Characterization of protein–ligand interactions by SABRE

Nuclear spin hyperpolarization through signal amplification by reversible exchange (SABRE), the non-hydrogenative version of para-hydrogen induced polarization, is demonstrated to enhance sensitivity for the detection of biomacromolecular interactions. A target ligand for the enzyme trypsin includes the binding motif for the protein, and at a distant location a heterocyclic nitrogen atom for interacting with a SABRE polarization transfer catalyst. This molecule, 4-amidinopyridine, is hyperpolarized with 50% para-hydrogen to yield enhancement values ranging from −87 and −34 in the ortho and meta positions of the heterocyclic nitrogen, to −230 and −110, for different solution conditions. Ligand binding is identified by flow-NMR, in a two-step process that separately optimizes the polarization transfer in methanol while detecting the interaction in a predominantly aqueous medium. A single scan Carr–Purcell–Meiboom–Gill (CPMG) experiment identifies binding by the change in R₂ relaxation rate. The SABRE hyperpolarization technique provides a cost effective means to enhance NMR of biological systems, for the identification of protein–ligand interactions and other applications.

Protein–ligand binding interactions are characterized by the para-H₂ based hyperpolarization technique SABRE and flow-NMR. Binding to the protein is identified by R₂ change of a ligand first interacting with the Ir polarization transfer catalyst.     

GanDTI: A multi-task neural network for drug-target interaction prediction

Drug discovery processes require drug-target interaction (DTI) prediction by virtual screenings with high accuracy. Compared with traditional methods, the deep learning method requires less time and domain expertise, while achieving higher accuracy. However, there is still room for improvement for higher performance with simplified structures. Meanwhile, this field is calling for multi-task models to solve different tasks. Here we report the GanDTI, an end-to-end deep learning model for both interaction classification and binding affinity prediction tasks. This model employs the compound graph and protein sequence data. It only consists of a graph neural network, an attention module and a multiple-layer perceptron, yet outperforms the state-of-the art methods to predict binding affinity and interaction classification on the DUD-E, human, and bindingDB benchmark datasets. This demonstrates our refined model is highly effective and efficient for DTI prediction and provides a new strategy for performance improvement.     

Peptoid-based reprogrammable template for cell-permeable inhibitors of protein–protein interactions

The development of inhibitors of intracellular protein–protein interactions (PPIs) is of great significance for drug discovery, but the generation of a cell-permeable molecule with high affinity to protein is challenging. Oligo(N-substituted glycines) (oligo-NSGs), referred to as peptoids, are attractive as potential intracellular PPI inhibitors owing to their high membrane permeability. However, their intrinsically flexible backbones make the rational design of inhibitors difficult. Here, we propose a peptoid-based rational approach to develop cell-permeable PPI inhibitors using oligo(N-substituted alanines) (oligo-NSAs). The rigid structures of oligo-NSAs enable independent optimization of each N-substituent to improve binding affinity and membrane permeability, while preserving the backbone shape. A molecule with optimized N-substituents inhibited a target PPI in cells, which demonstrated the utility of oligo-NSA as a reprogrammable template to develop intracellular PPI inhibitors.

A peptoid-based modular approach using oligo(N-substituted alanine) as a reprogrammable template enables independent optimization of N-substituents and facile development of cell-permeable inhibitors of protein–protein interactions.     

A proximity labeling method for protein–protein interactions on cell membrane

Antibodies targeting specific antigens are widely utilized in biological research to investigate protein interactions or to quantify target antigens. Here, we introduce antigen–antibody proximity labeling (AAPL), a novel method to map the antigen interaction sites as well as interactors of antibody-targeted proteins. As a proof of concept, AAPL was demonstrated using sodium/potassium transporting ATPase (ATP1A1) and epidermal growth factor receptor 2 (ERBB2)-specific antibodies that were modified with an Fe(iii) catalytic probe. Once bound to their target proteins, Fe(iii)-induced catalytic oxidation occurred in proximity of the antigen''s epitope. Oxidative proteomic analysis was then used to determine the degree of oxidation, the site of oxidation within the targeted antigen, and the interacting proteins that were in close proximity to the targeted antigen. An AAPL score was generated for each protein yielding the specificity of the oxidation and proximity of the interacting protein to the target antigen. As a final demonstration of its utility, the AAPL approach was applied to map the interactors of liver–intestine-cadherin (CDH17) in colon cancer cells.

