Biology subjects

Bonvin, A. M. J. J.

Publications and source records attributed to Bonvin, A. M. J. J..

29 records · Page 2Linked to original sources

bioRxiv2021

Cryptic SARS-CoV2-spike-with-sugar interactions revealed by 'universal' saturation transfer analysis

Many host pathogen interactions such as human viruses (including non-SARS-coronaviruses) rely on attachment to host cell-surface glycans. There are conflicting reports about whether the Spike protein of SARS-CoV-2 binds to sialic acid commonly found on host cell-surface N-linked glycans. In the absence of a biochemical assay, the ability to analyze the binding of glycans to heavily- modified proteins and resolve this issue is limited. Classical Saturation Transfer Difference (STD) NMR can be confounded by overlapping sugar resonances that compound with known experimental constraints. Here we present universal saturation transfer analysis (uSTA), an NMR method that builds on existing approaches to provide a general and automated workflow for studying protein-ligand interactions. uSTA reveals that B-origin-lineage-SARS-CoV-2 spike trimer binds sialoside sugars in an end on manner and modelling guided by uSTA localises binding to the spike N-terminal domain (NTD). The sialylated-polylactosamine motif is found on tetraantennary human N-linked-glycoproteins in deeper lung and may have played a role in zoonosis. Provocatively, sialic acid binding is abolished by mutations in some subsequent SARS- CoV-2 variants-of-concern. A very high resolution cryo-EM structure confirms the NTD location and end on mode; it rationalises the effect of NTD mutations and the structure-activity relationship of sialic acid analogues. uSTA is demonstrated to be a robust, rapid and quantitative tool for analysis of binding, even in the most demanding systems. Extended AbstractThe surface proteins found on both pathogens and host cells mediate entry (and exit) and influence disease progression and transmission. Both types can bear host-generated post- translational modifications such as glycosylation that are essential for function but can confound biophysical methods used for dissecting key interactions. Several human viruses (including non- SARS-coronaviruses) attach to host cell-surface N-linked glycans that include forms of sialic acid (sialosides). There remains, however, conflicting evidence as to if or how SARS-associated coronaviruses might use such a mechanism. Here, we demonstrate quantitative extension of saturation transfer protein NMR methods to a complete mathematical model of the magnetization transfer caused by interactions between protein and ligand. The method couples objective resonance-identification via a deconvolution algorithm with Bloch-McConnell analysis to enable a structural, kinetic and thermodynamic analysis of ligand binding beyond previously-perceived limits of exchange rates, concentration or system. Using an automated and openly available workflow this universal saturation transfer analysis (uSTA) can be readily-applied in a range of even heavily-modified systems in a general manner to now obtain quantitative binding interaction parameters (KD, kEx). uSTA proved critical in mapping direct interactions between natural sialoside sugar ligands and relevant virus-surface attachment glycoproteins - SARS-CoV-2-spike and influenza-H1N1-haemagglutinin variants - by quantitating ligand signal in spectral regions otherwise occluded by resonances from mobile protein glycans (that also include sialosides). In B- origin-lineage-SARS-CoV-2 spike trimer end on-binding to sialoside sugars was revealed contrasting with extended surface-binding for heparin sugar ligands; uSTA-derived constraints used in structural modelling suggested sialoside-glycan binding sites in a beta-sheet-rich region of spike N-terminal domain (NTD). Consistent with this, uSTA-glycan binding was minimally- perturbed by antibodies that neutralize the ACE2-binding domain (RBD) but strongly disrupted in spike from the B1.1.7/alpha and B1.351/beta variants-of-concern, which possess hotspot mutations in the NTD. Sialoside binding in B-origin-lineage-NTD was unequivocally pinpointed by cryo-EM to a site that is created from residues that are notably deleted in variants (e.g. H69,V70,Y145 in alpha). An analysis of beneficial genetic variances in cohorts of patients from early 2020 suggests a model in which this site in the NTD of B-origin-lineage-SARS-CoV-2 (but not in alpha/beta-variants) may have exploited a specific sialylated-polylactosamine motif found on tetraantennary human N-linked-glycoproteins in deeper lung. Together these confirm a novel binding mode mediated by the unusual NTD of SARS-CoV-2 and suggest how it may drive virulence and/or zoonosis via modulation of glycan attachment. Since cell-surface glycans are widely relevant to biology and pathology, uSTA can now provide ready, quantitative, widespread analysis of complex, host-derived and post-translationally modified proteins with putative ligands relevant to disease even in previously confounding complex systems.

