Graph Neural Network for Protein–Protein Interaction Prediction: A Comparative Study

Proteins are the fundamental biological macromolecules which underline practically all biological activities. Protein–protein interactions (PPIs), as they are known, are how proteins interact with other proteins in their environment to perform biological functions. Understanding PPIs reveals how cells behave and operate, such as the antigen recognition and signal transduction in the immune system. In the past decades, many computational methods have been developed to predict PPIs automatically, requiring less time and resources than experimental techniques. In this paper, we present a comparative study of various graph neural networks for protein–protein interaction prediction. Five network models are analyzed and compared, including neural networks (NN), graph convolutional neural networks (GCN), graph attention networks (GAT), hyperbolic neural networks (HNN), and hyperbolic graph convolutions (HGCN). By utilizing the protein sequence information, all of these models can predict the interaction between proteins. Fourteen PPI datasets are extracted and utilized to compare the prediction performance of all these methods. The experimental results show that hyperbolic graph neural networks tend to have a better performance than the other methods on the protein-related datasets.     

Machine Learning Models to Predict Protein–Protein Interaction Inhibitors

Protein–protein interaction (PPI) inhibitors have an increasing role in drug discovery. It is hypothesized that machine learning (ML) algorithms can classify or identify PPI inhibitors. This work describes the performance of different algorithms and molecular fingerprints used in chemoinformatics to develop a classification model to identify PPI inhibitors making the codes freely available to the community, particularly the medicinal chemistry research groups working with PPI inhibitors. We found that classification algorithms have different performances according to various features employed in the training process. Random forest (RF) models with the extended connectivity fingerprint radius 2 (ECFP4) had the best classification abilities compared to those models trained with ECFP6 o MACCS keys (166-bits). In general, logistic regression (LR) models had lower performance metrics than RF models, but ECFP4 was the representation most appropriate for LR. ECFP4 also generated models with high-performance metrics with support vector machines (SVM). We also constructed ensemble models based on the top-performing models. As part of this work and to help non-computational experts, we developed a pipeline code freely available.     

Modulating Protein–Protein Interactions by Cyclic and Macrocyclic Peptides. Prominent Strategies and Examples

Cyclic and macrocyclic peptides constitute advanced molecules for modulating protein–protein interactions (PPIs). Although still peptide derivatives, they are metabolically more stable than linear counterparts, and should have a lower degree of flexibility, with more defined secondary structure conformations that can be adapted to imitate protein interfaces. In this review, we analyze recent progress on the main methods to access cyclic/macrocyclic peptide derivatives, with emphasis in a few selected examples designed to interfere within PPIs. These types of peptides can be from natural origin, or prepared by biochemical or synthetic methodologies, and their design could be aided by computational approaches. Some advances to facilitate the permeability of these quite big molecules by conjugation with cell penetrating peptides, and the incorporation of β-amino acid and peptoid structures to improve metabolic stability, are also commented. It is predicted that this field of research could have an important future mission, running in parallel to the discovery of new, relevant PPIs involved in pathological processes.     

Native Structure-Based Peptides as Potential Protein–Protein Interaction Inhibitors of SARS-CoV-2 Spike Protein and Human ACE2 Receptor

Severe acute respiratory syndrome-coronavirus 2 (SARS-CoV-2) is a positive-strand RNA virus that causes severe respiratory syndrome in humans, which is now referred to as coronavirus disease 2019 (COVID-19). Since December 2019, the new pathogen has rapidly spread globally, with over 65 million cases reported to the beginning of December 2020, including over 1.5 million deaths. Unfortunately, currently, there is no specific and effective treatment for COVID-19. As SARS-CoV-2 relies on its spike proteins (S) to bind to a host cell-surface receptor angiotensin-converting enzyme-2(ACE2), and this interaction is proved to be responsible for entering a virus into host cells, it makes an ideal target for antiviral drug development. In this work, we design three very short peptides based on the ACE2 sequence/structure fragments, which may effectively bind to the receptor-binding domain (RBD) of S protein and may, in turn, disrupt the important virus-host protein–protein interactions, blocking early steps of SARS-CoV-2 infection. Two of our peptides bind to virus protein with affinity in nanomolar range, and as very short peptides have great potential for drug development.     

