Protein-protein Complexes Research Articles

A comprehensive survey of scoring functions for protein docking models

BackgroundWhile protein-protein docking is fundamental to our understanding of how proteins interact, scoring protein-protein complex conformations is a critical component of successful docking programs. Without accurate and efficient scoring functions to differentiate between native and non-native binding complexes, the accuracy of current docking tools cannot be guaranteed. Although many innovative scoring functions have been proposed, a good scoring function for docking remains elusive. Deep learning models offer alternatives to using explicit empirical or mathematical functions for scoring protein-protein complexes.ResultsIn this study, we perform a comprehensive survey of the state-of-the-art scoring functions by considering the most popular and highly performant approaches, both classical and deep learning-based, for scoring protein-protein complexes. The methods were also compared based on their runtime as it directly impacts their use in large-scale docking applications.ConclusionsWe evaluate the strengths and weaknesses of classical and deep learning-based approaches across seven public and popular datasets to aid researchers in understanding the progress made in this field.

AlphaFold and Docking Approaches for Antibody-Antigen and Other Targets: Insights From CAPRI Rounds 47-55.

Accurate modeling of the structures of protein-protein complexes and other biomolecular interactions represents a longstanding and important challenge for computational biology. The Critical Assessment of PRedicted Interactions (CAPRI) experiment has served for over two decades as a key means to assess and compare current approaches and methods through blind predictive scenarios, highlighting useful strategies, and new developments. Here we describe the performance of our laboratory's team in recent CAPRI rounds, which included submissions for 10 modeling targets. Our team utilized a range of docking and modeling approaches, including ZDOCK, Rosetta, and ZRANK, to model, refine, and score protein-protein and protein-DNA complexes. For recent targets we utilized adaptations of AlphaFold to generate models, leading to near-native models for an antibody-peptide target, and a highly accurate (but low ranked) model for an antibody-MHC complex. These results underscore the utility of AlphaFold-based protocols for predictive protein complex modeling, including for immune recognition, and highlight considerations regarding the use of AlphaFold confidence metrics in model selection.

Computational modeling of the anti-inflammatory complexes of IL37.

Interleukin (IL) 37 is an anti-inflammatory cytokine belonging to the IL1 protein family. Owing to its pivotal role in modulating immune responses, elucidating the IL37 complex structures holds substantial therapeutic promise for various autoimmune disorders and cancers. However, none of the structures of IL37 complexes have been experimentally characterized. This computational study aims to address this gap through molecular modeling and classical molecular dynamics simulations. We modeled all protein-protein complexes of IL37 using a range of methods from homology modeling to AlphaFold2 multimer predictions. Models that successfully recapitulated experimental features underwent further analysis through molecular dynamics simulations. As positive controls, binary and ternary complexes of IL18 from PDB were included for comparison. Several key findings emerged from the comparative analysis of IL37 and IL18 complexes. IL37 complexes exhibited higher mobility than the IL18 complexes. Simulations of the IL37-IL18Rα complex revealed altered receptor conformations capable of accommodating a dimeric IL37, with the N-terminal loop of IL37 contributing significantly to complex mobility. Additionally, the glycosyl chain on N297 of IL18Rα, which contours one edge of the cytokine binding surface, acted as a steric block against the N-terminal loop of IL37. Further, investigations into interactions between IL37 and IL18BP suggested that a binding mode homologous to IL18 was unstable for IL37, indicating an alternative binding mechanism. Altogether, this study accesses to the structure and dynamics of IL37 complexes, revealing the structural underpinnings of the IL37's modulatory effect on the IL18 signaling pathway.

Activation and Reactivity of the Deubiquitinylase OTU Cezanne-2 from MD Simulations and QM/MM Calculations.

