José Antonio Márquez
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
José Antonio Márquez has a total of 172 co-authors. Here are the top co-authors by number of shared publications:
Florine Dupeux
10 shared publicationsUlrich Zander
Grenoble Outstation, European Molecular Biology Laboratory; Unit of Virus Host-Cell Interactions (UMI 3265), University Grenoble Alpes–EMBL–CNRS, 71 Avenue des Martyrs, 38042 Grenoble, France
5 shared publicationsAnne-Sophie Humm
4 shared publications- Florine Dupeux
Univ. Grenoble Alpes, CNRS, CEA, Institut de Biologie Structurale
3 shared publications Martin Röwer
Grenoble Outstation, European Molecular Biology Laboratory; Unit of Virus Host-Cell Interactions (UMI 3265), University Grenoble Alpes–EMBL–CNRS, 71 Avenue des Martyrs, 38042 Grenoble, France
3 shared publications- Sihyun Sung
European Molecular Biology Laboratory, 71 Avenue des Martyrs, Grenoble 38000, France
3 shared publications
Latest Publications
- Open Access
- 28 October 2026
- Published by EMBL-EBI
Automated X-ray fragment and ligand screening for therapeutic innovation in infection biology
Fundamental knowledge in infection biology accumulates at an incredible pace, revealing new molecular targets that play key roles in infection, replication, and pathogen metabolism as well as essential host factors. These represent new opportunities for the development of novel anti-infective strategies. However, translation of fundamental discoveries into therapeutic applications has traditionally been hindered by a scarcity of approaches for thorough functional validation. Recent technological breakthroughs have made large-scale screening of small molecules against relevant targets by X-ray crystallography more accessible, helping accelerate the development of small molecule inhibitors through structure-based drug design. This approach can now be applied by researchers on a large scale to investigate the function of molecular targets and establish proof of concept of their therapeutic potential. In this webinar, we will present new technologies developed at EMBL Grenoble enabling fully automated, large-scale fragment and ligand screening and their application to the study and validation of key targets with strong therapeutic potential in infection biology.
- 28 October 2026
- Published by EMBL-EBI
Open AccessAutomated X-ray fragment and ligand screening for therapeutic innovation in infection biology
Fundamental knowledge in infection biology accumulates at an incredible pace, revealing new molecular targets that play key roles in infection, replication, and pathogen metabolism as well as essential host factors. These represent new opportunities for the development of novel anti-infective strategies. However, translation of fundamental discoveries into therapeutic applications has traditionally been hindered by a scarcity of approaches for thorough functional validation. Recent technological breakthroughs have made large-scale screening of small molecules against relevant targets by X-ray crystallography more accessible, helping accelerate the development of small molecule inhibitors through structure-based drug design. This approach can now be applied by researchers on a large scale to investigate the function of molecular targets and establish proof of concept of their therapeutic potential. In this webinar, we will present new technologies developed at EMBL Grenoble enabling fully automated, large-scale fragment and ligand screening and their application to the study and validation of key targets with strong therapeutic potential in infection biology.
- Open Access
- 19 March 2026
- Published by International Union of Crystallography (IUCr) in IUCrJ
- Vol. 13 (3) , 235-248
AXIS: a Lab-in-the-Loop machine learning approach for automated detection of macromolecular crystals
Macromolecular crystallography provides mechanistic understanding of biological processes and can be applied in drug design. Nowadays, the use of robotic systems for crystal growth and diffraction analysis is widespread and high-throughput protein-to-structure pipelines for ligand and fragment screening are revolutionizing the field. However, the identification of crystals is still largely carried out through manual inspection, sometimes involving tens of thousands of images, which represents a bottleneck in an otherwise highly automated process. Here we describe AXIS, an AI-based Crystal Identification System combining the DINOv2 computer vision model, state-of-the-art transfer learning and MARCO, the largest crystallization dataset available to date, for automated crystal detection. AXIS can operate with both visible and UV light images and integrates a Lab-in-the-Loop approach combining ML and expert inputs for iterative learning and specialization. AXIS enables automated annotation of large crystallization image datasets with performance and accuracy comparable to that of human experts, and the Lab-in-the-Loop approach introduced here enables efficient adaptation to local conditions, facilitating widespread application, which has been a major limitation to date. AXIS can help to correct human errors in image annotation and removes critical bottlenecks, particularly in the context of extensive crystallization screens or high-throughput applications like fragment and ligand screening, unlocking the potential for higher levels of automation that are key in both fundamental and translational research.