Modified catalytic antibodies targeting specific antigens are employed to investigate protein interactions and antigen interaction sites.     

Peptide-based inhibitors of protein–protein interactions: biophysical,structural and cellular consequences of introducing a constraint

Protein–protein interactions (PPIs) are implicated in the majority of cellular processes by enabling and regulating the function of individual proteins. Thus, PPIs represent high-value, but challenging targets for therapeutic intervention. The development of constrained peptides represents an emerging strategy to generate peptide-based PPI inhibitors, typically mediated by α-helices. The approach can confer significant benefits including enhanced affinity, stability and cellular penetration and is ingrained in the premise that pre-organization simultaneously pays the entropic cost of binding, prevents a peptide from adopting a protease compliant β-strand conformation and shields the hydrophilic amides from the hydrophobic membrane. This conceptual blueprint for the empirical design of peptide-based PPI inhibitors is an exciting and potentially lucrative way to effect successful PPI inhibitor drug-discovery. However, a plethora of more subtle effects may arise from the introduction of a constraint that include changes to binding dynamics, the mode of recognition and molecular properties. In this review, we summarise the influence of inserting constraints on biophysical, conformational, structural and cellular behaviour across a range of constraining chemistries and targets, to highlight the tremendous success that has been achieved with constrained peptides alongside emerging design opportunities and challenges.

This review summarizes the influence of inserting constraints on biophysical, conformational, structural and cellular behaviour for peptides targeting α-helix mediated protein–protein interactions.     

Query-guided protein–protein interaction inhibitor discovery

Protein–protein interactions (PPIs) are central to biological mechanisms, and can serve as compelling targets for drug discovery. Yet, the discovery of small molecule inhibitors of PPIs remains challenging given the large and typically shallow topography of the interacting protein surfaces. Here, we describe a general approach to the discovery of orthosteric PPI inhibitors that mimic specific secondary protein structures. Initially, hot residues at protein–protein interfaces are identified in silico or from experimental data, and incorporated into secondary structure-based queries. Virtual libraries of small molecules are then shape-matched against the queries, and promising ligands docked to target proteins. The approach is exemplified experimentally using two unrelated PPIs that are mediated by an α-helix (p53/hDM2) and a β-strand (GKAP/SHANK1-PDZ). In each case, selective PPI inhibitors are discovered with low μM activity as determined by a combination of fluorescence anisotropy and ¹H–¹⁵N HSQC experiments. In addition, hit expansion yields a series of PPI inhibitors with defined structure–activity relationships. It is envisaged that the generality of the approach will enable discovery of inhibitors of a wide range of unrelated secondary structure-mediated PPIs.

Small-molecule protein–protein interaction inhibitors were prioritised on the basis of shape similarity to secondary structure-based queries incorporating hot-spot residues.     

MGraphDTA: deep multiscale graph neural network for explainable drug–target binding affinity prediction

Predicting drug–target affinity (DTA) is beneficial for accelerating drug discovery. Graph neural networks (GNNs) have been widely used in DTA prediction. However, existing shallow GNNs are insufficient to capture the global structure of compounds. Besides, the interpretability of the graph-based DTA models highly relies on the graph attention mechanism, which can not reveal the global relationship between each atom of a molecule. In this study, we proposed a deep multiscale graph neural network based on chemical intuition for DTA prediction (MGraphDTA). We introduced a dense connection into the GNN and built a super-deep GNN with 27 graph convolutional layers to capture the local and global structure of the compound simultaneously. We also developed a novel visual explanation method, gradient-weighted affinity activation mapping (Grad-AAM), to analyze a deep learning model from the chemical perspective. We evaluated our approach using seven benchmark datasets and compared the proposed method to the state-of-the-art deep learning (DL) models. MGraphDTA outperforms other DL-based approaches significantly on various datasets. Moreover, we show that Grad-AAM creates explanations that are consistent with pharmacologists, which may help us gain chemical insights directly from data beyond human perception. These advantages demonstrate that the proposed method improves the generalization and interpretation capability of DTA prediction modeling.

