Predicting Protein Binding Research Articles

The state-of-the-art machine learning model for plasma protein binding prediction: Computational modeling with OCHEM and experimental validation

Plasma protein binding (PPB) is closely related to pharmacokinetics, pharmacodynamics and drug toxicity. Existing models for predicting PPB often suffer from low prediction accuracy and poor interpretability, especially for high PPB compounds, and are most often not experimentally validated. Here, we carried out a strict data curation protocol, and applied consensus modeling to obtain a model with a coefficient of determination of 0.90 and 0.91 on the training set and the test set, respectively. This model (available on the OCHEM platform https://ochem.eu/article/29) was further retrospectively validated for a set of 63 poly-fluorinated molecules and prospectively validated for a set of 25 highly diverse compounds, and its performance for both these sets was superior to that of the other previously reported models. Furthermore, we identified the physicochemical and structural characteristics of high and low PPB molecules for further structural optimization. Finally, we provide practical and detailed recommendations for structural optimization to decrease PPB binding of lead compounds.

Binding Promiscuity of Therapeutic Factor VIII.

The binding promiscuity of proteins defines their ability to indiscriminately bind multiple unrelated molecules. Binding promiscuity is implicated, at least in part, in the off-target reactivity, nonspecific biodistribution, immunogenicity, and/or short half-life of potentially efficacious protein drugs, thus affecting their clinical use. In this review, we discuss the current evidence for the binding promiscuity of factor VIII (FVIII), a protein used for the treatment of hemophilia A, which displays poor pharmacokinetics, and elevated immunogenicity. We summarize the different canonical and noncanonical interactions that FVIII may establish in the circulation and that could be responsible for its therapeutic liabilities. We also provide information suggesting that the FVIII light chain, and especially its C1 and C2 domains, could play an important role in the binding promiscuity. We believe that the knowledge accumulated over years of FVIII usage could be exploited for the development of strategies to predict protein binding promiscuity and therefore anticipate drug efficacy and toxicity. This would open a mutational space to reduce the binding promiscuity of emerging protein drugs while conserving their therapeutic potency.

LVPocket: integrated 3D global-local information to protein binding pockets prediction with transfer learning of protein structure classification

BackgroundPrevious deep learning methods for predicting protein binding pockets mainly employed 3D convolution, yet an abundance of convolution operations may lead the model to excessively prioritize local information, thus overlooking global information. Moreover, it is essential for us to account for the influence of diverse protein folding structural classes. Because proteins classified differently structurally exhibit varying biological functions, whereas those within the same structural class share similar functional attributes.ResultsWe proposed LVPocket, a novel method that synergistically captures both local and global information of protein structure through the integration of Transformer encoders, which help the model achieve better performance in binding pockets prediction. And then we tailored prediction models for data of four distinct structural classes of proteins using the transfer learning. The four fine-tuned models were trained on the baseline LVPocket model which was trained on the sc-PDB dataset. LVPocket exhibits superior performance on three independent datasets compared to current state-of-the-art methods. Additionally, the fine-tuned model outperforms the baseline model in terms of performance.Scientific contributionWe present a novel model structure for predicting protein binding pockets that provides a solution for relying on extensive convolutional computation while neglecting global information about protein structures. Furthermore, we tackle the impact of different protein folding structures on binding pocket prediction tasks through the application of transfer learning methods.Graphical

Role of Optimization in RNA-Protein-Binding Prediction.

RNA-binding proteins (RBPs) play an important role in regulating biological processes, such as gene regulation. Understanding their behaviors, for example, their binding site, can be helpful in understanding RBP-related diseases. Studies have focused on predicting RNA binding by means of machine learning algorithms including deep convolutional neural network models. One of the integral parts of modeling deep learning is achieving optimal hyperparameter tuning and minimizing a loss function using optimization algorithms. In this paper, we investigate the role of optimization in the RBP classification problem using the CLIP-Seq 21 dataset. Three optimization methods are employed on the RNA-protein binding CNN prediction model; namely, grid search, random search, and Bayesian optimizer. The empirical results show an AUC of 94.42%, 93.78%, 93.23% and 92.68% on the ELAVL1C, ELAVL1B, ELAVL1A, and HNRNPC datasets, respectively, and a mean AUC of 85.30 on 24 datasets. This paper's findings provide evidence on the role of optimizers in improving the performance of RNA-protein binding prediction.

