Biology subjects

Bonvin, A. M.

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

5 recordsLinked to original sources

bioRxiv2026

Patient-derived monoclonal anti-HPA-1a can induce platelet activation through FcγRIIa

Fetal and neonatal alloimmune thrombocytopenia (FNAIT) is a pregnancy-associated disorder caused by maternal alloantibodies targeting paternally inherited human platelet antigens (HPAs). These antibodies traverse the placenta and bind to fetal platelets, causing thrombocytopenia and potentially severe complications including intracranial hemorrhage (ICH). Beyond platelets, antibodies may also target other fetal cells such as endothelial cells and placental trophoblasts thereby exacerbating disease severity. Anti-HPA-1a antibodies, the primary cause of FNAIT, display functional heterogeneity due to differences in epitope specificity, glycosylation, which complicate disease severity prediction. Here, we report the structural and functional characterization of two novel anti-HPA-1a antibodies, D-204 and M-204, derived from a mother with a severe FNAIT case. In comparison to D-204 and other existing anti-HPA-1a antibodies such as B2G1 and 26.4, M-204 bound IIb{beta}3 with different kinetics, reaching only half-maximal binding, as determined by flow cytometry. Notably, only M-204 induced platelet aggregation. To investigate the underlying mechanism behind this aggregation, various antibody formats were generated (Fab, Fab2, bispecific, and IgG-Fc dead). Aggregation required both Fab-arms to engage HPA-1a and intact binding to Fc{gamma}RIIa. Structural models and crystal structures suggest the dual engagement of both Fab-arms is required for Fc{gamma}RIIA-engagement which is a unique feature of M-204 among known HPA-1 monoclonals. Our findings highlight the functional diversity of anti-HPA-1a antibodies and provide new insights into their pathogenic mechanisms.

immunology↗

bioRxiv2025

DeepRank-Ab: a dedicated scoring function for antibody-antigen complexes based on geometric deep learning

Gaining structural insights into the interactions between antibodies and their corresponding antigens is essential for understanding immune recognition and for guiding therapeutic antibody design. However, accurately modelling these complexes remains a significant challenge for both physics-based docking approaches and AI-based, co-folding methods such as AlphaFold3. These methods not only struggle to generate near native conformations, but, more critically, they often fail to score and rank such conformations correctly, revealing fundamental limitations when applied to antibody-antigen systems. To address these limitations, we present here DeepRank-Ab, a geometric deep learning-based scoring function tailored to the unique characteristics of antibody-antigen interfaces. Its development was enabled by a rigorously curated benchmark comprising more than 2.3 million decoys generated from 1,442 complexes, providing the diversity required for robust training and unbiased evaluation. Leveraging this resource, we systematically assessed multiple levels of graph representation, structural and energetic feature sets, and sampling strategies. Building on our previous DeepRank-GNN-esm work, our analysis identified that atom-level graph representation coupled with Voronoi-based surface decomposition and antibody-specific features is the most effective formulation for accurate scoring. Across multiple independent test sets, including models from unbound unbound docking and structures generated by AlphaFold, DeepRank-Ab consistently outperforms all evaluated methods, including AF3, HADDOCK and state of the art scoring functions such as FTDMP. It increases the AF3 Top 1 success rate by 35.5% and improves the mean Top 1 DockQ by more than a factor of two. DeepRank-Ab further generalizes robustly beyond its training distribution, achieving 100% Top 5 success rate on external antibody-antigen CAPRI targets, surpassing all tested methods. Together, these results demonstrate that DeepRank-Ab is a highly effective scoring method that substantially improves the identification of near-native antibody-antigen conformations.

bioinformatics↗

bioRxiv2024

Modelling Protein-Glycan Interactions with HADDOCK

The term glycan refers to a broad category of molecules composed of monosaccharides units linked to each other in a variety of ways, whose structural diversity is related to different functions in living organisms. Among others, glycans are recognized by proteins with the aim of carrying information and for signalling purposes. Determining the three-dimensional structures of protein-protein-glycan complexes is essential both for the understanding of the mechanisms glycans are involved in, and for applications such as drug design. In this context, molecular docking approaches are of undoubted importance as complementary approaches to experiments. In this study, we show how HADDOCK can be efficiently used for the prediction of protein-glycan complexes. Using a benchmark of 89 complexes, starting from their bound or unbound forms, and assuming some knowledge of the binding site on the protein, our protocol reaches a 70% and 40% top 5 success rate on bound and unbound datasets, respectively. We show that the main limiting factor is related to the complexity of the glycan to be modelled and the associated conformational flexibility.

bioinformatics↗

bioRxiv2023

DeepRank-GNN-esm: A Graph Neural Network for Scoring Protein-Protein Models using Protein Language Model

MotivationProtein-Protein interactions (PPIs) play critical roles in numerous cellular processes. By modelling the three-dimensional structures of the correspond protein complexes valuable insights can be obtained, providing, for example, starting points for drug and protein design. One challenge in the modelling process is however the identification of near-native models from the large pool of generated models. To this end we previously developed DeepRank-GNN, a graph neural network that integrates structural and sequence information to enable effective pattern learning at PPI interfaces. Its main features are related to the Position Specific Scoring Matrices (PSSM), which are computationally expensive to generate and significantly limit the algorithms usability. ResultsWe introduce here DeepRank-GNN-esm that includes as additional features protein language model embeddings from the EMS-2 model. We show that the ESM-2 embeddings can actually replace the PSSM features at no cost in-, or even better performance on two PPI-related tasks: scoring docking poses and detecting crystal artifacts. This new DeepRank version bypasses thus the need of generating PSSM, greatly improving the usability of the software and opening new application opportunities for systems for which PSSM profiles cannot be obtained or are irrelevant (e.g. antibody-antigen complexes). Availability and implementationDeepRank-GNN-esm is freely available from https://github.com/DeepRank/DeepRank-GNN-esm

bioinformatics↗

bioRxiv2020

Integrative Modeling of Membrane-associated Protein Assemblies

Historically, membrane protein systems have been considered as one of the most challenging systems to study with experimental structural biology techniques. Over the past years, increased number of experimental structures of membrane proteins have become available thanks in particular to advances in solid-state NMR spectroscopy and cryo-electron microscopy. This has opened the route to modeling the complexes that those membrane proteins form by methods such as docking. Most approaches developed to date are, however, not capable of incorporating the topological information provided by the membrane into the modeling process. Here, we present an integrative computational protocol for the modeling of membrane-associated protein assemblies, specifically complexes consisting of a membrane-embedded protein and a soluble partner. It combines efficient, artificial intelligence-based rigid-body docking by LightDock with a flexible final refinement with HADDOCK to remove potential clashes at the interface. We make use of an equilibrated coarse-grained lipid bilayer to represent the information encoded in the membrane in the form of artificial beads, which allows to target the docking towards the binding-competent regions. We demonstrate the performance of this membrane-driven protocol on eighteen membrane-associated complexes, whose interface lies between the membrane and either the cytosolic or periplasmic regions. In addition, we evaluate how different membrane definitions impact the performance of the docking protocol and provide a comparison, in terms of success rate, to another state-of-the-art docking software, ZDOCK. Finally, we discuss the quality of the generated models and propose possible future developments. Our membrane docking protocol should allow to shed light on the still rather dark fraction of the interactome consisting of membrane proteins.

bioinformatics↗