MGraphDTA is designed to capture the local and global structure of a compound simultaneously for drug–target affinity prediction and can provide explanations that are consistent with pharmacologists.     

Alchemical absolute protein–ligand binding free energies for drug design

The recent advances in relative protein–ligand binding free energy calculations have shown the value of alchemical methods in drug discovery. Accurately assessing absolute binding free energies, although highly desired, remains a challenging endeavour, mostly limited to small model cases. Here, we demonstrate accurate first principles based absolute binding free energy estimates for 128 pharmaceutically relevant targets. We use a novel rigorous method to generate protein–ligand ensembles for the ligand in its decoupled state. Not only do the calculations deliver accurate protein–ligand binding affinity estimates, but they also provide detailed physical insight into the structural determinants of binding. We identify subtle rotamer rearrangements between apo and holo states of a protein that are crucial for binding. When compared to relative binding free energy calculations, obtaining absolute binding free energies is considerably more challenging in large part due to the need to explicitly account for the protein in its apo state. In this work we present several approaches to obtain apo state ensembles for accurate absolute ΔG calculations, thus outlining protocols for prospective application of the methods for drug discovery.

Molecular dynamics based absolute protein–ligand binding free energies can be calculated accurately and at large scale to facilitate drug discovery.     

VoteDock: Consensus docking method for prediction of protein–ligand interactions

Molecular recognition plays a fundamental role in all biological processes, and that is why great efforts have been made to understand and predict protein–ligand interactions. Finding a molecule that can potentially bind to a target protein is particularly essential in drug discovery and still remains an expensive and time‐consuming task. In silico, tools are frequently used to screen molecular libraries to identify new lead compounds, and if protein structure is known, various protein–ligand docking programs can be used. The aim of docking procedure is to predict correct poses of ligand in the binding site of the protein as well as to score them according to the strength of interaction in a reasonable time frame. The purpose of our studies was to present the novel consensus approach to predict both protein–ligand complex structure and its corresponding binding affinity. Our method used as the input the results from seven docking programs (Surflex, LigandFit, Glide, GOLD, FlexX, eHiTS, and AutoDock) that are widely used for docking of ligands. We evaluated it on the extensive benchmark dataset of 1300 protein–ligands pairs from refined PDBbind database for which the structural and affinity data was available. We compared independently its ability of proper scoring and posing to the previously proposed methods. In most cases, our method is able to dock properly approximately 20% of pairs more than docking methods on average, and over 10% of pairs more than the best single program. The RMSD value of the predicted complex conformation versus its native one is reduced by a factor of 0.5 Å. Finally, we were able to increase the Pearson correlation of the predicted binding affinity in comparison with the experimental value up to 0.5. © 2010 Wiley Periodicals, Inc. J Comput Chem 32: 568–581, 2011     

Design and training of a neural network for predicting the solvent accessibility of proteins

A feed-forward neural network has been developed to predict the solvent accessibility/accessible surface area (ASA) of proteins using improved design and training methods. Several network issues ranging from the coding of ASA states to the problem of local minima of learning curve, have been addressed. Successful new approaches to overcome these problems are presented. Set of trained network weights for each ASA threshold is provided. It has been established that the prediction accuracy results with neural network are better than other reported results of ASA prediction, despite a high test to training data ratio.     

Predicting protein-protein interactions using graph invariants and a neural network

The PDZ domain of proteins mediates a protein-protein interaction by recognizing the hydrophobic C-terminal tail of the target protein. One of the challenges put forth by the DREAM (Discussions on Reverse Engineering Assessment and Methods) 2009 Challenge consists of predicting a position weight matrix (PWM) that describes the specificity profile of five PDZ domains to their target peptides. We consider the primary structures of each of the five PDZ domains as a numerical sequence derived from graph-theoretic models of each of the individual amino acids in the protein sequence. Using available PDZ domain databases to obtain known targets, the graph-theoretic based numerical sequences are then used to train a neural network to recognize their targets. Given the challenge sequences, the target probabilities are computed and a corresponding position weight matrix is derived. In this work we present our method. The results of our method placed second in the DREAM 2009 challenge.     