biophysics↗

bioRxiv2021

Native or non-native protein-protein docking models? Molecular dynamics to the rescue.

Molecular docking excels at creating a plethora of potential models of protein-protein complexes. To correctly distinguish the favourable, native-like models from the remaining ones remains, however, a challenge. We assessed here if a protocol based on molecular dynamics (MD) simulations would allow to distinguish native from non-native models to complement scoring functions used in docking. To this end, first models for 25 protein-protein complexes were generated using HADDOCK. Next, MD simulations complemented with machine learning were used to discriminate between native and non-native complexes based on a combination of metrics reporting on the stability of the initial models. Native models showed higher stability in almost all measured properties, including the key ones used for scoring in the CAPRI competition, namely the positional root mean square deviations and fraction of native contacts from the initial docked model. A Random Forest classifier was trained, reaching 0.85 accuracy in correctly distinguishing native from non-native complexes. Reasonably modest simulation lengths in the order of 50 to 100 ns are already sufficient to reach this accuracy, which makes this approach applicable in practice.

bioinformatics↗

bioRxiv2021

DeepRank: A deep learning framework for data mining 3D protein-protein interfaces

Three-dimensional (3D) structures of protein complexes provide fundamental information to decipher biological processes at the molecular scale. The vast amount of experimentally and computationally resolved protein-protein interfaces (PPIs) offers the possibility of training deep learning models to aid the predictions of their biological relevance. We present here DeepRank, a general, configurable deep learning framework for data mining PPIs using 3D convolutional neural networks (CNNs). DeepRank maps features of PPIs onto 3D grids and trains a user-specified CNN on these 3D grids. DeepRank allows for efficient training of 3D CNNs with data sets containing millions of PPIs and supports both classification and regression. We demonstrate the performance of DeepRank on two distinct challenges: The classification of biological versus crystallographic PPIs, and the ranking of docking models. For both problems DeepRank is competitive or outperforms state-of-the-art methods, demonstrating the versatility of the framework for research in structural biology.

bioinformatics↗

bioRxiv2020

proABC-2: PRediction Of AntiBody Contacts v2 and its application to information-driven docking

Monoclonal antibodies (mAbs) are essential tools in the contemporary therapeutic armoury. Understanding how these recognize their antigen is a fundamental step in their rational design and engineering. The rising amount of publicly available data is catalysing the development of computational approaches able to offer valuable, faster and cheaper alternatives to classical experimental methodologies used for the study of antibody-antigen complexes. Here we present proABC-2, an update of the original random-forest antibody paratope predictor, based on a convolutional neural network algorithm. We also demonstrate how the predictions can be fruitfully used to drive the docking in HADDOCK. The proABC-2 server is freely available at: https://bianca.science.uu.nl/proabc2/.

bioinformatics↗

bioRxiv2019

An overview of data-driven HADDOCK strategies in CAPRI rounds 38-45.