- 19 March 2026
- Published by International Union of Crystallography (IUCr) in IUCrJ
- Vol. 13 (3) , 235-248
Open AccessAXIS: a Lab-in-the-Loop machine learning approach for automated detection of macromolecular crystals
Macromolecular crystallography provides mechanistic understanding of biological processes and can be applied in drug design. Nowadays, the use of robotic systems for crystal growth and diffraction analysis is widespread and high-throughput protein-to-structure pipelines for ligand and fragment screening are revolutionizing the field. However, the identification of crystals is still largely carried out through manual inspection, sometimes involving tens of thousands of images, which represents a bottleneck in an otherwise highly automated process. Here we describe AXIS, an AI-based Crystal Identification System combining the DINOv2 computer vision model, state-of-the-art transfer learning and MARCO, the largest crystallization dataset available to date, for automated crystal detection. AXIS can operate with both visible and UV light images and integrates a Lab-in-the-Loop approach combining ML and expert inputs for iterative learning and specialization. AXIS enables automated annotation of large crystallization image datasets with performance and accuracy comparable to that of human experts, and the Lab-in-the-Loop approach introduced here enables efficient adaptation to local conditions, facilitating widespread application, which has been a major limitation to date. AXIS can help to correct human errors in image annotation and removes critical bottlenecks, particularly in the context of extensive crystallization screens or high-throughput applications like fragment and ligand screening, unlocking the potential for higher levels of automation that are key in both fundamental and translational research.
- Open Access
AXIS: A Lab-in-the-Loop Machine Learning approach for generalized detection of macromolecular crystals
Macromolecular crystallography provides mechanistic understanding of biological processes and can be applied in drug design. Nowadays, the use of robotic systems for crystal growth and diffraction analysis is widespread and high throughput protein-to-structure pipelines for ligand and fragment screening are revolutionizing the field. However, the identification of crystals is still largely carried out through manual inspection, sometimes involving tens of thousands of images, which represents a bottleneck in an otherwise highly automated process. Here we describe AXIS , an A I-based C rystal I dentification S ystem combining the DINOv2 computer vision model, state-of-the-art transfer learning and MARCO, the largest crystallization dataset available to date, for automated crystal detection. AXIS can operate both with visible and UV light images and integrates a Lab-In-The-Loop approach combining ML and expert inputs for continuous learning and specialization. AXIS enables automated annotation of large crystallization image datasets with performance and accuracy comparable to that of human experts and the Lab-In-The-Loop approach introduced here enables efficient adaptation to local conditions facilitating widespread application, which has been a major limitation to date. AXIS can help correct human errors in image annotation and removes critical bottlenecks, particularly in the context of extensive crystallization screens or high throughput applications like fragment and ligand screening unlocking the potential for higher levels of automation that are key both in fundamental and translational research. Appendix A. Graphical
Open AccessAXIS: A Lab-in-the-Loop Machine Learning approach for generalized detection of macromolecular crystals
Macromolecular crystallography provides mechanistic understanding of biological processes and can be applied in drug design. Nowadays, the use of robotic systems for crystal growth and diffraction analysis is widespread and high throughput protein-to-structure pipelines for ligand and fragment screening are revolutionizing the field. However, the identification of crystals is still largely carried out through manual inspection, sometimes involving tens of thousands of images, which represents a bottleneck in an otherwise highly automated process. Here we describe AXIS , an A I-based C rystal I dentification S ystem combining the DINOv2 computer vision model, state-of-the-art transfer learning and MARCO, the largest crystallization dataset available to date, for automated crystal detection. AXIS can operate both with visible and UV light images and integrates a Lab-In-The-Loop approach combining ML and expert inputs for continuous learning and specialization. AXIS enables automated annotation of large crystallization image datasets with performance and accuracy comparable to that of human experts and the Lab-In-The-Loop approach introduced here enables efficient adaptation to local conditions facilitating widespread application, which has been a major limitation to date. AXIS can help correct human errors in image annotation and removes critical bottlenecks, particularly in the context of extensive crystallization screens or high throughput applications like fragment and ligand screening unlocking the potential for higher levels of automation that are key both in fundamental and translational research. Appendix A. Graphical
- Open Access
It’s Not Always Crystal Clear: In-Pocket Analysis for Understanding Binding Site Microenvironment Properties
Structural biology enabled key advances in biochemistry, drug discovery, and AI-driven three-dimensional structure modelling. Despite its importance, reliable and general computational tools for chemical validation of structural models remain limited, especially when small-molecule ligands are involved. This challenge is exacerbated when experimental resolution is insufficient to define interaction chemistry, leaving critical interpretations ambiguous. We introduce In-Pocket, a quantum mechanical optimisation method biased by experimental data that evaluates the chemical nature of protein-ligand binding interfaces. In-Pocket improves model quality, corrects chemically inconsistent structures, and identifies the energetically most favourable ligand-conformers within biomacromolecular complexes. The approach is generally applicable to a wide range of systems, including those containing transition metals. Benchmarking demonstrates that In-Pocket enhances the reliability of legacy and newly solved structures. With automated, user-friendly routine chemical validation, In-Pocket offers unique opportunities to improve the integrity of structural databases and impact diverse domains: from molecular biology to rational drug design and application of generative AI.