Cezanne-2 (Cez2) is a deubiquitinylating (DUB) enzyme involved in the regulation of ubiquitin-driven cellular signaling and selectively targets Lys11-linked polyubiquitin chains. As a representative member of the ovarian tumor (OTU) subfamily DUBs, it performs cysteine proteolytic isopeptide bond cleavage; however, its exact catalytic mechanism is not yet resolved. In this work, we used different computational approaches to get molecular insights into the Cezanne-2 catalytic mechanism. Extensive molecular dynamics (MD) simulations were performed for 12 μs to model free Cez2 and the diubiquitin (diUb) substrate-bound protein-protein complex in two different charge states of Cez2, each corresponding to a distinct reactive state in its catalytic cycle. The simulations were analyzed in terms of the relevant structural parameters for productive enzymatic catalysis. Reactive diUb-Cez2 complex configurations were identified, which lead to isopeptide bond cleavage and stabilization of the tetrahedral oxyanion intermediate. The reliability of these complexes was further assessed by quantum mechanics/molecular mechanics (QM/MM) optimizations. The results show that Cez2 follows a modified cysteine protease mechanism involving a catalytic Cys210/His367 dyad, with the oxyanion hole to be a part of the "C-loop," and polarization of His367 by the formation of a strictly conserved water bridge with Glu173. The third residue has a dual role in catalysis as it mediates substrate binding and polarization of the catalytic dyad. A similar mechanism was identified for Cezanne-1, the paralogue of Cez2. In general, our simulations provide valuable molecular information that may help in the rational design of selective inhibitors of Cez2 and closely related enzymes.

Bluues_cplx: Electrostatics at Protein-Protein and Protein-Ligand Interfaces.

(1) Background: Electrostatics plays a capital role in protein-protein and protein-ligand interactions. Implicit solvent models are widely used to describe electrostatics and complementarity at interfaces. Electrostatic complementarity at the interface is not trivial, involving surface potentials rather than the charges of surfacial contacting atoms. (2) Results: The program bluues_cplx, here used in conjunction with the software NanoShaper to compute molecular surfaces, has been used to compute the electrostatic properties of 756 protein-protein and 189 protein-ligand complexes along with the corresponding isolated molecules. (3) Methods: The software we make available here uses Generalized Born (GB) radii, computed by a molecular surface integral, to output several descriptors of electrostatics at protein (and in general, molecular) interfaces. We illustrate the usage of the software by analyzing a dataset of protein-protein and protein-ligand complexes, thus extending and refining previous analyses of electrostatic complementarity at protein interfaces. (4) Conclusions: The complete analysis of a molecular complex is performed in tens of seconds on a PC, and the results include the list of surfacial contacting atoms, their charges and Pearson correlation coefficient, the list of contacting surface points with the electrostatic potential (computed for the isolated molecules) and Pearson correlation coefficient, the electrostatic and hydrophobic free energy with different contributions for the isolated molecules, their complex and the difference for all terms. The software is readily usable for any molecular complex in solution.

A potential therapeutic strategy by fungal laccase targeting novel binding sites on human cytomegalovirus DNA polymerase.

Human cytomegalovirus (HCMV) is a common herpesvirus that can severely affect transplant recipients, those with AIDS, and newborns. Existing synthetic medications face limitations, including toxicity, processing issues, and viral resistance. As part of this study, the efficacy of the extracellular enzyme laccase isolated from a widely available mushroom (Pleurotus pulmonarius) was compared to that of ganciclovir, a common antiviral, used against HCMV. The study found that laccase can synergistically inhibit HCMV replication by targeting new inhibitory sites on the UL54 protein. Viral replication requires significant energy, increasing cellular respiration. The antiviral effect of laccase was linked to reduced expression of genes regulating cellular respiration, which coincided with decreased viral DNA copies. Additionally, in silico analysis has identified a novel binding site for the laccase enzyme in the C-terminal region of HCMV DNA polymerase specifically between amino acids 1004 and 1242, which effectively obstructs the binding of the essential viral replication regulatory accessory protein UL44, thereby hindering successful replication. Molecular dynamics simulations were performed under standardized conditions mimicking a cellular environment, revealing a stable protein-protein docking complex. This study may aid in developing novel antiviral strategies by utilizing laccase's target specificity to regulate host cellular pathways against Herpesviridae.

Modes of Binding of Small Molecules Dictate the Interruption of RBD-ACE2 Complex of SARS-CoV-2.