An Unusual Protein–Protein Interaction through Coupled Unfolding and Binding

Aptides, a novel class of high‐affinity peptides, recognize diverse molecular targets with high affinity and specificity. The solution structure of the aptide APT specifically bound to fibronectin extradomain B (EDB), which represents an unusual protein–protein interaction that involves coupled unfolding and binding, is reported. APT binding is accompanied by unfolding of the C‐terminal β strand of EDB, thereby permitting APT to interact with the freshly exposed hydrophobic interior surfaces of EDB. The β‐hairpin scaffold of APT drives the interaction by a β‐strand displacement mechanism, such that an intramolecular β sheet is replaced by an intermolecular β sheet. The unfolding of EDB perturbs the tight domain association between EDB and FN8 of fibronectin, thus highlighting its potential use as a scaffold that switches between stretched and bent conformations.     

Visualizing an Ultra‐Weak Protein–Protein Interaction in Phosphorylation Signaling

Proteins interact with each other to fulfill their functions. The importance of weak protein–protein interactions has been increasingly recognized. However, owing to technical difficulties, ultra‐weak interactions remain to be characterized. Phosphorylation can take place via a K_D≈25 mM interaction between two bacterial enzymes. Using paramagnetic NMR spectroscopy and with the introduction of a novel Gd^III‐based probe, we determined the structure of the resulting complex to atomic resolution. The structure accounts for the mechanism of phosphoryl transfer between the two enzymes and demonstrates the physical basis for their ultra‐weak interaction. Further, molecular dynamics (MD) simulations suggest that the complex has a lifetime in the micro‐ to millisecond regimen. Hence such interaction is termed a fleeting interaction. From mathematical modeling, we propose that an ultra‐weak fleeting interaction enables rapid flux of phosphoryl signal, providing a high effective protein concentration.     

Molecular Dynamics Scoring of Protein–Peptide Models Derived from Coarse-Grained Docking

One of the major challenges in the computational prediction of protein–peptide complexes is the scoring of predicted models. Usually, it is very difficult to find the most accurate solutions out of the vast number of sometimes very different and potentially plausible predictions. In this work, we tested the protocol for Molecular Dynamics (MD)-based scoring of protein–peptide complex models obtained from coarse-grained (CG) docking simulations. In the first step of the scoring procedure, all models generated by CABS-dock were reconstructed starting from their original C-alpha trace representations to all-atom (AA) structures. The second step included geometry optimization of the reconstructed complexes followed by model scoring based on receptor–ligand interaction energy estimated from short MD simulations in explicit water. We used two well-known AA MD force fields, CHARMM and AMBER, and a CG MARTINI force field. Scoring results for 66 different protein–peptide complexes show that the proposed MD-based scoring approach can be used to identify protein–peptide models of high accuracy. The results also indicate that the scoring accuracy may be significantly affected by the quality of the reconstructed protein receptor structures.     

MolADI: A Web Server for Automatic Analysis of Protein–Small Molecule Dynamic Interactions

Protein–ligand interaction analysis is important for drug discovery and rational protein design. The existing online tools adopt only a single conformation of the complex structure for calculating and displaying the interactions, whereas both protein residues and ligand molecules are flexible to some extent. The interactions evolved with time in the trajectories are of greater interest. MolADI is a user-friendly online tool which analyzes the protein–ligand interactions in detail for either a single structure or a trajectory. Interactions can be viewed easily with both 2D graphs and 3D representations. MolADI is available as a web application.     

A Small‐Molecule Protein–Protein Interaction Inhibitor of PARP1 That Targets Its BRCT Domain

Poly(ADP‐ribose)polymerase‐1 (PARP1) is a BRCT‐containing enzyme (BRCT=BRCA1 C‐terminus) mainly involved in DNA repair and damage response and a validated target for cancer treatment. Small‐molecule inhibitors that target the PARP1 catalytic domain have been actively pursued as anticancer drugs, but are potentially problematic owing to a lack of selectivity. Compounds that are capable of disrupting protein–protein interactions of PARP1 provide an alternative by inhibiting its activities with improved selectivity profiles. Herein, by establishing a high‐throughput microplate‐based assay suitable for screening potential PPI inhibitors of the PARP1 BRCT domain, we have discovered that (±)‐gossypol, a natural product with a number of known biological activities, possesses novel PARP1 inhibitory activity both in vitro and in cancer cells and presumably acts through disruption of protein–protein interactions. As the first known cell‐permeable small‐molecule PPI inhibitor of PAPR1, we further established that (?)‐gossypol was likely the causative agent of PARP1 inhibition by promoting the formation of a 1:2 compound/PARP1 complex by reversible formation of a covalent imine linkage.     