Biomimetic fusion: Platyper's dual vision for predicting protein-surface interactions.

Predicting protein binding with the material surface still remains a challenge. Here, a novel approach, platypus dual perception neural network (Platyper), was developed to describe the interactions in protein-surface systems involving bioceramics with BMPs. The resulting model integrates a graph convolutional neural network (GCN) based on interatomic potentials with a convolutional neural network (CNN) model based on images of molecular structures. This dual-vision approach, inspired by the platypus's adaptive sensory system, addresses the challenge of accurately predicting the complex binding and unbinding dynamics in steered molecular dynamics (SMD) simulations. The model's effectiveness is demonstrated through its application in predicting surface interactions in protein-ligand systems. Notably, Platyper improves computational efficiency compared to classical SMD-based methods and overcomes the limitations of GNN-based methods for large-scale atomic simulations. The incorporation of heat maps enhances model's interpretability, providing valuable insights into its predictive capabilities. Overall, Platyper represents a promising advancement in the accurate and efficient prediction of protein-surface interactions in the context of bioceramics and growth factors.

Adaptive language model training for molecular design

The vast size of chemical space necessitates computational approaches to automate and accelerate the design of molecular sequences to guide experimental efforts for drug discovery. Genetic algorithms provide a useful framework to incrementally generate molecules by applying mutations to known chemical structures. Recently, masked language models have been applied to automate the mutation process by leveraging large compound libraries to learn commonly occurring chemical sequences (i.e., using tokenization) and predict rearrangements (i.e., using mask prediction). Here, we consider how language models can be adapted to improve molecule generation for different optimization tasks. We use two different generation strategies for comparison, fixed and adaptive. The fixed strategy uses a pre-trained model to generate mutations; the adaptive strategy trains the language model on each new generation of molecules selected for target properties during optimization. Our results show that the adaptive strategy allows the language model to more closely fit the distribution of molecules in the population. Therefore, for enhanced fitness optimization, we suggest the use of the fixed strategy during an initial phase followed by the use of the adaptive strategy. We demonstrate the impact of adaptive training by searching for molecules that optimize both heuristic metrics, drug-likeness and synthesizability, as well as predicted protein binding affinity from a surrogate model. Our results show that the adaptive strategy provides a significant improvement in fitness optimization compared to the fixed pre-trained model, empowering the application of language models to molecular design tasks.

Spatial Layouts of Low-Entropy Hydration Shells Guide Protein Binding.

Protein-protein binding enables orderly biological self-organization and is therefore considered a miracle of nature. Protein‒protein binding is driven by electrostatic forces, hydrogen bonding, van der Waals force, and hydrophobic interactions. Among these physical forces, only hydrophobic interactions can be considered long-range intermolecular attractions between proteins due to the electrostatic shielding of surrounding water molecules. Low-entropy hydration shells around proteins drive hydrophobic attraction among them that essentially coordinate protein‒protein binding. Here, an innovative method is developed for identifying low-entropy regions of hydration shells of proteins by screening off pseudohydrophilic groups on protein surfaces and revealing that large low-entropy regions of the hydration shells typically cover the binding sites of individual proteins. According to an analysis of determined protein complex structures, shape matching between a large low-entropy hydration shell region of a protein and that of its partner at the binding sites is revealed as a universal law. Protein‒protein binding is thus found to be mainly guided by hydrophobic collapse between the shape-matched low-entropy hydration shells that is verified by bioinformatics analyses of hundreds of structures of protein complexes, which cover four test systems. A simple algorithm is proposed to accurately predict protein binding sites.

PeSTo: parameter-free geometric deep learning for accurate prediction of protein binding interfaces

Proteins are essential molecular building blocks of life, responsible for most biological functions as a result of their specific molecular interactions. However, predicting their binding interfaces remains a challenge. In this study, we present a geometric transformer that acts directly on atomic coordinates labeled only with element names. The resulting model—the Protein Structure Transformer, PeSTo—surpasses the current state of the art in predicting protein-protein interfaces and can also predict and differentiate between interfaces involving nucleic acids, lipids, ions, and small molecules with high confidence. Its low computational cost enables processing high volumes of structural data, such as molecular dynamics ensembles allowing for the discovery of interfaces that remain otherwise inconspicuous in static experimentally solved structures. Moreover, the growing foldome provided by de novo structural predictions can be easily analyzed, providing new opportunities to uncover unexplored biology.