The spike protein is a vital target for therapeutic advancement to inhibit viral entrance. Given that the connection between Spike and ACE2 constitutes the initial phase of SARS-CoV-2 pathogenesis, obstructing this interaction presents a promising therapeutic approach. This work aims to find compounds from DrugBank that can modulate the stability of the spike RBD-ACE2 protein-protein complex. Employing a therapeutic repurposing strategy, we conducted molecular docking of over 9000 DrugBank compounds against the Spike RBD-ACE2 complex, on ten variants, including the wild-type. We also evaluated the intricate stability of the RBD-ACE2 proteins by molecular dynamics simulations, hydrogen bond analysis, RMSD analysis, radius of gyration analysis, and the QM-MM approach. We assessed the efficacy of the top ten candidates for each variant as an inhibitor. Our findings demonstrated for the first time that DrugBank small molecules can interact in three distinct modalities inside the extensive protein-protein interface of RBD and ACE2 complexes. The top ten analyses identified specific DrugBank candidates for each variant and molecules capable of binding to multiple variants. This comprehensive computational technique enables the screening and forecasting of hits for any big and shallow protein-protein interface drug targets.

PPB-Affinity: Protein-Protein Binding Affinity dataset for AI-based protein drug discovery

Prediction of protein-protein binding (PPB) affinity plays an important role in large-molecular drug discovery. Deep learning (DL) has been adopted to predict the changes of PPB binding affinities upon mutations, but there was a scarcity of studies predicting the PPB affinity itself. The major reason is the paucity of open-source dataset with PPB affinity data. To address this gap, the current study introduced a large comprehensive PPB affinity (PPB-Affinity) dataset. The PPB-Affinity dataset contains key information such as crystal structures of protein-protein complexes (with or without protein mutation patterns), PPB affinity, receptor protein chain, ligand protein chain, etc. To the best of our knowledge, this is the largest publicly available PPB affinity dataset, and we believe it will significantly advance drug discovery by streamlining the screening of potential large-molecule drugs. We also developed a deep-learning benchmark model with this dataset to predict the PPB affinity, providing a foundational comparison for the research community.

Computational insight into crucial interaction between Pcf11 and Ydh1 for pre-mRNA 3ˈ-end processing

Pre-mRNA processing in eukaryotes involves capping, splicing, cleavage, and polyadenylation. Various proteins regulating this key transcriptional event in humans share considerable homology with Saccharomyces cerevisiae proteins. Among these proteins, Pcf11 is a crucial component of the yeast CF IA sub-unit, and Ydh1 is part of the CPF sub-unit. Both these proteins have a significant role during the pre-mRNA processing of the nascent transcription. Our in silico analysis highlights probable interaction between residues of Pcf11-Ydh1 and their role in mRNA processing events. These outcomes provide evidence for direct interaction between the domain from residues 116 to 204 of Pcf11 (Pcf11116-204) with the N-terminal region of Ydh1 (residues 1-246; Ydh11-246). Molecular docking and MD simulations shed light on the structure and dynamics of the protein-protein complex that includes binding affinity and binding interface of Pcf11 and Ydh1 interaction. These outcomes would pave the way to further design in vitro and in vivo studies to determine the function of the Pcf11116-204 domain, which has not been previously analyzed; this would also facilitate deciphering the crucial role of Ydh1 and Pcf11 in assembling the cleavage and polyadenylation complex for executing co-transcriptional processing to generate mature mRNA.

Physicochemical features of subunit interfaces and their role in self-assembly across the ferritin superfamily.

Ferritins are ubiquitous and play a critical role in iron homeostasis. They are classified into four main subfamilies: classical, bacterial, bacterioferritin, and Dps. These are characterized by subunits with a four-helical bundle domain and interact through three distinct regions-one antiparallel interface (IntA) and two perpendicular interfaces (IntB and IntC), collectively forming a cage-like structure. Here, we attempt to characterize the variability of these interfaces across subfamilies. We found that IntA is essential for the dimeric unit assembly and is likely to assemble first, followed by the smaller interfaces of IntB and IntC (in any order), which are crucial for cage formation. These interfaces are unique in that they are less packed, although chemically stable, and their size lies between that of protein-protein complex and obligate homodimers. This study provides a detailed exploration of the ferritin interfaces, offering insights into their assembly and their importance as carrier proteins.

