Simulating large molecular systems over long timescales requires force fields that are both accurate and efficient. In recent years, E(3) equivariant neural networks have lifted the tension between computational efficiency and accuracy of force fields, but they are still several orders of magnitude more expensive than established molecular mechanics (MM) force fields. Here, we propose Grappa, a machine learning framework to predict MM parameters from the molecular graph, employing a graph attentional neural network and a transformer with symmetry-preserving positional encoding. The resulting Grappa force field outperforms tabulated and machine-learned MM force fields in terms of accuracy at the same computational efficiency and can be used in existing Molecular Dynamics (MD) engines like GROMACS and OpenMM. It predicts energies and forces of small molecules, peptides, and RNA at state-of-the-art MM accuracy, while also reproducing experimentally measured values for J-couplings. With its simple input features and high data-efficiency, Grappa is well suited for extensions to uncharted regions of chemical space, which we show on the example of peptide radicals. We demonstrate Grappa's transferability to macromolecules in MD simulations from a small fast-folding protein up to a whole virus particle. Our force field sets the stage for biomolecular simulations closer to chemical accuracy, but with the same computational cost as established protein force fields.

Predicting reaction barriers for arbitrary configurations based on only a limited set of density functional theory (DFT) calculations would render the design of catalysts or the simulation of reactions within complex materials highly efficient. We here propose Gaussian process regression (GPR) as a method of choice if DFT calculations are limited to hundreds or thousands of barrier calculations. For the case of hydrogen atom transfer in proteins, an important reaction in chemistry and biology, we obtain a mean absolute error of 3.23 kcal mol-1 for the range of barriers in the data set using SOAP descriptors and similar values using the marginalized graph kernel. Thus, the two GPR models can robustly estimate reaction barriers within the large chemical and conformational space of proteins. Their predictive power is comparable to a graph neural network-based model, and GPR even outcompetes the latter in the low data regime. We propose GPR as a valuable tool for an approximate but data-efficient model of chemical reactivity in a complex and highly variable environment.

Proton pump inhibitors have become top-selling drugs worldwide. Serendipitously discovered as prodrugs that are activated by protonation in acidic environments, proton pump inhibitors inhibit stomach acid secretion by covalently modifying the gastric proton pump. Despite their widespread use, alternative activation mechanisms and potential target proteins in non-acidic environments remain poorly understood. Employing a chemoproteomic approach, we found that the proton pump inhibitor rabeprazole selectively forms covalent conjugates with zinc-binding proteins. Focusing on DENR, a protein with a C4 zinc cluster (that is, zinc coordinated by four cysteines), we show that rabeprazole is activated by the zinc ion and subsequently conjugated to zinc-coordinating cysteines. Our results suggest that drug binding, activation and conjugation take place rapidly within the zinc coordination sphere. Finally, we provide evidence that other proton pump inhibitors can be activated in the same way. We conclude that zinc acts as a Lewis acid, obviating the need for low pH, to promote the activation and conjugation of proton pump inhibitors in non-acidic environments.

MOTIVATION:Collagen fibrils are fundamental building blocks of connective tissues, yet generating accurate molecular models of their structure remains challenging due to their hierarchical organization and complex crosslinking patterns. RESULTS:ColBuilder has been developed to automate the generation of atomistic models of crosslinked collagen fibrils and facilitate the setup of molecular simulations. The tool integrates homology modeling, higher order structure generation and optimization to build complete fibril structures with precise control over sequence composition, crosslinking patterns, and dimensions. Users can explore different collagen sequences, manipulate crosslink chemistry through mixed ratios and densities, and generate fibrils of varying diameter and length. All-atom molecular dynamics simulations of 335nm-long fibrils validate the generated structures, showing excellent agreement with experimental measurements of D-band periodicity and force-extension behavior. AVAILABILITY AND IMPLEMENTATION:ColBuilder is available both as an open-source command-line application and through a web interface at https://colbuilder.mpip-mainz.mpg.de.