Our information-driven docking approach HADDOCK has demonstrated a sustained performance since the start of its participation to CAPRI. This is due, in part, to its ability to integrate data into the modelling process, and to the robustness of its scoring function. We participated in CAPRI both as server and as manual predictors.\n\nIn CAPRI rounds 38-45, we have used various strategies depending on the information at hand. These ranged from imposing restraints to a few residues identified from literature as being important for the interaction, to binding pockets identified from homologous complexes or template-based refinement / CA-CA restraint-guided docking from identified templates. When relevant, symmetry restraints were used to limit the conformational sampling. We also tested for a large decamer target a new implementation of the MARTINI coarse-grained force field in HADDOCK. Overall in the current rounds, we obtained acceptable or better predictions for 13 and 11 server and manual submissions, respectively, out of the 22 interfaces. Our server performance (acceptable models) was better (59%) than the manual (50%) one, in which we typically experiment with various combinations of protocols and data sources. Again, our simple scoring function based on a linear combination of intermolecular van der Waals and electrostatic energies and an empirical desolvation term demonstrated a good performance in the scoring experiment with a 63% success rate across all 22 interfaces.\n\nAn analysis of model quality indicates that, while we are consistently performing well in generating acceptable models, there is room for improvement for generating/identifying higher quality models.

bioinformatics↗

bioRxiv2019

Less is more: Coarse-grained integrative modeling of large biomolecular assemblies with HADDOCK

Predicting the 3D structure of protein interactions remains a challenge in the field of computational structural biology. This is in part due to difficulties in sampling the complex energy landscape of multiple interacting flexible polypeptide chains. Coarse-graining approaches, which reduce the number of degrees of freedom of the system, help address this limitation by smoothing the energy landscape, allowing an easier identification of the global energy minimum. They also accelerate the calculations, allowing to model larger assemblies. Here, we present the implementation of the MARTINI coarse-grained force field for proteins into HADDOCK, our integrative modelling platform. Docking and refinement are performed at the coarse-grained level and the resulting models are then converted back to atomistic resolution through a distance restraints-guided morphing procedure. Our protocol, tested on the largest complexes of the protein docking benchmark 5, shows an overall ~7-fold speed increase compared to standard all-atom calculations, while maintaining a similar accuracy and yielding substantially more near-native solutions. To showcase the potential of our method, we performed simultaneous 7 body docking to model the 1:6 KaiC-KaiB complex, integrating mutagenesis and hydrogen/deuterium exchange data from mass spectrometry with symmetry restraints, and validated the resulting models against a recently published cryo-EM structure.

bioinformatics↗

bioRxiv2019

LightDock goes information-driven

The use of experimental information has been demonstrated to increase the success rate of computational macromolecular docking. Many methods use information to post-filter the simulation output while others drive the simulation based on experimental restraints, which can become problematic for more complex scenarios such as multiple binding interfaces. We present a novel two-step method for including interface information into protein docking simulations within the LightDock framework. Prior to the simulation, irrelevant regions from the receptor are excluded for sampling (filter of initial swarms) and initial ligand poses are pre-oriented based on ligand input information. We demonstrate the applicability of this approach on the new 55 cases of the Protein-Protein Docking Benchmark 5, using different amounts of information. Even with incomplete information, a significant improvement in performance is obtained compared to blind ab initio docking.\n\nThe software is supported and freely available from https://github.com/brianjimenez/lightdock and analysis data from https://github.com/brianjimenez/lightdock_bm5.

bioinformatics↗

bioRxiv2018

iScore: A novel graph kernel-based function for scoring protein-protein docking models