Characterization of a widespread sugar phosphate-processing bacterial microcompartment

Many prokaryotes form Bacterial Microcompartments (BMCs) that encapsulate segments of specialized metabolic pathways to enhance catalysis. The various functions of metabolosomes, catabolic BMCs, are dictated by the signature enzyme that processes initial substrates of the confined pathway. The components and native functions of several metabolosomes have been experimentally characterized; however one of the most prevalent across all bacteria has yet to be studied. Sugar Phosphate Utilizing (SPU) BMC loci encode enzymes predicted to be involved in sugar phosphate metabolism. The SPU genetic loci are found in organisms occupying habitats ranging from soils to hot springs, highlighting the ubiquity of the SPU BMC. We bioinformatically characterized seven SPU subtypes, all which contain an enzyme unique to SPU BMCs, a deoxyribose 5-phosphate aldolase (DERA). Here, we define the fundamental characteristics of SPU BMCs and have expressed, purified, and characterized a set of SPU core enzymes. These include a protein-protein complex formed between a SPU BMC DERA and a predicted ribose 5-phosphate isomerase. Further, we show that the SPU BMC DERA is catalytically active and propose that it acts as the universal signature enzyme for the SPU BMC, with implications for fundamental understanding and biotechnological applications of SPU BMCs.

Engineering the Mechanical Stability of a Therapeutic Complex between Affibody and Programmed Death-Ligand 1 by Anchor Point Selection.

Protein-protein complexes can vary in mechanical stability depending on the direction from which force is applied. Here, we investigated the mechanical stability of a complex between a binding scaffold called Affibody and an immune checkpoint protein Programmed Death-Ligand 1 (PD-L1). We used AFM single-molecule force spectroscopy with bioorthogonal clickable peptide handles, shear stress bead adhesion assays, molecular modeling, and steered molecular dynamics (SMD) to understand the pulling point dependency of the mechanostability of the Affibody:(PD-L1) complex. We observed a wide range of rupture forces depending on the anchor point. Pulling from residue #22 on Affibody generated an intermediate state attributed to partially unfolded PD-L1, while pulling from Affibody's N-terminus generated a force-activated catch bond. Pulling from residue #22 or #47 on Affibody generated high rupture forces, with the complex breaking at up to ∼190 pN under loading rates of ∼104-105 pN/s, representing a ∼4-fold increase as compared with low-force N-terminal pulling. SMD simulations showed relative tendencies in rupture forces that were consistent with experiments and, through visualization of force propagation networks, provided mechanistic insights. These results demonstrate how the mechanical properties of protein-protein interfaces can be controlled by informed choice of site-specific bioconjugation points within molecules, with implications for optimal bioconjugation strategies in drug delivery vehicles.

Integration of molecular coarse-grained model into geometric representation learning framework for protein-protein complex property prediction

Structure-based machine learning algorithms have been utilized to predict the properties of protein-protein interaction (PPI) complexes, such as binding affinity, which is critical for understanding biological mechanisms and disease treatments. While most existing algorithms represent PPI complex graph structures at the atom-scale or residue-scale, these representations can be computationally expensive or may not sufficiently integrate finer chemical-plausible interaction details for improving predictions. Here, we introduce MCGLPPI, a geometric representation learning framework that combines graph neural networks (GNNs) with MARTINI molecular coarse-grained (CG) models to predict PPI overall properties accurately and efficiently. Extensive experiments on three types of downstream PPI property prediction tasks demonstrate that at the CG-scale, MCGLPPI achieves competitive performance compared with the counterparts at the atom- and residue-scale, but with only a third of computational resource consumption. Furthermore, CG-scale pre-training on protein domain-domain interaction structures enhances its predictive capabilities for PPI tasks. MCGLPPI offers an effective and efficient solution for PPI overall property predictions, serving as a promising tool for the large-scale analysis of biomolecular interactions.

Chemical Tools for Probing the Ub/Ubl Conjugation Cascades.