Open AccessIt’s Not Always Crystal Clear: In-Pocket Analysis for Understanding Binding Site Microenvironment Properties
Structural biology enabled key advances in biochemistry, drug discovery, and AI-driven three-dimensional structure modelling. Despite its importance, reliable and general computational tools for chemical validation of structural models remain limited, especially when small-molecule ligands are involved. This challenge is exacerbated when experimental resolution is insufficient to define interaction chemistry, leaving critical interpretations ambiguous. We introduce In-Pocket, a quantum mechanical optimisation method biased by experimental data that evaluates the chemical nature of protein-ligand binding interfaces. In-Pocket improves model quality, corrects chemically inconsistent structures, and identifies the energetically most favourable ligand-conformers within biomacromolecular complexes. The approach is generally applicable to a wide range of systems, including those containing transition metals. Benchmarking demonstrates that In-Pocket enhances the reliability of legacy and newly solved structures. With automated, user-friendly routine chemical validation, In-Pocket offers unique opportunities to improve the integrity of structural databases and impact diverse domains: from molecular biology to rational drug design and application of generative AI.
- Open Access
EnsembleFlex: Protein Structure Ensemble Analysis Made Easy
EnsembleFlex is a novel tool designed to perform dual-scale flexibility analysis of protein structure ensembles, encompassing both backbone and side-chain dynamics. It integrates dedicated superposition methodologies, two dimension reduction techniques, customizable clustering, an automated ligand binding site analysis and analysis of conserved water molecules, enhancing the accuracy of ensemble flexibility analysis and visualisation. With a user-friendly, no-coding graphical interface as well as a code-based option, EnsembleFlex is accessible to a broad range of researchers. The tool complements experimental ensemble data analysis with predictive methods based on elastic network models, bridging the gap between experimental and computational approaches in protein flexibility analysis, and does so from a global backbone-based perspective, as well as from a focused side-chain and binding site-centred point of view. The need for reliable, streamlined, fast, and user-friendly tools for structural ensemble analysis is critical in today’s research environment, and EnsembleFlex addresses these needs efficiently, offering an accessible solution that integrates powerful analysis methods in a single package. Its wide applicability across various domains of molecular design underscores the importance of streamlined flexibility analysis in scientific research.
Open AccessEnsembleFlex: Protein Structure Ensemble Analysis Made Easy
EnsembleFlex is a novel tool designed to perform dual-scale flexibility analysis of protein structure ensembles, encompassing both backbone and side-chain dynamics. It integrates dedicated superposition methodologies, two dimension reduction techniques, customizable clustering, an automated ligand binding site analysis and analysis of conserved water molecules, enhancing the accuracy of ensemble flexibility analysis and visualisation. With a user-friendly, no-coding graphical interface as well as a code-based option, EnsembleFlex is accessible to a broad range of researchers. The tool complements experimental ensemble data analysis with predictive methods based on elastic network models, bridging the gap between experimental and computational approaches in protein flexibility analysis, and does so from a global backbone-based perspective, as well as from a focused side-chain and binding site-centred point of view. The need for reliable, streamlined, fast, and user-friendly tools for structural ensemble analysis is critical in today’s research environment, and EnsembleFlex addresses these needs efficiently, offering an accessible solution that integrates powerful analysis methods in a single package. Its wide applicability across various domains of molecular design underscores the importance of streamlined flexibility analysis in scientific research.
