Silvia Gervasoni

Bio: Silvia Gervasoni is an academic researcher from University of Bristol. The author has contributed to research in topics: Zinc. The author has co-authored 1 publications. Previous affiliations of Silvia Gervasoni include University of Milan.

Multiscale Workflow for Modeling Ligand Complexes of Zinc Metalloproteins.

Abstract: Zinc metalloproteins are ubiquitous, with protein zinc centers of structural and functional importance, involved in interactions with ligands and substrates and often of pharmacological interest. Biomolecular simulations are increasingly prominent in investigations of protein structure, dynamics, ligand interactions, and catalysis, but zinc poses a particular challenge, in part because of its versatile, flexible coordination. A computational workflow generating reliable models of ligand complexes of biological zinc centers would find broad application. Here, we evaluate the ability of alternative treatments, using (nonbonded) molecular mechanics (MM) and quantum mechanics/molecular mechanics (QM/MM) at semiempirical (DFTB3) and density functional theory (DFT) levels of theory, to describe the zinc centers of ligand complexes of six metalloenzyme systems differing in coordination geometries, zinc stoichiometries (mono- and dinuclear), and the nature of interacting groups (specifically the presence of zinc-sulfur interactions). MM molecular dynamics (MD) simulations can overfavor octahedral geometries, introducing additional water molecules to the zinc coordination shell, but this can be rectified by subsequent semiempirical (DFTB3) QM/MM MD simulations. B3LYP/MM geometry optimization further improved the accuracy of the description of coordination distances, with the overall effectiveness of the approach depending upon factors, including the presence of zinc-sulfur interactions that are less well described by semiempirical methods. We describe a workflow comprising QM/MM MD using DFTB3 followed by QM/MM geometry optimization using DFT (e.g., B3LYP) that well describes our set of zinc metalloenzyme complexes and is likely to be suitable for creating accurate models of zinc protein complexes when structural information is more limited.

Mechanistic Insights into Enzyme Catalysis from Explaining Machine-Learned Quantum Mechanical and Molecular Mechanical Minimum Energy Pathways

Abstract: With the increasing popularity of machine learning (ML) applications, the demand for explainable artificial intelligence techniques to explain ML models developed for computational chemistry has also emerged. In this study, we present the development of the Boltzmann-weighted cumulative integrated gradients (BCIG) approach for effective explanation of mechanistic insights into ML models trained on high-level quantum mechanical and molecular mechanical (QM/MM) minimum energy pathways. Using the acylation reactions of the Toho-1 β-lactamase and two antibiotics (ampicillin and cefalexin) as the model systems, we show that the BCIG approach could quantitatively attribute the energetic contribution in one system and the relative reactivity of individual steps across different systems to specific chemical processes such as the bond making/breaking and proton transfers. The proposed BCIG contribution attribution method quantifies chemistry-interpretable insights in terms of contributions from each elementary chemical process, which is in agreement with the validating QM/MM calculations and our intuitive mechanistic understandings of the model reactions.

Pyrazolones Potentiate Colistin Activity against MCR-1-Producing Resistant Bacteria: Computational and Microbiological Study

Abstract: The polymyxin colistin is a last line antibiotic for extensively resistant Gram-negative bacteria. Colistin binding to lipid A disrupts the Gram-negative outer membrane, but mobile colistin resistance (mcr) gene family members confer resistance by catalyzing phosphoethanolamine (PEA) transfer onto lipid A, neutralizing its negative charge to reduce colistin interactions. Multiple mcr isoforms have been identified in clinical and environmental isolates, with mcr-1 being the most widespread and mcr-3 being common in South and East Asia. Preliminary screening revealed that treatment with pyrazolones significantly reduced mcr-1, but not mcr-3, mediated colistin resistance. Molecular dynamics (MD) simulations of the catalytic domains of MCR-1 and a homology model of MCR-3, in different protonation states of active site residues H395/H380 and H478/H463, indicate that the MCR-1 active site has greater water accessibility than MCR-3, but that this is less influenced by changes in protonation. MD-optimized structures of MCR-1 and MCR-3 were used in virtual screening of 20 pyrazolone derivatives. Docking of these into the MCR-1/MCR-3 active sites identifies common residues likely to be involved in protein–ligand interactions, specifically the catalytic threonine (MCR-1 T285, MCR-3 T277) site of PEA addition, as well as differential interactions with adjacent amino acids. Minimal inhibitory concentration assays showed that the pyrazolone with the lowest predicted binding energy (ST3f) restores colistin susceptibility of mcr-1, but not mcr-3, expressing Escherichia coli. Thus, simulations indicate differences in the active site structure between MCR-1 and MCR-3 that may give rise to differences in pyrazolone binding and so relate to differential effects upon producer E. coli. This work identifies pyrazolones as able to restore colistin susceptibility of mcr-1-producing bacteria, laying the foundation for further investigations of their activity as phosphoethanolamine transferase inhibitors as well as of their differential activity toward mcr isoforms.