Cooperative stabilisation of 14-3-3σ protein–protein interactions via covalent protein modification

14-3-3 proteins are an important family of hub proteins that play important roles in many cellular processes via a large network of interactions with partner proteins. Many of these protein–protein interactions (PPI) are implicated in human diseases such as cancer and neurodegeneration. The stabilisation of selected 14-3-3 PPIs using drug-like ‘molecular glues’ is a novel therapeutic strategy with high potential. However, the examples reported to date have a number of drawbacks in terms of selectivity and potency. Here, we report that WR-1065, the active species of the approved drug amifostine, covalently modifies 14-3-3σ at an isoform-unique cysteine residue, Cys38. This modification leads to isoform-specific stabilisation of two 14-3-3σ PPIs in a manner that is cooperative with a well characterised molecular glue, fusicoccin A. Our findings reveal a novel stabilisation mechanism for 14-3-3σ, an isoform with particular involvement in cancer pathways. This mechanism can be exploited to harness the enhanced potency conveyed by covalent drug molecules and dual ligand cooperativity. This is demonstrated in two cancer cell lines whereby the cooperative behaviour of fusicoccin A and WR-1065 leads to enhanced efficacy for inducing cell death and attenuating cell growth.

The aminothiol WR-1065 covalently modifies 14-3-3σ to stabilse its interactions with p53 and ERα. It enhances the effect of fusicoccin A via a cooperative mechanism that leads to 14-3-3 partner-protein specific activty against cancer cells.      

Discovering protein–protein interaction stabilisers by native mass spectrometry

Protein–protein interactions (PPIs) are key therapeutic targets. Most PPI-targeting drugs in the clinic inhibit these important interactions; however, stabilising PPIs is an attractive alternative in cases where a PPI is disrupted in a disease state. The discovery of novel PPI stabilisers has been hindered due to the lack of tools available to monitor PPI stabilisation. Moreover, for PPI stabilisation to be detected, both the stoichiometry of binding and the shift this has on the binding equilibria need to be monitored simultaneously. Here, we show the power of native mass spectrometry (MS) in the rapid search for PPI stabilisers. To demonstrate its capability, we focussed on three PPIs between the eukaryotic regulatory protein 14-3-3σ and its binding partners estrogen receptor ERα, the tumour suppressor p53, and the kinase LRRK2, whose interactions upon the addition of a small molecule, fusicoccin A, are differentially stabilised. Within a single measurement the stoichiometry and binding equilibria between 14-3-3 and each of its binding partners was evident. Upon addition of the fusicoccin A stabiliser, a dramatic shift in binding equilibria was observed with the 14-3-3:ERα complex compared with the 14-3-3:p53 and 14-3-3:LRRK2 complexes. Our results highlight how native MS can not only distinguish the ability of stabilisers to modulate PPIs, but also give important insights into the dynamics of ternary complex formation. Finally, we show how native MS can be used as a screening tool to search for PPI stabilisers, highlighting its potential role as a primary screening technology in the hunt for novel therapeutic PPI stabilisers.

Stabilising protein–protein interactions is challenging, yet therapeutically important. Native mass spectrometry can be used to monitor binding equilibria, allowing identification and measurement of novel protein–protein interaction stabilisers.     

PLIMSTEX: a novel mass spectrometric method for the quantification of protein–ligand interactions in solution

Protein–ligand interactions by mass spectrometry, titration, and H/D exchange (PLIMSTEX) is a new mass spectrometric method for determining association constants and binding stoichiometry for interactions of proteins with various ligands, as well as for quantifying the conformational changes associated with ligand binding to proteins. The association constants determined with PLIMSTEX agree with literature values within a factor of six, establishing its validity for protein interactions involving metal ions, small organic molecules, peptides, and proteins. PLIMSTEX provides solution, not gas-phase, properties by taking advantage of ESI and MALDI mass spectrometry to measure accurately the mass of a protein as it undergoes amide H/D exchange. The approach sidesteps the problem of relating gas-phase abundances of the protein or protein–ligand complex ions to their solution concentrations. With on-column concentration and desalting, high picomole quantities of proteins are sufficient for reproducible mass detection, and the concentration of the protein can be as low as 10⁻⁸ M. It is amenable to different protein/ligand systems in physiologically relevant media. No specially labeled protein or ligand is needed. PLIMSTEX offers minimal perturbation of the binding equilibrium because it uses no denaturants, no additional spectroscopy or reaction probes, and no physical separation of ligand and protein during binding.     