Design of Synthetic Hydrogel Compositions for Noncovalent Protein Recognition.

Multifunctional hydrogels composed of segments with ionizable, hydrophilic, and hydrophobic monomers have been optimized for sensing, bioseparation, and therapeutic applications. While the "biological identity" of bound proteins from biofluids underlies device performance in each context, design rules that predict protein binding outcomes from hydrogel design parameters are lacking. Uniquely, hydrogel designs that influence protein affinity (e.g., ionizable monomers, hydrophobic moieties, conjugated ligands, cross-linking) also affect physical properties (e.g., matrix stiffness, volumetric swelling). Here, we evaluated the influence of hydrophobic comonomer steric bulk and quantity on the protein recognition characteristics of ionizable microscale hydrogels (microgels) while controlling for swelling. Using a library synthesis approach, we identified compositions that balance the practical balance between protein-microgel affinity and the loaded mass at saturation. Intermediate quantities (10-30 mol %) of hydrophobic comonomer increased the equilibrium binding of certain model proteins (lysozyme, lactoferrin) in buffer conditions that favored complementary electrostatic interactions. Solvent-accessible surface area analysis of model proteins identified arginine content as highly predictive of model proteins' binding to our library of hydrogels containing acidic and hydrophobic comonomers. Taken together, we established an empirical framework for characterizing the molecular recognition properties of multifunctional hydrogels. Our study is the first to identify solvent-accessible arginine as an important predictor for protein binding to hydrogels containing both acidic and hydrophobic subunits.

Development, testing, parameterisation, and calibration of a human PBPK model for the plasticiser, di-(2-ethylhexyl) terephthalate (DEHTP) using in silico, in vitro and human biomonitoring data.

A physiologically based pharmacokinetic model for di-(2-ethylhexyl) terephthalate (DEHTP) based on a refined model for di-(2-propylheptyl) phthalate (DPHP) was developed to interpret the metabolism and biokinetics of DEHTP following a single oral dose of 50mg to three male volunteers. In vitro and in silico methods were used to generate parameters for the model. For example, measured intrinsic hepatic clearance scaled from in vitro to in vivo and plasma unbound fraction and tissue:blood partition coefficients (PCs) were predicted algorithmically. Whereas the development and calibration of the DPHP model was based upon two data streams, blood concentrations of parent chemical and first metabolite and the urinary excretion of metabolites, the model for DEHTP was calibrated against a single data stream, the urinary excretion of metabolites. Despite the model form and structure being identical significant quantitative differences in lymphatic uptake between the models were observed. In contrast to DPHP the fraction of ingested DEHTP entering lymphatic circulation was much greater and of a similar magnitude to that entering the liver with evidence for the dual uptake mechanisms discernible in the urinary excretion data. Further, the absolute amounts absorbed by the study participants, were much higher for DEHTP relative to DPHP. The in silico algorithm for predicting protein binding performed poorly with an error of more than two orders of magnitude. The extent of plasma protein binding has important implications for the persistence of parent chemical in venous blood-inferences on the behaviour of this class of highly lipophilic chemicals, based on calculations of chemical properties, should be made with extreme caution. Attempting read across for this class of highly lipophilic chemicals should be undertaken with caution since basic adjustments to PCs and metabolism parameters would be insufficient, even when the structure of the model itself is appropriate. Therefore, validation of a model parameterized entirely with in vitro and in silico derived parameters would need to be calibrated against several human biomonitoring data streams to constitute a data rich source chemical to afford confidence for future evaluations of other similar chemicals using the read-across approach.

CAT-Site: Predicting Protein Binding Sites Using a Convolutional Neural Network.