Nucleic acids, such as DNA, are integral components of biological systems in that they steer many cellular processes and biotechnological applications. In addition, their monomer-precise sequence and accurately predictable structure render them an excellent model for exploring fundamental problems in nanotechnology and polymer science. In the field of polymer mechanochemistry, predetermined breaking points, called mechanophores, are used to endow macromolecules with chain-scission selectivity when subjected to external forces. However, this approach entails cumbersome chemical synthesis and limited outcome analysis. Here, we show the mechanophore-free, near-nucleotide-precise scission of nicked double-stranded DNA in a combined experimental and computational approach. We leverage next-generation sequencing to achieve monomer-level precision in assessing chain scission. Additionally, we monitor and control the scission distribution on the polymer’s backbone. Our research highlights the potential of DNA as a model polymer in the field of polymer mechanochemistry.

Small Ubiquitin-related modifiers of the SUMO family regulate thousands of proteins in eukaryotic cells. Many SUMO substrates, effectors and enzymes carry short motifs (SIMs) that mediate low affinity interactions with SUMO proteins. This raises the question how specificity is achieved in target selection, SUMO paralogue choice and SUMO-dependent interactions. A unique but poorly understood feature of SUMO proteins is their intrinsically disordered N-terminus. We reveal a function for N-termini of human, C. elegans, and yeast SUMO proteins as intramolecular inhibitors of SUMO-SIM interactions. Mutational analyses, NMR spectroscopy, and Molecular Dynamics simulations indicate that SUMO’s N-terminus can inhibit SIM binding by fast and fuzzy interactions with SUMO‘s core. Deletion of the C. elegans SUMO1 N-terminus leads to p53-dependent apoptosis during germline development, indicating an important role of SUMO’s N-termini in DNA damage repair. Our findings reveal a mechanism of disorder-based autoinhibition that contributes to the specificity of SUMOylation and SUMO-dependent interactions.

Three-dimensional (3D) bioprinting is a promising technology which typically uses bioinks to pattern cells and their scaffolds. The selection of cytocompatible inks is critical for the printing success. In laserbased 3D bioprinting, photoresist molecules are used as bioinks. We propose that cytotoxicity can be a consequence of the interaction of photoresists with lipid membranes and their permeation into the cell. Here, molecular dynamics simulations and in vitro assays address this issue, retrieving partition coefficients, free energies, and permeabilities for eight commonly-used photoresists in model lipid bilayers. Crossing the hydrophobic center of the membrane constitutes the rate limiting step during permeation. In addition, three photoresists feature a preferential localization site at the acyl chain headgroup interface. Photoresist permeabilities range over eight orders of magnitude, with some molecules being membrane-permeable on bioprinting timescales. Moreover, permeation correlates well with the oil-water partition coefficients and is severely hampered by the lipid ordering imposed by the lipid saturation. Overall, the mechanism of interaction of photoresists with model lipid bilayers is provided here, helping to classify them according to their residence in the membrane and permeation through it. This is useful information to guide the selection of cytocompatible photoresists for 3D bioprinting.

Stronger chemical bonds withstand higher mechanical forces; thus, the rupture of single bonds is preferred over the rupture of double or triple bonds or aromatic rings. We investigated bond scission in poly(dialkyl-p-phenylene ethynylene)s (PPEs), a fully conjugated polymer. In a scale-bridging approach using electron-paramagnetic resonance spectroscopy and gel permeation chromatography of cryomilled samples, in combination with density functional theory calculations and coarse-grained simulations, we conclude that mechanical force cleaves the sp-sp2 bond of PPEs (bond dissociation energy as high as 600 kJ mol(-1)). Bond scission primarily occurs in shear bands with locally increased shear stresses. The scission occurs in the middle of the PPE chains. Breaking sp-sp(2) bonds into free radicals thus is feasible but requires significant mechanical force and an efficient stress concentration.

Pathogens use sophisticated adhesion mechanisms to remain attached to the host's surfaces. A key example of this is the adhesion of erythrocytes infected with Plasmodium, the parasite which causes malaria, to the microvasculature. Remarkably, in the case of pregnancy-associated malaria, the adherence of parasitized erythrocytes to the placenta is enhanced by the shear of the flowing blood, suggesting a catch-bond adhesion mechanism. The adhesion is mediated by a parasite protein called VAR2CSA which anchors such infected erythrocytes to the proteoglycan matrix of the placenta.