- Open Access
- 12 March 2024
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 13
X-Ray Structure and enzymatic study of a Bacterial NADPH oxidase highlight the activation mechanism of eukaryotic NOX
NADPH oxidases (NOX) are transmembrane proteins that professionally produce reactive oxygen species (ROS) and are distributed widely in both eukaryotes and prokaryotes. Eukaryotes use the ROS products for innate immune defense and signaling; the seven human isoforms of NOX participate in critical physiological and pathophysiological processes. Recently solved structures of two human NOX isoforms provide much new information, but do not fully elucidate controls on the electron transfer pathway from NAD(P)H substrate through FAD and heme cofactors to the final ROS product. SpNOX, a bacterial NOX homolog from Streptococcus pneumoniae, shows robust constitutive activity in detergent solution, making it a good prototype for exploring electron transfer in the NOX family. Here we present crystal structures of wildtype and mutant full-length and dehydrogenase (DH) domain-only constructs of SpNOX. The isolated DH domain acts as a flavin reductase, and both DH and full-length constructs use either NADPH or NADH as substrate. Our data supports hydride transfer from NAD(P)H to FAD as the rate limiting step in electron transfer. Using the DH domain, we demonstrate the role of F397 in allowing access of nicotinamide to the flavin isoalloxazine, while in the full-length construct, we used mutants and flavin analogs to confirm the contribution of both domains to flavin binding observed in the structure. Comparison with homologous enzymes suggests distal heme access may influence the final electron acceptor, while the relative position of DH and TM does not necessarily correlate with activity of a given NOX family member. Based on this comparison, SpNOX appears to be a good model of active NOX2, which allows us to propose an explanation for NOX2’s requirement for activation.
- 12 March 2024
- Published by eLife Sciences Publications, Ltd in eLife preprints
- Vol. 13
Open AccessX-Ray Structure and enzymatic study of a Bacterial NADPH oxidase highlight the activation mechanism of eukaryotic NOX
NADPH oxidases (NOX) are transmembrane proteins that professionally produce reactive oxygen species (ROS) and are distributed widely in both eukaryotes and prokaryotes. Eukaryotes use the ROS products for innate immune defense and signaling; the seven human isoforms of NOX participate in critical physiological and pathophysiological processes. Recently solved structures of two human NOX isoforms provide much new information, but do not fully elucidate controls on the electron transfer pathway from NAD(P)H substrate through FAD and heme cofactors to the final ROS product. SpNOX, a bacterial NOX homolog from Streptococcus pneumoniae, shows robust constitutive activity in detergent solution, making it a good prototype for exploring electron transfer in the NOX family. Here we present crystal structures of wildtype and mutant full-length and dehydrogenase (DH) domain-only constructs of SpNOX. The isolated DH domain acts as a flavin reductase, and both DH and full-length constructs use either NADPH or NADH as substrate. Our data supports hydride transfer from NAD(P)H to FAD as the rate limiting step in electron transfer. Using the DH domain, we demonstrate the role of F397 in allowing access of nicotinamide to the flavin isoalloxazine, while in the full-length construct, we used mutants and flavin analogs to confirm the contribution of both domains to flavin binding observed in the structure. Comparison with homologous enzymes suggests distal heme access may influence the final electron acceptor, while the relative position of DH and TM does not necessarily correlate with activity of a given NOX family member. Based on this comparison, SpNOX appears to be a good model of active NOX2, which allows us to propose an explanation for NOX2’s requirement for activation.
- Open Access
X-Ray Structure and enzymatic study of a Bacterial NADPH oxidase highlight the activation mechanism of eukaryotic NOX
NADPH oxidases (NOX) are transmembrane proteins that professionally produce reactive oxygen species (ROS) and are distributed widely in both eukaryotes and prokaryotes. Eukaryotes use the ROS products for innate immune defense and signaling; the seven human isoforms of NOX participate in critical physiological and pathophysiological processes. Recently solved structures of two human NOX isoforms provide much new information, but do not fully elucidate controls on the electron transfer pathway from NAD(P)H substrate through FAD and heme cofactors to the final ROS product. SpNOX, a bacterial NOX homolog from Streptococcus pneumoniae , shows robust constitutive activity in detergent solution, making it a good prototype for exploring electron transfer in the NOX family. Here we present crystal structures of wildtype and mutant full-length and dehydrogenase (DH) domain-only constructs of SpNOX. The isolated DH domain acts as a flavin reductase, and both DH and full-length constructs use either NADPH or NADH as substrate. Our data supports hydride transfer from NAD(P)H to FAD as the rate limiting step in electron transfer. Using the DH domain, we demonstrate the role of F397 in allowing access of nicotinamide to the flavin isoalloxazine, while in the full-length construct, we used mutants and flavin analogs to confirm the contribution of both domains to flavin binding observed in the structure. Comparison with homologous enzymes suggests distal heme access may influence the final electron acceptor, while the relative position of DH and TM does not necessarily correlate with activity of a given NOX family member. Based on this comparison, SpNOX appears to be a good model of active NOX2, which allows us to propose an explanation for NOX2’s requirement for activation.
