross-validation (MCCV). The model is further employed to analyze two large groups of HLA- and SH3-binding peptides based upon computationally modeled structures. Satisfactorily, although the PLS model is originally trained with dissociation constants (Kd ) of proteinpeptide binding, it shows a good correlation with other two affinity qualities, i.e. SPOT signal intensities (BLU) and half maximal competitive concentrations (IC50 ). Furthermore, we perform systematic comparisons of our method with several widely used, representative affinity predictors, including molecular mechanics-based MM-PB/SA, knowledge-based DFIRE and docking score HADDOCK, on a small panel of elaborately selected proteinpeptide systems. It is demonstrated that (i) the QSAR-improved statistical potential exhibits a comparable predictive performance with but can work faster than these traditional methods, and (ii) the crystal structure-derived statistical potential also supports the modeled and solution structures of proteinpeptide complexes. We expect that this hybrid method can be exploited as a new scoring tool to facilitate, for example, peptide docking and virtual screening.

Advances in the Prediction of Protein–Peptide Binding Affinities: Implications for Peptide‐Based Drug Discovery

Peptides hold great promise as novel medicinal and biologic agents, and computational methods can help unlock that promise. In particular, structure-based peptide design can be used to identify and optimize peptide ligands. Successful structure-based design, in turn, requires accurate and fast methods for predicting protein-peptide binding affinities. Here, we review the development of such methods, emphasizing structure-based methods that assume rigid-body association and the single-structure approximation. We also briefly review recent applications of computational free energy prediction methods to enable and guide novel peptide drug and biomarker discovery. We close the review with a brief perspective on the future of computational, structure-based protein-peptide binding affinity prediction.

Use of fast conformational sampling to improve the characterization of VEGF A–peptide interactions

Protein–peptide interaction is fundamentally important for signal transduction, transcription regulation, protein degradation, cell regeneration, and immune response. Here, we report the use of a fast conformational sampling strategy to improve the prediction of protein–peptide binding affinity. This method generates hundreds of alternative conformers for a protein–peptide complex and then performs classical MM-PB/SA analysis over these conformers to derive a consistent binding energy expression for the complex. We show a proof-of-concept study on vascular endothelial growth factor A (VEGF A) interaction with its peptide ligands. The structures of VEGF A complexed with 13 peptides are modeled with a virtual mutagenesis protocol and their binding energies are subsequently calculated by using the conformational sampling-based method. A good linear correlation between the calculated and experimental values is observed, and we demonstrate that the correlation could be further improved by fitting the decomposed energy terms to experimentally measured affinity. Furthermore, the obtained results are discussed in detail in order to elucidate the structural basis and energetic implication underlying VEGF A–peptide recognition and association. We also give a detailed comparison between the proposed method and other widely used approaches, from which it is suggested that our method exhibits a good compromise between the effectiveness and efficiency in evaluating protein–peptide affinity.

AlGaN/GaN high electron mobility transistors for protein–peptide binding affinity study

Antibody-immobilized AlGaN/GaN high electron mobility transistors (HEMTs) were used to detect a short peptide consisting of 20 amino acids. One-binding-site model and two-binding-site model were used for the analysis of the electrical signals, revealing the number of binding sites on an antibody and the dissociation constants between the antibody and the short peptide. In the binding-site models, the surface coverage ratio of the short peptide on the sensor surface is relevant to the electrical signals resulted from the peptide–antibody binding on the HEMTs. Two binding sites on an antibody were observed and two dissociation constants, 4.404×10−11M and 1.596×10−9M, were extracted from the binding-site model through the analysis of the surface coverage ratio of the short peptide on the sensor surface. We have also shown that the conventional method to extract the dissociation constant from the linear regression of curve-fitting with Langmuir isotherm equation may lead to an incorrect information if the receptor has more than one binding site for the ligand. The limit of detection (LOD) of the sensor observed in the experimental result (∼10pM of the short peptide) is very close to the LOD (around 2.7–3.4pM) predicted from the value of the smallest dissociation constants. The sensitivity of the sensor is not only dependent on the transistors, but also highly relies on the affinity of the ligand-receptor pair. The results demonstrate that the AlGaN/GaN HEMTs cannot only be used for biosensors, but also for the biological affinity study.

Continued development of an empirical function for predicting and rationalizing protein–protein binding affinities

Here we summarize recent work on the continued development of our fast and simple empirical equation for predicting and structurally rationalizing protein–protein and protein–peptide binding affinities. Our empirical expression consists of six regression-weighted physical descriptors and derives from two key simplifying assumptions: (1) the assumption of rigid-body association and (2) the assumption that all contributions not explicitly considered in the equation make a net contribution to binding of ≈ 0 kcal. Within the strict framework of rigid-body association, we tested relative binding affinity predictions using our empirical equation against the corresponding experimental binding free energy data for 197 interface alanine mutants. Our methodology produced excellent agreement between prediction and experiment for 79% of the mutations considered. These encouraging results further suggest the basic validity of our approach. Further analysis suggests that the majority of the failed predictions can be accounted for in terms of mutation induced violations of assumption (2). In particular, we hypothesize that assumed away charge and aromatic side chain-mediated electrostatic interface interactions play a key role in protein–protein recognition and that such interactions must be explicitly considered for a more generally valid approach to physics-based binding affinity prediction.

Side-chain conformational space analysis (SCSA): A multi conformation-based QSAR approach for modeling and prediction of protein–peptide binding affinities

In this article, the concept of multi conformation-based quantitative structure-activity relationship (MCB-QSAR) is proposed, and based upon that, we describe a new approach called the side-chain conformational space analysis (SCSA) to model and predict protein-peptide binding affinities. In SCSA, multi-conformations (rather than traditional single-conformation) have received much attention, and the statistical average information on multi-conformations of side chains is determined using self-consistent mean field theory based upon side chain rotamer library. Thereby, enthalpy contributions (including electrostatic, steric, hydrophobic interaction and hydrogen bond) and conformational entropy effects to the binding are investigated in terms of occurrence probability of residue rotamers. Then, SCSA was applied into the dataset of 419 HLA-A 0201 binding peptides, and nonbonding contributions of each position in peptide ligands are well determined. For the peptides, the hydrogen bond and electrostatic interactions of the two ends are essential to the binding specificity, van der Waals and hydrophobic interactions of all the positions ensure strong binding affinity, and the loss of conformational entropy at anchor positions partially counteracts other favorable nonbonding effects.

QSAR method for prediction of protein-peptide binding affinity: Application to MHC class I molecule HLA-A*0201

The support vector machine (SVM), which is a novel algorithm from the machine learning community, was used to develop quantitative structure–activity relationship (QSAR) models for predicting the binding affinity of 152 nonapeptides, which can bind to class I MHC HLA-A*201 molecule. Each peptide was represented by a large pool of descriptors including constitutional, topological descriptors and physical–chemical properties. The heuristic method (HM) was then used to search the descriptor space for selecting the proper ones responsible for binding affinity. The four descriptors were obtained to build linear models based on HM and nonlinear models based on SVM method. The best results are found using SVM: root mean-square (RMS) errors for training, test and whole data set were 0.383, 0.385 and 0.384, respectively. This paper allow the prediction of the binding affinity of new, untested peptides and, through the analysis of contribution of each parameter of different residue at specific position of peptidic ligands, to understand nature of the forces governing binding behavior and suggest new ideas for further synthesis of high-affinity peptides.