A neural network potential based on pairwise resolved atomic forces and energies.

Molecular simulations have become a key tool in molecular and materials design. Machine learning (ML)-based potential energy functions offer the prospect of simulating complex molecular systems efficiently at quantum chemical accuracy. In previous work, we have introduced the ML-based PairF-Net approach to neural network potentials, that adopts a pairwise interatomic scheme to predicting forces within a molecular system. Here, we further develop the PairF-Net model to intrinsically incorporate energy conservation and couple the model to a molecular mechanical (MM) environment within the OpenMM package. The updated PairF-Net model yields energy and force predictions and dynamical distributions in good agreement with the rMD17 dataset of ten small organic molecules in the gas-phase. We further show that these in vacuo ML models of small molecules can be applied to force predictions in aqueous solution via hybrid ML/MM simulations. We present a new benchmark dataset for these ten molecules in solution, obtained from QM/MM simulations, which we denote as rMD17-aq (https://zenodo.org/records/10048644); and assess the ability of PairF-Net to reproduce the molecular energy, atomic forces and dynamical distributions of these solution conformations via ML/MM simulations.

Boosting the Accuracy and Chemical Space Coverage of the Detection of Small Colloidal Aggregating Molecules Using the BAD Molecule Filter.

The ability to conduct effective high throughput screening (HTS) campaigns in drug discovery is often hampered by the detection of false positives in these assays due to small colloidally aggregating molecules (SCAMs). SCAMs can produce artifactual hits in HTS by nonspecific inhibition of the protein target. In this work, we present a new computational prediction tool for detecting SCAMs based on their 2D chemical structure. The tool, called the boosted aggregation detection (BAD) molecule filter, employs decision tree ensemble methods, namely, the CatBoost classifier and the light gradient-boosting machine, to significantly improve the detection of SCAMs. In developing the filter, we explore models trained on individual data sets, a consensus approach using these models, and, third, a merged data set approach, each tailored for specific drug discovery needs. The individual data set method emerged as most effective, achieving 93% sensitivity and 90% specificity, outperforming existing state-of-the-art models by 20 and 5%, respectively. The consensus models offer broader chemical space coverage, exceeding 90% for all testing sets. This feature is an important aspect particularly for early stage medicinal chemistry projects, and provides information on applicability domain. Meanwhile, the merged data set models demonstrated robust performance, with a notable sensitivity of 79% in the comprehensive 10-fold cross-validation test set. A SHAP analysis of model features indicates the importance of hydrophobicity and molecular complexity as primary factors influencing the aggregation propensity. The BAD molecule filter is readily accessible for the public usage on https://molmodlab-aau.com/Tools.html. This filter provides a new, more robust tool for aggregate prediction in the early stages of drug discovery to optimize hit rates and reduce associated testing and validation overheads.

Molecular dynamics simulations as a guide for modulating small molecule aggregation.

Small colloidally aggregating molecules (SCAMs) can be problematic for biological assays in drug discovery campaigns. However, the self-associating properties of SCAMs have potential applications in drug delivery and analytical biochemistry. Consequently, the ability to predict the aggregation propensity of a small organic molecule is of considerable interest. Chemoinformatics-based filters such as ChemAGG and Aggregator Advisor offer rapid assessment but are limited by the assay quality and structural diversity of their training set data. Complementary to these tools, we explore here the ability of molecular dynamics (MD) simulations as a physics-based method capable of predicting the aggregation propensity of diverse chemical structures. For a set of 32 molecules, using simulations of 100 ns in explicit solvent, we find a success rate of 97% (one molecule misclassified) as opposed to 75% by Aggregator Advisor and 72% by ChemAGG. These short timescale MD simulations are representative of longer microsecond trajectories and yield an informative spectrum of aggregation propensities across the set of solutes, capturing the dynamic behaviour of weakly aggregating compounds. Implicit solvent simulations using the generalized Born model were less successful in predicting aggregation propensity. MD simulations were also performed to explore structure-aggregation relationships for selected molecules, identifying chemical modifications that reversed the predicted behaviour of a given aggregator/non-aggregator compound. While lower throughput than rapid cheminformatics-based SCAM filters, MD-based prediction of aggregation has potential to be deployed on the scale of focused subsets of moderate size, and, depending on the target application, provide guidance on removing or optimizing a compound's aggregation propensity.

Modeling pyranose ring pucker in carbohydrates using machine learning and semi-empirical quantum chemical methods.