Identifying binding sites on the protein surface is an important part of computer-assisted drug design processes. Reliable prediction of binding sites not only assists with docking algorithms, but it can also explain the possible side-effects of a potential drug as well as its efficiency. In this work, we propose a novel workflow for predicting possible binding sites of a ligand on a protein surface. We use proteins from the PDBbind and sc-PDB databases, from which we combine available ligand information for similar proteins using all the possible ligands rather than only a special sub-selection to generalize the work of existing research. After performing protein clustering and merging of ligands of similar proteins, we use a three-dimensional convolutional neural network that takes into account the spatial structure of a protein. Lastly, we combine ligandability predictions for points on protein surfaces into joint binding sites. Analysis of our model's performance shows that its achieved sensitivity is 0.829, specificity is 0.98, and F1 score is 0.517, and that for 54% of larger and pharmacologically relevant binding sites, the distance between their real and predicted centers amounts to less than 4 Å.

Chickens, more than humans, focus the diversity of their immunoglobulin genes on the complementarity-determining region but utilise amino acids, indicative of a more cross-reactive antibody repertoire.

The mechanisms of B-cell diversification differ greatly between aves and mammals, but both produce B cells and antibodies capable of supporting an effective immune response. To see how differences in the generation of diversity might affect overall repertoire diversity, we have compared the diversity characteristics of immunoglobulin genes from domestic chickens to those from humans. Both use V(D)J gene rearrangement and somatic hypermutation, but only chickens use somatic gene conversion. A range of diversity analysis tools were used to investigate multiple aspects of amino acid diversity at both the germline and repertoire levels. The effect of differing amino acid usages on antibody characteristics was assessed. At both the germline and repertoire levels, chickens exhibited lower amino acid diversity in comparison to the human immunoglobulin genes, especially outside of the complementarity-determining region (CDR). Chickens were also found to possess much larger and more hydrophilic CDR3s with a higher predicted protein binding potential, suggesting that the antigen-binding site in chicken antibodies is more flexible and more polyreactive than that seen in human antibodies.

Language models for the prediction of SARS-CoV-2 inhibitors.

The COVID-19 pandemic highlights the need for computational tools to automate and accelerate drug design for novel protein targets. We leverage deep learning language models to generate and score drug candidates based on predicted protein binding affinity. We pre-trained a deep learning language model (BERT) on ∼9.6 billion molecules and achieved peak performance of 603 petaflops in mixed precision. Our work reduces pre-training time from days to hours, compared to previous efforts with this architecture, while also increasing the dataset size by nearly an order of magnitude. For scoring, we fine-tuned the language model using an assembled set of thousands of protein targets with binding affinity data and searched for inhibitors of specific protein targets, SARS-CoV-2 Mpro and PLpro. We utilized a genetic algorithm approach for finding optimal candidates using the generation and scoring capabilities of the language model. Our generalizable models accelerate the identification of inhibitors for emerging therapeutic targets.

Predicting residues involved in anti-DNA autoantibodies with limited neural networks.

Computer-aided rational vaccine design (RVD) and synthetic pharmacology are rapidly developing fields that leverage existing datasets for developing compounds of interest. Computational proteomics utilizes algorithms and models to probe proteins for functional prediction. A potentially strong target for computational approach is autoimmune antibodies, which are the result of broken tolerance in the immune system where it cannot distinguish “self” from “non-self” resulting in attack of its own structures (proteins and DNA, mainly). The information on structure, function, and pathogenicity of autoantibodies may assist in engineering RVD against autoimmune diseases. Current computational approaches exploit large datasets curated with extensive domain knowledge, most of which include the need for many resources and have been applied indirectly to problems of interest for DNA, RNA, and monomer protein binding. We present a novel method for discovering potential binding sites. We employed long short-term memory (LSTM) models trained on FASTA primary sequences to predict protein binding in DNA-binding hydrolytic antibodies (abzymes). We also employed CNN models applied to the same dataset for comparison with LSTM. While the CNN model outperformed the LSTM on the primary task of binding prediction, analysis of internal model representations of both models showed that the LSTM models recovered sub-sequences that were strongly correlated with sites known to be involved in binding. These results demonstrate that analysis of internal processes of LSTM models may serve as a powerful tool for primary sequence analysis.Graphical abstract

Vibrational Spectroscopy, Quantum Computational and Molecular Docking Studies on 2-[(1H-Benzimidazol-1-yl)-methyl]benzoic Acid