Mimicking extracellular matrices holds great potential for tissue engineering in biological and biomedical applications. A key compound for the mechanical stability of these matrices is collagen, which also plays an important role in many intra- and intercellular processes. Two-photon 3D laser printing offers structuring of these matrices with subcellular resolution. So far, efforts on 3D microprinting of collagen have been limited to simple geometries and customized set-ups. Herein, an easily accessible approach is presented using a collagen type I methacrylamide (ColMA) ink system which can be stored at room temperature and be precisely printed using a commercial two-photon 3D laser printer. The formulation and printing parameters are carefully optimized enabling the manufacturing of defined 3D microstructures. Furthermore, these printed microstructures show a fully reversible response upon heating and cooling in multiple cycles, indicating successful collagen folding and unfolding. This experimental observation has been supported by molecular dynamics simulations. Thus, the study opens new perspectives for designing new responsive biomaterials for 4D (micro)printing.

We introduce a generative model for protein backbone design utilizing geometric products and higher order message passing. In particular, we propose Clifford Frame Attention (CFA), an extension of the invariant point attention (IPA) architecture from AlphaFold2, in which the backbone residue frames and geometric features are represented in the projective geometric algebra. This enables to construct geometrically expressive messages between residues, including higher order terms, using the bilinear operations of the algebra. We evaluate our architecture by incorporating it into the framework of FrameFlow, a state-of-the-art flow matching model for protein backbone generation. The proposed model achieves high designability, diversity and novelty, while also sampling protein backbones that follow the statistical distribution of secondary structure elements found in naturally occurring proteins, a property so far only insufficiently achieved by many state-of-the-art generative models.

Collagen, the most abundant protein in the human body, must withstand high mechanical loads due to its structural role in tendons, skin, bones, and other connective tissue. It was recently found that tensed collagen creates mechanoradicals by homolytic bond scission. We here employ scale-bridging simulations to determine the influence of collagen's mesoscale fibril structure on molecular breakages, combining atomistic molecular dynamics simulations with a newly developed mesoscopic ultra-coarse-grained description of a collagen fibril. Our simulations identify a conserved structural feature, a length difference of the two helices between pairs of crosslinks, to play a critical role. The release of the extra hidden length enables collagen to buffer mechanical stress. At the same time, this topology funnels ruptures such that the potentially harmful mechanoradicals are readily stabilized, buffering the arising oxidative stress. Our results suggest collagen's hidden length to exploit a sweet spot in the trade-off between breakage specificity and strength.

Lubricin, an intrinsically disordered glycoprotein, plays a pivotal role in facilitating smooth movement and ensuring the enduring functionality of synovial joints. The central domain of this protein serves as a source of this excellent lubrication and is characterized by its highly glycosylated, negatively charged, and disordered structure. However, the influence of O-glycans on the viscosity of lubricin remains unclear. In this study, we employ molecular dynamics simulations in the absence and presence of shear, along with continuum simulations, to elucidate the intricate interplay between O-glycans and lubricin and the impact of O-glycans on lubricin's conformational properties and viscosity. We found the presence of O-glycans to induce a more extended conformation in fragments of the disordered region of lubricin. These O-glycans contribute to a reduction in solution viscosity but at the same time weaken shear thinning at high shear rates, compared to nonglycosylated systems with the same density. This effect is attributed to the steric and electrostatic repulsion between the fragments, which prevents their conglomeration and structuring. Our computational study yields a mechanistic mechanism underlying previous experimental observations of lubricin and paves the way to a more rational understanding of its function in the synovial fluid.

Hydrogen atom transfer (HAT) reactions are important in many biological systems. As these reactions are hard to observe experimentally, it is of high interest to shed light on them using simulations. Here, we present a machine learning model based on graph neural networks for the prediction of energy barriers of HAT reactions in proteins. As input, the model uses exclusively non-optimized structures as obtained from classical simulations. It was trained on more than 17,000 energy barriers calculated using hybrid density functional theory. We built and evaluated the model in the context of HAT in collagen, but we show that the same workflow can easily be applied to HAT reactions in other biological or synthetic polymers. We obtain for relevant reactions (small reaction distances) a model with good predictive power (R2 ∼ 0.9 and mean absolute error of < 3 kcal/mol). As the inference speed is high, this model enables evaluations of dozens of chemical situations within seconds. When combined with molecular dynamics in a kinetic Monte-Carlo scheme, the model paves the way toward reactive simulations.