Open AccessX-Ray Structure and enzymatic study of a Bacterial NADPH oxidase highlight the activation mechanism of eukaryotic NOX
NADPH oxidases (NOX) are transmembrane proteins that professionally produce reactive oxygen species (ROS) and are distributed widely in both eukaryotes and prokaryotes. Eukaryotes use the ROS products for innate immune defense and signaling; the seven human isoforms of NOX participate in critical physiological and pathophysiological processes. Recently solved structures of two human NOX isoforms provide much new information, but do not fully elucidate controls on the electron transfer pathway from NAD(P)H substrate through FAD and heme cofactors to the final ROS product. SpNOX, a bacterial NOX homolog from Streptococcus pneumoniae , shows robust constitutive activity in detergent solution, making it a good prototype for exploring electron transfer in the NOX family. Here we present crystal structures of wildtype and mutant full-length and dehydrogenase (DH) domain-only constructs of SpNOX. The isolated DH domain acts as a flavin reductase, and both DH and full-length constructs use either NADPH or NADH as substrate. Our data supports hydride transfer from NAD(P)H to FAD as the rate limiting step in electron transfer. Using the DH domain, we demonstrate the role of F397 in allowing access of nicotinamide to the flavin isoalloxazine, while in the full-length construct, we used mutants and flavin analogs to confirm the contribution of both domains to flavin binding observed in the structure. Comparison with homologous enzymes suggests distal heme access may influence the final electron acceptor, while the relative position of DH and TM does not necessarily correlate with activity of a given NOX family member. Based on this comparison, SpNOX appears to be a good model of active NOX2, which allows us to propose an explanation for NOX2’s requirement for activation.
- Open Access
Plasticity of the binding pocket in peptide transporters underpins promiscuous substrate recognition
Summary: Proton-coupled oligopeptide transporters (POTs) are promiscuous transporters of the Major Facilitator Superfamily, that constitute the main route of entry for a wide range of dietary peptides and orally administrated peptidomimetic drugs. Given their clinical and pathophysiological relevance, several bacterial and mammalian POT homologs have been extensively studied on a structural and molecular level. However, the molecular basis of recognition and transport of the wide range of peptide substrates has remained elusive. Here we present 14 X-ray structures of the bacterial POT DtpB in complex with chemically diverse di- and tripeptides, providing novel insights into the plasticity of the conserved central binding cavity. We analyzed binding affinities for more than 80 peptides and monitored uptake by a fluorescence-based transport assay. To probe if all natural 8400 di- and tripeptides can bind to DtpB, we employed state-of-the-art molecular docking and machine learning and conclude that peptides of a specific subset with compact hydrophobic residues are the best DtpB binders.
Open AccessPlasticity of the binding pocket in peptide transporters underpins promiscuous substrate recognition
Summary: Proton-coupled oligopeptide transporters (POTs) are promiscuous transporters of the Major Facilitator Superfamily, that constitute the main route of entry for a wide range of dietary peptides and orally administrated peptidomimetic drugs. Given their clinical and pathophysiological relevance, several bacterial and mammalian POT homologs have been extensively studied on a structural and molecular level. However, the molecular basis of recognition and transport of the wide range of peptide substrates has remained elusive. Here we present 14 X-ray structures of the bacterial POT DtpB in complex with chemically diverse di- and tripeptides, providing novel insights into the plasticity of the conserved central binding cavity. We analyzed binding affinities for more than 80 peptides and monitored uptake by a fluorescence-based transport assay. To probe if all natural 8400 di- and tripeptides can bind to DtpB, we employed state-of-the-art molecular docking and machine learning and conclude that peptides of a specific subset with compact hydrophobic residues are the best DtpB binders.
- 19 October 2022
- Published by Elsevier in European Journal of Medicinal Chemistry
- Vol. 244, 114857
Rational design of the zonulin inhibitor AT1001 derivatives as potential anti SARS-CoV-2
- 19 October 2022
- Published by Elsevier in European Journal of Medicinal Chemistry
- Vol. 244, 114857
Rational design of the zonulin inhibitor AT1001 derivatives as potential anti SARS-CoV-2
- 4 April 2022
- Published by Springer Nature in The EMBO Journal
- Vol. 41 (7) , e110799
Erratum To: A thermodynamic switch modulates abscisic acid receptor sensitivity
- 4 April 2022
- Published by Springer Nature in The EMBO Journal
- Vol. 41 (7) , e110799
Erratum To: A thermodynamic switch modulates abscisic acid receptor sensitivity