Experimental and theoretical investigations on the optimized geometrical structure, electronic and vibrational features of 2-[(1H-benzimidazol-1-yl)-methyl]benzoic acid are provided using the B3LYP/6-311++G(d,p) basis set. The Vibrational Energy Distribution Analysis (VEDA) program was used to perform the vibrational assignments and calculate the Potential Energy Distribution (PED). The acquired FT-IR and FT Raman data were used to complete the vibrational assignment and characterization of the compound fundamental modes. Theoretical and actual NMR chemical shifts were found to be quite similar. The UV-vis spectrum of 21HBMBA, as well as effects of solvents, have been investigated. The calculated HOMO and LUMO energies reveal that charge transfer happens within the molecule and MEP surface to be a chemically reactive area appropriate for drug action. Furthermore, a thorough examination of Non-Bonding Orbitals, excitation energies, AIM charges, Fukui functions and the Electron Localization Function (ELF) is carried out. The research is also expanded to compute first-order hyperpolarizability and forecast NLO characteristics. The details of the docking studies aided in the prediction of protein binding.

Challenges and Advantages of Accounting for Backbone Flexibility in Prediction of Protein-Protein Complexes.

Predicting protein binding is a core problem of computational biophysics. That this objective can be partly achieved with some amount of success using docking algorithms based on rigid protein models is remarkable, although going further requires allowing for protein flexibility. However, accurately capturing the conformational changes upon binding remains an enduring challenge for docking algorithms. Here, we adapt our Upside folding model, where side chains are represented as multi-position beads, to explore how flexibility may impact predictions of protein-protein complexes. Specifically, the Upside model is used to investigate where backbone flexibility helps, which types of interactions are important, and what is the impact of coarse graining. These efforts also shed light on the relative challenges posed by folding and docking. After training the Upside energy function for docking, the model is competitive with the established all-atom methods. However, allowing for backbone flexibility during docking is generally detrimental, as the presence of comparatively minor (3-5 Å) deviations relative to the docked structure has a large negative effect on performance. While this issue appears to be inherent to current forcefield-guided flexible docking methods, systems involving the co-folding of flexible loops such as antibody-antigen complexes represent an interesting exception. In this case, binding is improved when backbone flexibility is allowed using the Upside model.

Computational approach for binding prediction of SARS-CoV-2 with neutralizing antibodies

Coronavirus disease 2019 (COVID-19) caused by a novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread worldwide as a severe pandemic and caused enormous global health and economical damage. Since December 2019, more than 197 million cases have been reported, causing 4.2 million deaths. In the settings of pandemic it is an urgent necessity for the development of an effective COVID-19 treatment. While in-vitro screening of hundreds of antibodies isolated from convalescent patients is challenging due to its high cost, use of computational methods may provide an attractive solution in selecting the top candidates. Here, we developed a computational approach (SARS-AB) for binding prediction of spike protein SARS-CoV-2 with monoclonal antibodies. We validated our approach using existing structures in the protein data bank (PDB), and demonstrated its prediction power in antibody-spike protein binding prediction. We further tested its performance using antibody sequences from the literature where crystal structure is not available, and observed a high prediction accuracy (AUC = 99.6%). Finally, we demonstrated that SARS-AB can be used to design effective antibodies against novel SARS-CoV-2 mutants that might escape the current antibody protections. We believe that SARS-AB can significantly accelerate the discovery of neutralizing antibodies against SARS-CoV-2 and its mutants.

In silico analysis of AhyI protein and AI-1 inhibition using N-cis-octadec-9z-enoyl-l-homoserine lactone inhibitor in Aeromonas hydrophila