Graph-learning guided mechanistic insights into imipenem hydrolysis in GES carbapenemases

Abstract: Pathogen resistance to carbapenem antibiotics compromises effective treatments of superbug infections. One major source of carbapenem resistance is the bacterial production of carbapenemases which effectively hydrolyze carbapenem drugs. In this computational study, the deacylation reaction of imipenem (IPM) by GES-5 carbapenemases (GES) is modeled to unravel the mechanistic factors that facilitate carbapenem resistance. Hybrid quantum mechanical/molecular mechanical (QM/MM) calculations are applied to sample the GES/IPM deacylation barriers on the minimum energy pathways (MEPs). In light of the recent emergence of graph-based deep-learning techniques, we construct graph representations of the GES/IPM active site. An edge-conditioned graph convolutional neural network (ECGCNN) is trained on the acyl-enzyme conformational graphs to learn the underlying correlations between the GES/IPM conformations and the deacylation barriers. A perturbative approach is proposed to interpret the latent representations from the graph-learning (GL) model and extract essential mechanistic understanding with atomistic detail. In general, our study combining QM/MM MEPs calculations and GL models explains mechanistic landscapes underlying the IPM resistance driven by GES carbapenemases. We also demonstrate that GL methods could effectively assist the post-analysis of QM/MM calculations whose data span high dimensionality and large sample-size.

Metal3D: a general deep learning framework for accurate metal ion location prediction in proteins

Abstract: Metal ions are essential cofactors for many proteins and play a crucial role in many applications such as enzyme design or design of protein-protein interactions because they are biologically abundant, tether to the protein using strong interactions, and have favorable catalytic properties. Computational design of metalloproteins is however hampered by the complex electronic structure of many biologically relevant metals such as zinc . In this work, we develop two tools - Metal3D (based on 3D convolutional neural networks) and Metal1D (solely based on geometric criteria) to improve the location prediction of zinc ions in protein structures. Comparison with other currently available tools shows that Metal3D is the most accurate zinc ion location predictor to date with predictions within 0.70 ± 0.64 Å of experimental locations. Metal3D outputs a confidence metric for each predicted site and works on proteins with few homologes in the protein data bank. Metal3D predicts a global zinc density that can be used for annotation of computationally predicted structures and a per residue zinc density that can be used in protein design workflows. Currently trained on zinc, the framework of Metal3D is readily extensible to other metals by modifying the training data.

Accurate prediction of transition metal ion location via deep learning

Abstract: Metal ions are essential cofactors for many proteins. In fact, currently, about half of the structurally characterized proteins contain a metal ion. Metal ions play a crucial role for many applications such as enzyme design or design of protein-protein interactions because they are biologically abundant, tether to the protein using strong interactions, and have favorable catalytic properties e.g. as Lewis acid. Computational design of metalloproteins is however hampered by the complex electronic structure of many biologically relevant metals such as zinc that can often not be accurately described using a classical force field. In this work, we develop two tools - Metal3D (based on 3D convolutional neural networks) and Metal1D (solely based on geometric criteria) to improve the identification and localization of zinc and other metal ions in experimental and computationally predicted protein structures. Comparison with other currently available tools shows that Metal3D is the most accurate metal ion location predictor to date outperforming geometric predictors including Metal1D by a wide margin using a single structure as input. Metal3D outputs a confidence metric for each predicted site and works on proteins with few homologes in the protein data bank. The predicted metal ion locations for Metal3D are within 0.70 ± 0.64 Å of the experimental locations with half of the sites below 0.5 Å. Metal3D predicts a global metal density that can be used for annotation of structures predicted using e.g. AlphaFold2 and a per residue metal density that can be used in protein design workflows for the location of suitable metal binding sites and rotamer sampling to create novel metalloproteins. Metal3D is available as easy to use webapp, notebook or commandline interface.