Pyranose ring pucker is a key coordinate governing the structure, interactions and reactivity of carbohydrates. We assess the ability of the machine learning potentials, ANI-1ccx and ANI-2x, and the GFN2-xTB semiempirical quantum chemical method, to model ring pucker conformers of five monosaccharides and oxane in the gas phase. Relative to coupled-cluster quantum mechanical calculations, we find that ANI-1ccx most accurately reproduces the ring pucker energy landscape for these molecules, with a correlation coefficient r2 of 0.83. This correlation in relative energies lowers to values of 0.70 for ANI-2x and 0.60 for GFN2-xTB. The ANI-1ccx also provides the most accurate estimate of the energetics of the 4 C1 -to-1 C4 minimum energy pathway for the six molecules. All three models reproduce chair more accurately than non-chair geometries. Analysis of small model molecules suggests that the ANI-1ccx model favors puckers with equatorial hydrogen bonding substituents; that ANI-2x and GFN2-xTB models overstabilize conformers with axially oriented groups; and that the endo-anomeric effect is overestimated by the machine learning models and underestimated via the GFN2-xTB method. While the pucker conformers considered in this study correspond to a gas phase environment, the accuracy and computational efficiency of the ANI-1ccx approach in modeling ring pucker in vacuo provides a promising basis for future evaluation and application to condensed phase environments.

A Selective Review and Virtual Screening Analysis of Natural Product Inhibitors of the NLRP3 Inflammasome.

The NLRP3 inflammasome is currently an exciting target for drug discovery due to its role in various inflammatory diseases; however, to date, no NLRP3 inhibitors have reached the clinic. Several studies have used natural products as hit compounds to facilitate the design of novel selective NLRP3 inhibitors. Here, we review selected natural products reported in the literature as NLRP3 inhibitors, with a particular focus on those targeting gout. To complement this survey, we also report a virtual screen of the ZINC20 natural product database, predicting favored chemical features that can aid in the design of novel small molecule NLRP3 inhibitors.

Lead Optimization of Dehydroemetine for Repositioned Use in Malaria.

Drug repositioning offers an effective alternative to de novo drug design to tackle the urgent need for novel antimalarial treatments. The antiamoebic compound emetine dihydrochloride has been identified as a potent in vitro inhibitor of the multidrug-resistant strain K1 of Plasmodium falciparum (50% inhibitory concentration [IC50], 47 nM ± 2.1 nM [mean ± standard deviation]). Dehydroemetine, a synthetic analogue of emetine dihydrochloride, has been reported to have less-cardiotoxic effects than emetine. The structures of two diastereomers of dehydroemetine were modeled on the published emetine binding site on the cryo-electron microscopy (cryo-EM) structure with PDB code 3J7A (P. falciparum 80S ribosome in complex with emetine), and it was found that (-)-R,S-dehydroemetine mimicked the bound pose of emetine more closely than did (-)-S,S-dehydroisoemetine. (-)-R,S-dehydroemetine (IC50 71.03 ± 6.1 nM) was also found to be highly potent against the multidrug-resistant K1 strain of P. falciparum compared with (-)-S,S-dehydroisoemetine (IC50, 2.07 ± 0.26 µM), which loses its potency due to the change of configuration at C-1'. In addition to its effect on the asexual erythrocytic stages of P. falciparum, the compound exhibited gametocidal properties with no cross-resistance against any of the multidrug-resistant strains tested. Drug interaction studies showed (-)-R,S-dehydroemetine to have synergistic antimalarial activity with atovaquone and proguanil. Emetine dihydrochloride and (-)-R,S-dehydroemetine failed to show any inhibition of the hERG potassium channel and displayed activity affecting the mitochondrial membrane potential, indicating a possible multimodal mechanism of action.

How Do Small Molecule Aggregates Inhibit Enzyme Activity? A Molecular Dynamics Study.