Conjugation of ubiquitin (Ub) and structurally related ubiquitin-like proteins (Ubls), essential for many cellular processes, employs multi-step reactions orchestrated by specific E1, E2 and E3 enzymes. The E1 enzyme activates the Ub/Ubl C-terminus in an ATP-dependent process that results in the formation of a thioester linkage with the E1 active site cysteine. The thioester-activated Ub/Ubl is transferred to the active site of an E2 enzyme which then interacts with an E3 enzyme to promote conjugation to the target substrate. The E1-E2-E3 enzymatic cascades utilize labile intermediates, extensive conformational changes, and vast combinatorial diversity of short-lived protein-protein complexes to conjugate Ub/Ubl to various substrates in a regulated manner. In this review, we discuss various chemical tools and methods used to study the consecutive steps of Ub/Ubl activation and conjugation, which are often too elusive for direct studies. We focus on methods developed to probe enzymatic activities and capture and characterize stable mimics of the transient intermediates and transition states, thereby providing insights into fundamental mechanisms in the Ub/Ubl conjugation pathways.

Mechanisms of Copper Selectivity and Release by the Metallochaperone CusF: Insights from CO-Binding, Rapid-Freeze-Quench EXAFS, and Unnatural Amino Acid Substitution.

Metallochaperones are small proteins that shuttle essential metal ions such as Cu selectively to their cellular targets. CusF has unusual Cu(I) coordination, bound by two methionines, one histidine and a capping tryptophan residue, W44. Here we compare the CO binding reactivity of the wild type (WT) protein and its W44A, F, and M variants. Fourier-transform infrared (FTIR) indicates that W44A provides unhindered access for CO, while W44M is unreactive. WT is also largely unreactive to CO suggesting that the tryptophan cap is effective in shielding the Cu(I) center from exogenous adduct formation, while the Phe variant shows partial reactivity suggestive of an equilibrium between cap-on and cap-off conformers. Rates of metal transfer to the partner CusB are consistent with the π-cation cap providing both selectivity and redox protection. Unnatural amino acid substitutions of the W44 ligand with cyano-Phe and Br-Phe underpin the conclusion that the Phe ligand is a less effective capping residue. Finally, density functional theory (DFT) calculations validate the CO-binding strategy. Overall, the study suggests that CusF uses the tryptophan cap to protect against exogenous ligand (O2) attack while the mechanism of protein-protein complex formation allows the cap to swing out of the way, and thus have minimal effect on the rates of metal transfer.

Assessment of Binding Affinity in the Complexes of CoV-S-Protein’s RBD and the ACE2 Using Convolutional Neural Networks

The experimentally obtained structures of 48 complexes of the ACE2 receptor with the S protein’s RBD of the coronaviruses SARS-CoV and SARS-CoV-2 (including mutant forms of the latter) were assessed and the dissociation constant was calculated for them. Prediction of binding affinity was carried out using ProBAN, a neural network algorithm, previously developed by the authors, and a number of other algorithms for estimating the Gibbs free energy such as Prodigy, FoldX, DFIRE and RosettaDock. A comparison of the evaluation results shows that ProBAN has the best prediction quality (Pearson correlation − 0.56, MAE − 0.7 kcal/mol) of all the analyzed algorithms. The results obtained suggest better quality of affinity prediction for other protein-protein complexes. Information about the complexes under study and prediction results are available in the repository at the link: https://github.com/EABogdanova/ProBAN_RBD-ACE2.

Plastocyanin and Cytochrome f Complex Structures Obtained by NMR, Molecular Dynamics, and AlphaFold 3 Methods Compared to Cryo-EM Data.

Plastocyanin is a small mobile protein that facilitates electron transfer through the formation of short-lived protein-protein complexes with cytochrome bf and photosystem 1. Due to the transient nature of plastocyanin-cytochrome f complex, the lack of a long-lived tight complex makes it impossible to determine its structure by X-ray diffraction analysis. Up to today, a number of slightly different structures of such complexes have been obtained by experimental and computer methods. Now, artificial intelligence gives us the possibility to predict the structures of intermolecular complexes. In this study, we compare encounter and final complexes obtained by Brownian and molecular dynamics methods, as well as the structures predicted by AlphaFold 3, with NMR and cryo-EM data. Surprisingly, the best match for the plastocyanin electron density obtained by cryo-EM was demonstrated by an AlphaFold 3 structure. The orientation of plastocyanin in this structure almost completely coincides with its orientation obtained by molecular dynamics calculation, and, at the same time, it is different from the orientation of plastocyanin predicted on the basis of NMR data. This is even more unexpected given that only NMR structures for the plastocyanin-cytochrome f complex are available in the PDB database, which was used to train AlphaFold 3.