AhyI is homologous to the protein LuxI and is conserved throughout bacterial species including Aeromonas hydrophila. A. hydrophila causes opportunistic infections in fish and other aquatic organisms. Furthermore, this pathogennot only poses a great risk for the aquaculture industry, but also for human public health. AhyI (expressing acylhomoserine lactone) is responsible for the biosynthesis of autoinducer-1 (AI-1), commonly referred to as a quorum sensing (QS) signaling molecule, which plays an essential role in bacterial communication. Studying protein structure is essential for understanding molecular mechanisms of pathogenicity in microbes. Here, we have deduced a predicted structure of AhyI protein and characterized its function using in silico methods to aid the development of new treatments for controlling A.hydrophila infections. In addition to modeling AhyI, an appropriate inhibitor molecule was identified via high throughput virtual screening (HTVS) using mcule drug-like databases.The AhyI-inhibitor N-cis-octadec-9Z-enoyl-l-Homoserine lactone was selected withthe best drug score. In order to understand the pocket sites (ligand binding sites) and their interaction with the selected inhibitor, docking (predicted protein binding complex) servers were used and the selected ligand was docked with the predicted AhyI protein model. Remarkably, N-cis-octadec-9Z-enoyl-l-Homoserine lactone established interfaces with the protein via16 residues (V24, R27, F28, R31, W34, V36, D45, M77, F82, T101, R102, L103, 104, V143, S145, and V168), which are involved with regulating mechanisms of inhibition. These proposed predictions suggest that this inhibitor molecule may be used as a novel drug candidate for the inhibition of auto-inducer-1 (AI-1) activity.The N-cis-octadec-9Z-enoyl-l-Homoserine lactone inhibitor molecule was studied on cultured bacteria to validate its potency against AI-1 production. At a concentration of 40 μM, optimal inhibition efficiency of AI-1 was observedin bacterial culture media.These results suggest that the inhibitor molecule N-cis-octadec-9Z-enoyl-l-Homoserine lactone is a competitive inhibitor of AI-1 biosynthesis.

Plasma protein binding prediction focusing on residue-level features and circularity of cyclic peptides by deep learning.

MotivationIn recent years, cyclic peptide drugs have been receiving increasing attention because they can target proteins that are difficult to be tackled by conventional small-molecule drugs or antibody drugs. Plasma protein binding rate () is a significant pharmacokinetic property of a compound in drug discovery and design. However, due to structural differences, previous computational prediction methods developed for small-molecule compounds cannot be successfully applied to cyclic peptides, and methods for predicting the PPB rate of cyclic peptides with high accuracy are not yet available.ResultsCyclic peptides are larger than small molecules, and their local structures have a considerable impact on PPB; thus, molecular descriptors expressing residue-level local features of cyclic peptides, instead of those expressing the entire molecule, as well as the circularity of the cyclic peptides should be considered. Therefore, we developed a prediction method named CycPeptPPB using deep learning that considers both factors. First, the macrocycle ring of cyclic peptides was decomposed residue by residue. The residue-based descriptors were arranged according to the sequence information of the cyclic peptide. Furthermore, the circular data augmentation method was used, and the circular convolution method CyclicConv was devised to express the cyclic structure. CycPeptPPB exhibited excellent performance, with mean absolute error (MAE) of 4.79% and correlation coefficient (R) of 0.92 for the public drug dataset, compared to the prediction performance of the existing PPB rate prediction software ().Availability and implementationThe data underlying this article are available in the online supplementary material. The source code of CycPeptPPB is available at https://github.com/akiyamalab/cycpeptppb.Supplementary information Supplementary data are available at Bioinformatics online.

DeepCLIP: predicting the effect of mutations on protein-RNA binding with deep learning.

Nucleotide variants can cause functional changes by altering protein–RNA binding in various ways that are not easy to predict. This can affect processes such as splicing, nuclear shuttling, and stability of the transcript. Therefore, correct modeling of protein–RNA binding is critical when predicting the effects of sequence variations. Many RNA-binding proteins recognize a diverse set of motifs and binding is typically also dependent on the genomic context, making this task particularly challenging. Here, we present DeepCLIP, the first method for context-aware modeling and predicting protein binding to RNA nucleic acids using exclusively sequence data as input. We show that DeepCLIP outperforms existing methods for modeling RNA-protein binding. Importantly, we demonstrate that DeepCLIP predictions correlate with the functional outcomes of nucleotide variants in independent wet lab experiments. Furthermore, we show how DeepCLIP binding profiles can be used in the design of therapeutically relevant antisense oligonucleotides, and to uncover possible position-dependent regulation in a tissue-specific manner. DeepCLIP is freely available as a stand-alone application and as a webtool at http://deepclip.compbio.sdu.dk.