Small molecule compounds which form colloidal aggregates in solution are problematic in early drug discovery; adsorption of the target protein by these aggregates can lead to false positives in inhibition assays. In this work, we probe the molecular basis of this inhibitory mechanism using molecular dynamics simulations. Specifically, we examine in aqueous solution the adsorption of the enzymes ß-lactamase and PTP1B onto aggregates of the drug miconazole. In accordance with experiment, molecular dynamics simulations observe formation of miconazole aggregates as well as subsequent association of these aggregates with ß-lactamase and PTP1B. When complexed with aggregate, the proteins do not exhibit significant alteration in protein tertiary structure or dynamics on the microsecond time scale of the simulations, but they do indicate persistent occlusion of the protein active site by miconazole molecules. MD simulations further suggest this occlusion can occur via surficial interactions of protein with miconazole but also potentially by envelopment of the protein by miconazole. The heterogeneous polarity of the miconazole aggregate surface seems to underpin its activity as an invasive and nonspecific inhibitory agent. A deeper understanding of these protein/aggregate systems has implications not only for drug design but also for their exploitation as tools in drug delivery and analytical biochemistry.

Negative Cooperativity in NAD(P)H Quinone Oxidoreductase 1 (NQO1).

NAD(P)H quinone oxidoreductase-1 (NQO1) is a homodimeric protein that acts as a detoxifying enzyme or as a chaperone protein. Dicourmarol interacts with NQO1 at the NAD(P)H binding site and can both inhibit enzyme activity and modulate the interaction of NQO1 with other proteins. We show that the binding of dicoumarol and related compounds to NQO1 generates negative cooperativity between the monomers. This does not occur in the presence of the reducing cofactor, NAD(P)H, alone. Alteration of Gly150 (but not Gly149 or Gly174) abolished the dicoumarol-induced negative cooperativity. Analysis of the dynamics of NQO1 with the Gaussian network model indicates a high degree of collective motion by monomers and domains within NQO1. Ligand binding is predicted to alter NQO1 dynamics both proximal to the ligand binding site and remotely, close to the second binding site. Thus, drug-induced modulation of protein motion might contribute to the biological effects of putative inhibitors of NQO1.

Ring Puckering Landscapes of Glycosaminoglycan-Related Monosaccharides from Molecular Dynamics Simulations.

The conformational flexibility of the glycosaminoglycans (GAGs) is known to be key in their binding and biological function, for example in regulating coagulation and cell growth. In this work, we employ enhanced sampling molecular dynamics simulations to probe the ring conformations of GAG-related monosaccharides, including a range of acetylated and sulfated GAG residues. We first perform unbiased MD simulations of glucose anomers and the epimers glucuronate and iduronate. These calculations indicate that in some cases, an excess of 15 µs is required for adequate sampling of ring pucker due to the high energy barriers between states. However, by applying our recently developed msesMD simulation method (multidimensional swarm-enhanced sampling molecular dynamics), we were able to quantitatively and rapidly reproduce these ring pucker landscapes. From msesMD simulations, the puckering free energy profiles were then compared for 15 further monosaccharides related to GAGs; this includes to our knowledge the first simulation study of sulfation effects on ß-GalNAc ring puckering. For the force field employed, we find that in general the calculated pucker free energy profiles for sulfated sugars were similar to the corresponding unsulfated profiles. This accords with recent experimental studies suggesting that variation in ring pucker of sulfated GAG residues is primarily dictated by interactions with surrounding residues rather than by intrinsic conformational preference. As an exception to this, however, we predict that 4-O-sulfation of ß-GalNAc leads to reduced ring rigidity, with a significant lowering in energy of the 1C4 ring conformation; this observation may have implications for understanding the structural basis of the biological function of ß-GalNAc-containing glycosaminoglycans such as dermatan sulfate.

Non-enzymatic recombination of RNA: Ligation in loops.

BACKGROUND: While the RNA world hypothesis is widely accepted, it is still far from complete: the existence of self-replicating ribozyme, consisting of potentially hundreds of nucleotides, is a core assumption for the majority of RNA world models. The appearance of such long RNA molecules under prebiotic conditions is not self-evident. Recombination seems to be a plausible way of creating RNA diversity, resulting in the appearance of functional RNAs, capable of self-replicating. METHODS: We report here on the study of recombination process modelled with two 96 nts RNA fragments. Detection of recombination products was performed with RT-PCR followed by TA-cloning and Sanger sequencing. RESULTS: A wide range of recombinant products was detected. We found that (i) the most efficient ligation was observed for RNA species forming bulges or internal loops, with ligation partners located within the loop; (ii) a strong preference was observed for formation of a few types of major products with a large variety of minor products; (iii) ligation could occur with participation of either 2',3'-cyclophosphate or 5'-ppp; (iv) the presence of key reaction components, i.e. 5'ppp-RNAs, enabled the formation of additional types of product; (v) molecular dynamics simulations of one of the most abundant products suggests that the ligation results in a preferable formation of 2'-5'- rather than 3'-5'-linkages. CONCLUSIONS: The study demonstrates regularities of new RNA molecules formation with non-enzymatic recombination process. GENERAL SIGNIFICANCE: Our findings provide new data supporting the RNA World hypothesis and show the way of new RNA sequences emergence under prebiotic conditions.