Proteome-wide copy-number estimation from transcriptomics

Protein copy numbers constrain systems-level properties of regulatory networks, but proportional proteomic data remain scarce compared to RNA-seq. We related mRNA to protein statistically using best-available data from quantitative proteomics and transcriptomics for 4366 genes in 369 cell lines. The approach starts with a protein’s median copy number and hierarchically appends mRNA–protein and mRNA–mRNA dependencies to define an optimal gene-specific model linking mRNAs to protein. For dozens of cell lines and primary samples, these protein inferences from mRNA outmatch stringent null models, a count-based protein-abundance repository, empirical mRNA-to-protein ratios, and a proteogenomic DREAM challenge winner. The optimal mRNA-to-protein relationships capture biological processes along with hundreds of known protein-protein complexes, suggesting mechanistic relationships. We use the method to identify a viral-receptor abundance threshold for coxsackievirus B3 susceptibility from 1489 systems-biology infection models parameterized by protein inference. When applied to 796 RNA-seq profiles of breast cancer, inferred copy-number estimates collectively re-classify 26–29% of luminal tumors. By adopting a gene-centered perspective of mRNA–protein covariation across different biological contexts, we achieve accuracies comparable to the technical reproducibility of contemporary proteomics.

MPA-MutPred: a novel strategy for accurately predicting the binding affinity change upon mutation in membrane protein complexes.

Mutations in the interface of membrane protein (MP) complexes are key contributors to a broad spectrum of human diseases, primarily due to changes in their binding affinities. While various methods exist for predicting the mutation-induced changes in binding affinity (ΔΔG) in protein-protein complexes, none are specific to MP complexes. This study proposes a novel strategy for ΔΔG prediction in MP complexes, which combines linear and nonlinear models, to obtain a more robust model with improved prediction accuracy. We used multiple linear regression to extract informative features that influence the binding affinity in MP complexes, which included changes in the stability of the complex, conservation score, electrostatic interaction, relatively accessible surface area, and interface contacts. Further, using gradient boosting regressor on the selected features, we developed MPA-MutPred, a novel method specific for predicting the ΔΔG of membrane protein-protein complexes, and it is freely accessible at https://web.iitm.ac.in/bioinfo2/MPA-MutPred/. Our method achieved a correlation of 0.75 and a mean absolute error (MAE) of 0.73kcal/mol in the jack-knife test conducted on a dataset of 770 mutants. We further validated the method using a blind test set of 86 mutations, obtaining a correlation of 0.85 and an MAE of 0.77kcal/mol. We anticipate that this method can be used for large-scale studies to understand the influence of binding affinity change on disease-causing mutations in MP complexes, thereby aiding in the understanding of disease mechanisms and the identification of potential therapeutic targets.

Computational screening of the effects of mutations on protein-protein off-rates and dissociation mechanisms by τRAMD

The dissociation rate, or its reciprocal, the residence time (τ), is a crucial parameter for understanding the duration and biological impact of biomolecular interactions. Accurate prediction of τ is essential for understanding protein-protein interactions (PPIs) and identifying potential drug targets or modulators for tackling diseases. Conventional molecular dynamics simulation techniques are inherently constrained by their limited timescales, making it challenging to estimate residence times, which typically range from minutes to hours. Building upon its successful application in protein-small molecule systems, τ-Random Acceleration Molecular Dynamics (τRAMD) is here investigated for estimating dissociation rates of protein-protein complexes. τRAMD enables the observation of unbinding events on the nanosecond timescale, facilitating rapid and efficient computation of relative residence times. We tested this methodology for three protein-protein complexes and their extensive mutant datasets, achieving good agreement between computed and experimental data. By combining τRAMD with MD-IFP (Interaction Fingerprint) analysis, dissociation mechanisms were characterized and their sensitivity to mutations investigated, enabling the identification of molecular hotspots for selective modulation of dissociation kinetics. In conclusion, our findings underscore the versatility of τRAMD as a simple and computationally efficient approach for computing relative protein-protein dissociation rates and investigating dissociation mechanisms, thereby aiding the design of PPI modulators.