Stabilizer‐Guided Inhibition of Protein–Protein Interactions

The discovery of novel protein–protein interaction (PPI) modulators represents one of the great molecular challenges of the modern era. PPIs can be modulated by either inhibitor or stabilizer compounds, which target different though proximal regions of the protein interface. In principle, protein–stabilizer complexes can guide the design of PPI inhibitors (and vice versa). In the present work, we combine X‐ray crystallographic data from both stabilizer and inhibitor co‐crystal complexes of the adapter protein 14‐3‐3 to characterize, down to the atomic scale, inhibitors of the 14‐3‐3/Tau PPI, a potential drug target to treat Alzheimer’s disease. The most potent compound notably inhibited the binding of phosphorylated full‐length Tau to 14‐3‐3 according to NMR spectroscopy studies. Our work sets a precedent for the rational design of PPI inhibitors guided by PPI stabilizer–protein complexes while potentially enabling access to new synthetically tractable stabilizers of 14‐3‐3 and other PPIs.     

Structure‐Based Design of Inhibitors of Protein–Protein Interactions: Mimicking Peptide Binding Epitopes

Protein–protein interactions (PPIs) are involved at all levels of cellular organization, thus making the development of PPI inhibitors extremely valuable. The identification of selective inhibitors is challenging because of the shallow and extended nature of PPI interfaces. Inhibitors can be obtained by mimicking peptide binding epitopes in their bioactive conformation. For this purpose, several strategies have been evolved to enable a projection of side chain functionalities in analogy to peptide secondary structures, thereby yielding molecules that are generally referred to as peptidomimetics. Herein, we introduce a new classification of peptidomimetics (classes A–D) that enables a clear assignment of available approaches. Based on this classification, the Review summarizes strategies that have been applied for the structure‐based design of PPI inhibitors through stabilizing or mimicking turns, β‐sheets, and helices.     

Selective and Potent Proteomimetic Inhibitors of Intracellular Protein–Protein Interactions

Inhibition of protein–protein interactions (PPIs) represents a major challenge in chemical biology and drug discovery. α‐Helix mediated PPIs may be amenable to modulation using generic chemotypes, termed “proteomimetics”, which can be assembled in a modular manner to reproduce the vectoral presentation of key side chains found on a helical motif from one partner within the PPI. In this work, it is demonstrated that by using a library of N‐alkylated aromatic oligoamide helix mimetics, potent helix mimetics which reproduce their biophysical binding selectivity in a cellular context can be identified.     

Combining Molecular Dynamic Information and an Aspherical-Atom Data Bank in the Evaluation of the Electrostatic Interaction Energy in Multimeric Protein-Ligand Complex: A Case Study for HIV-1 Protease

Computational analysis of protein–ligand interactions is of crucial importance for drug discovery. Assessment of ligand binding energy allows us to have a glimpse of the potential of a small organic molecule to be a ligand to the binding site of a protein target. Available scoring functions, such as in docking programs, all rely on equations that sum each type of protein–ligand interactions in order to predict the binding affinity. Most of the scoring functions consider electrostatic interactions involving the protein and the ligand. Electrostatic interactions constitute one of the most important part of total interactions between macromolecules. Unlike dispersion forces, they are highly directional and therefore dominate the nature of molecular packing in crystals and in biological complexes and contribute significantly to differences in inhibition strength among related enzyme inhibitors. In this study, complexes of HIV-1 protease with inhibitor molecules (JE-2147 and darunavir) were analyzed by using charge densities from the transferable aspherical-atom University at Buffalo Databank (UBDB). Moreover, we analyzed the electrostatic interaction energy for an ensemble of structures, using molecular dynamic simulations to highlight the main features of electrostatic interactions important for binding affinity.     

Probing the Small‐Molecule Inhibition of an Anticancer Therapeutic Protein‐Protein Interaction Using a Solid‐State Nanopore

Nanopore sensing is an emerging technology for the single‐molecule‐based detection of various biomolecules. In this study, we probed the anticancer therapeutic p53 transactivation domain (p53TAD)/MDM2 interaction and its inhibition with a small‐molecule MDM2 antagonist, Nutlin‐3, using low‐noise solid‐state nanopores. Although the translocation of positively charged MDM2 through a nanopore was detected at the applied negative voltage, this MDM2 translocation was almost completely blocked upon formation of the MDM2/GST‐p53TAD complex owing to charge conversion. In combination with NMR data, the nanopore measurements showed that the addition of Nutlin‐3 rescued MDM2 translocation, indicating that Nutlin‐3 disrupted the MDM2/GST‐p53TAD complex, thereby releasing MDM2. Taken together, our results reveal that solid‐state nanopores can be a valuable platform for the ultrasensitive, picomole‐scale screening of small‐molecule drugs against protein–protein interaction (PPI) targets.     