Protein complexes play a central role in many aspects of biological function. Knowledge of the three-dimensional (3D) structures of protein complexes is critical for gaining insights into the structural basis of interactions and their roles in the biomolecular pathways that orchestrate key cellular processes. Because of the expense and effort associated with experimental determination of 3D structures of protein complexes, computational docking has evolved as a valuable tool to predict the 3D structures of biomolecular complexes. Despite recent progress, reliably distinguishing near-native docking conformations from a large number of candidate conformations, the so-called scoring problem, remains a major challenge. Here we present iScore, a novel approach to scoring docked conformations that combines HADDOCK energy terms with a score obtained using a graph representation of the protein-protein interfaces and a measure of evolutionary conservation. It achieves a scoring performance competitive with, or superior to that of the state-of-the-art scoring functions on independent data sets consisting docking software-specific data sets and the CAPRI score set built from a wide variety of docking approaches. iScore ranks among the top scoring approaches on the CAPRI score set (13 targets) when compared with the 37 scoring groups in CAPRI. The results demonstrate the utility of combining evolutionary and topological, and physicochemical information for scoring docked conformations. This work represents the first successful demonstration of graph kernel to protein interfaces for effective discrimination of near-native and non-native conformations of protein complexes. It paves the way for the further development of computational methods for predicting the structure of protein complexes.

bioinformatics↗

bioRxiv2018

pdb-tools: a swiss army knife for molecular structures

The pdb-tools are a collection of Python scripts for working with molecular structure data in the PDB format. They allow users to edit, convert, and validate PDB files, from the command-line, in a simple but efficient manner. The pdb-tools are implemented in Python, without any external dependencies, and are freely available under the open-source Apache License at https://github.com/haddocking/pdb-tools/ and on PyPI (https://pypi.org/project/pdb-tools/).

bioinformatics↗

bioRxiv2018

An enhanced-sampling MD-based protocol for molecular docking

Understanding molecular recognition of proteins by small molecules is key for drug design. Despite the number of experimental structures of ligand-protein complexes keeps growing, the number of available targets remains limited compared to the druggable genome, and structural diversity is generally low, which affects the chemical variance of putative lead compounds. From a computational perspective, molecular docking is widely used to mimic ligand-protein association in silico. Ensemble-docking approaches include flexibility through a set of different conformations of the protein obtained either experimentally or from computer simulations, e.g. molecular dynamics. However, structures prone to host (the correct) ligands are generally poorly sampled by standard molecular dynamics simulations of the apo protein. In order to address this limitation, we introduce a computational approach based on metadynamics simulations (EDES - Ensemble-Docking with Enhanced-sampling of pocket Shape) to generate druggable conformations of proteins only exploiting their apo structures. This is achieved by defining a set of collective variables that effectively sample different shapes of the binding site, ultimately mimicking the steric effect due to ligands to generate holo-like binding site geometries. We assessed the method on two challenging proteins undergoing different extents of conformational changes upon ligand binding. In both cases our protocol generated a significant fraction of structures featuring a low RMSD from the experimental holo conformation. Moreover, ensemble docking calculations using those conformations yielded native-like poses among the top ranked ones for both targets. This proof of concept study paves the route towards an automated workflow to generate druggable conformations of proteins, which should become a precious tool for structure-based drug design.

biophysics↗

bioRxiv2018

iSEE: Interface Structure, Evolution and Energy-based machine learning predictor of binding affinity changes upon mutations

Quantitative evaluation of binding affinity changes upon mutations is crucial for protein engineering and drug design. Machine learning-based methods are gaining increasing momentum in this field. Due to the limited number of experimental data, using a small number of sensitive predictive features is vital to the generalization and robustness of such machine learning methods. Here we introduce a fast and reliable predictor of binding affinity changes upon single point mutation, based on a random forest approach. Our method, iSEE, uses a limited number of interface Structure, Evolution and Energy-based features for the prediction. iSEE achieves, using only 31 features, a high prediction performance with a Pearson correlation coefficient (PCC) of 0.80 and a root mean square error of 1.41 kcal mol-1 on a diverse training dataset consisting of 1102 mutations in 57 protein-protein complexes. It competes with existing state-of-the-art methods on two blind test datasets. Predictions for a new dataset of 540 mutations in 58 protein complexes from the recently published SKEMPI 2.0 database reveals that none of the current methods perform well (PCC<0.4), although their combination does improve the predictions. Feature analysis for iSEE underlines the significance of evolutionary conservations for quantitative prediction of mutation effects.

bioinformatics↗