Molecular basis for acyl carrier protein–ketoreductase interaction in trans-acyltransferase polyketide synthases

The biosynthesis of polyketides by type I modular polyketide synthases (PKS) relies on co-ordinated interactions between acyl carrier protein (ACP) domains and catalytic domains within the megasynthase. Despite the importance of these interactions, and their implications for biosynthetic engineering efforts, they remain poorly understood. Here, we report the molecular details of the interaction interface between an ACP domain and a ketoreductase (KR) domain from a trans-acyltransferase (trans-AT) PKS. Using a high-throughput mass spectrometry (MS)-based assay in combination with scanning alanine mutagenesis, residues contributing to the KR-binding epitope of the ACP domain were identified. Application of carbene footprinting revealed the ACP-binding site on the KR domain surface, and molecular docking simulations driven by experimental data allowed production of an accurate model of the complex. Interactions between ACP and KR domains from trans-AT PKSs were found to be specific for their cognate partner, indicating highly optimised interaction interfaces driven by evolutionary processes. Using detailed knowledge of the ACP:KR interaction epitope, an ACP domain was engineered to interact with a non-cognate KR domain partner. The results provide novel, high resolution insights into the ACP:KR interface and offer valuable rules for future engineering efforts of biosynthetic assembly lines.

The interaction epitope between a cognate KR–ACP domain pairing from a trans-AT polyketide synthase is elucidated in molecular detail, providing unique insights into recognition and specificity of the interface.     

Learning size-adaptive molecular substructures for explainable drug–drug interaction prediction by substructure-aware graph neural network

Drug–drug interactions (DDIs) can trigger unexpected pharmacological effects on the body, and the causal mechanisms are often unknown. Graph neural networks (GNNs) have been developed to better understand DDIs. However, identifying key substructures that contribute most to the DDI prediction is a challenge for GNNs. In this study, we presented a substructure-aware graph neural network, a message passing neural network equipped with a novel substructure attention mechanism and a substructure–substructure interaction module (SSIM) for DDI prediction (SA-DDI). Specifically, the substructure attention was designed to capture size- and shape-adaptive substructures based on the chemical intuition that the sizes and shapes are often irregular for functional groups in molecules. DDIs are fundamentally caused by chemical substructure interactions. Thus, the SSIM was used to model the substructure–substructure interactions by highlighting important substructures while de-emphasizing the minor ones for DDI prediction. We evaluated our approach in two real-world datasets and compared the proposed method with the state-of-the-art DDI prediction models. The SA-DDI surpassed other approaches on the two datasets. Moreover, the visual interpretation results showed that the SA-DDI was sensitive to the structure information of drugs and was able to detect the key substructures for DDIs. These advantages demonstrated that the proposed method improved the generalization and interpretation capability of DDI prediction modeling.

SA-DDI is designed to learn size-adaptive molecular substructures for drug–drug interaction prediction and can provide explanations that are consistent with pharmacologists.     

Two‐dimensional replica‐exchange method for predicting protein–ligand binding structures

We have developed a two‐dimensional replica‐exchange method for the prediction of protein–ligand binding structures. The first dimension is the umbrella sampling along the reaction coordinate, which is the distance between a protein binding pocket and a ligand. The second dimension is the solute tempering, in which the interaction between a ligand and a protein and water is weakened. The second dimension is introduced to make a ligand follow the umbrella potential more easily and enhance the binding events, which should improve the sampling efficiency. As test cases, we applied our method to two protein‐ligand complex systems (MDM2 and HSP 90‐alpha). Starting from the configuration in which the protein and the ligand are far away from each other in each system, our method predicted the ligand binding structures in excellent agreement with the experimental data from Protein Data Bank much faster with the improved sampling efficiency than the replica‐exchange umbrella sampling method that we have previously developed. © 2013 Wiley Periodicals, Inc.