Quantifying Protein–Protein Interactions by Molecular Counting with Mass Photometry

Interactions between biomolecules control the processes of life in health and their malfunction in disease, making their characterization and quantification essential. Immobilization‐ and label‐free analytical techniques are desirable because of their simplicity and minimal invasiveness, but they struggle with quantifying tight interactions. Here, we show that mass photometry can accurately count, distinguish by molecular mass, and thereby reveal the relative abundances of different unlabelled biomolecules and their complexes in mixtures at the single‐molecule level. These measurements determine binding affinities over four orders of magnitude at equilibrium for both simple and complex stoichiometries within minutes, as well as the associated kinetics. These results introduce mass photometry as a rapid, simple and label‐free method for studying sub‐micromolar binding affinities, with potential for extension towards a universal approach for characterizing complex biomolecular interactions.     

Probing Functional Heteromeric Chemokine Protein–Protein Interactions through Conformation‐Assisted Oxime Ligation

Protein–protein interactions (PPIs) govern most processes in living cells. Current drug development strategies are aimed at disrupting or stabilizing PPIs, which require a thorough understanding of PPI mechanisms. Examples of such PPIs are heteromeric chemokine interactions that are potentially involved in pathological disorders such as cancer, atherosclerosis, and HIV. It remains unclear whether this functional modulation is mediated by heterodimer formation or by the additive effects of mixed chemokines on their respective receptors. To address this issue, we report the synthesis of a covalent RANTES‐PF4 heterodimer (termed OPRAH) by total chemical synthesis and oxime ligation, with an acceleration of the final ligation step driven by PPIs between RANTES and PF4. Compared to mixed separate chemokines, OPRAH exhibited increased biological activity, thus providing evidence that physical formation of the heterodimer indeed mediates enhanced function.     

Constrained Peptides with Target‐Adapted Cross‐Links as Inhibitors of a Pathogenic Protein–Protein Interaction

Bioactive conformations of peptides can be stabilized by macrocyclization, resulting in increased target affinity and activity. Such macrocyclic peptides proved useful as modulators of biological functions, in particular as inhibitors of protein–protein interactions (PPI). However, most peptide‐derived PPI inhibitors involve stabilized α‐helices, leaving a large number of secondary structures unaddressed. Herein, we present a rational approach towards stabilization of an irregular peptide structure, using hydrophobic cross‐links that replace residues crucially involved in target binding. The molecular basis of this interaction was elucidated by X‐ray crystallography and isothermal titration calorimetry. The resulting cross‐linked peptides inhibit the interaction between human adaptor protein 14‐3‐3 and virulence factor exoenzyme S. Taking into consideration that irregular peptide structures participate widely in PPIs, this approach provides access to novel peptide‐derived inhibitors.     

Determination of the Protein-Protein Interactions within Acyl Carrier Protein (MmcB)-Dependent Modifications in the Biosynthesis of Mitomycin

Mitomycin has a unique chemical structure and contains densely assembled functionalities with extraordinary antitumor activity. The previously proposed mitomycin C biosynthetic pathway has caused great attention to decipher the enzymatic mechanisms for assembling the pharmaceutically unprecedented chemical scaffold. Herein, we focused on the determination of acyl carrier protein (ACP)-dependent modification steps and identification of the protein–protein interactions between MmcB (ACP) with the partners in the early-stage biosynthesis of mitomycin C. Based on the initial genetic manipulation consisting of gene disruption and complementation experiments, genes mitE, mmcB, mitB, and mitF were identified as the essential functional genes in the mitomycin C biosynthesis, respectively. Further integration of biochemical analysis elucidated that MitE catalyzed CoA ligation of 3-amino-5-hydroxy-bezonic acid (AHBA), MmcB-tethered AHBA triggered the biosynthesis of mitomycin C, and both MitB and MitF were MmcB-dependent tailoring enzymes involved in the assembly of mitosane. Aiming at understanding the poorly characterized protein–protein interactions, the in vitro pull-down assay was carried out by monitoring MmcB individually with MitB and MitF. The observed results displayed the clear interactions between MmcB and MitB and MitF. The surface plasmon resonance (SPR) biosensor analysis further confirmed the protein–protein interactions of MmcB with MitB and MitF, respectively. Taken together, the current genetic and biochemical analysis will facilitate the investigations of the unusual enzymatic mechanisms for the structurally unique compound assembly and inspire attempts to modify the chemical scaffold of mitomycin family antibiotics.