Evaluation of analogues of furan-amidines as inhibitors of NQO2.

Inhibitors of the enzyme NQO2 (NRH: quinone oxidoreductase 2) are of potential use in cancer chemotherapy and malaria. We have previously reported that non-symmetrical furan amidines are potent inhibitors of NQO2 and here novel analogues are evaluated. The furan ring has been changed to other heterocycles (imidazole, N-methylimidazole, oxazole, thiophene) and the amidine group has been replaced with imidate, reversed amidine, N-arylamide and amidoxime to probe NQO2 activity, improve solubility and decrease basicity of the lead furan amidine. All compounds were fully characterised spectroscopically and the structure of the unexpected product N-hydroxy-4-(5-methyl-4-phenylfuran-2-yl)benzamidine was established by X-ray crystallography. The analogues were evaluated for inhibition of NQO2, which showed lower activity than the lead furan amidine. The observed structure-activity relationship for the furan-amidine series with NQO2 was rationalized by preliminary molecular docking and binding mode analysis. In addition, the oxazole-amidine analogue inhibited the growth of Plasmodium falciparum with an IC50 value of 0.3 µM.

Approximate quantum chemical methods for modelling carbohydrate conformation and aromatic interactions: ß-cyclodextrin and its adsorption on a single-layer graphene sheet.

Noncovalent functionalization of graphene by carbohydrates such as ß-cyclodextrin (ßCD) has the potential to improve graphene dispersibility and its use in biomedical applications. Here we explore the ability of approximate quantum chemical methods to accurately model ßCD conformation and its interaction with graphene. We find that DFTB3, SCC-DFTB and PM3CARB-1 methods provide the best agreement with density functional theory (DFT) in calculation of relative energetics of gas-phase ßCD conformers; however, the remaining NDDO-based approaches we considered underestimate the stability of the trans,gauche vicinal diol conformation. This diol orientation, corresponding to a clockwise hydrogen bonding arrangement in the glucosyl residue of ßCD, is present in the lowest energy ßCD conformer. Consequently, for adsorption on graphene of clockwise or counterclockwise hydrogen bonded forms of ßCD, calculated with respect to this unbound conformer, the DFTB3 method provides closer agreement with DFT values than PM7 and PM6-DH2 approaches. These findings suggest approximate quantum chemical methods as potentially useful tools to guide the design of carbohydrate-graphene interactions, but also highlights the specific challenge to NDDO-based methods in capturing the relative energetics of carbohydrate hydrogen bond networks.

Free Energy Calculations using a Swarm-Enhanced Sampling Molecular Dynamics Approach.

Free energy simulations are an established computational tool in modelling chemical change in the condensed phase. However, sampling of kinetically distinct substates remains a challenge to these approaches. As a route to addressing this, we link the methods of thermodynamic integration (TI) and swarm-enhanced sampling molecular dynamics (sesMD), where simulation replicas interact cooperatively to aid transitions over energy barriers. We illustrate the approach by using alchemical alkane transformations in solution, comparing them with the multiple independent trajectory TI (IT-TI) method. Free energy changes for transitions computed by using IT-TI grew increasingly inaccurate as the intramolecular barrier was heightened. By contrast, swarm-enhanced sampling TI (sesTI) calculations showed clear improvements in sampling efficiency, leading to more accurate computed free energy differences, even in the case of the highest barrier height. The sesTI approach, therefore, has potential in addressing chemical change in systems where conformations exist in slow exchange.

Molecular Dynamics of Cellulose Amphiphilicity at the Graphene-Water Interface.

Molecular dynamics (MD) simulations have been applied to study the interactions between hydrophobic and hydrophilic faces of ordered cellulose chains and a single layer of graphene in explicit aqueous solvent. The hydrophobic cellulose face is predicted to form a stable complex with graphene. This interface remains solvent-excluded over the course of simulations; the cellulose chains contacting graphene in general preserve intra- and interchain hydrogen bonds and a tg orientation of hydroxymethyl groups. Greater flexibility is observed in the more solvent-exposed cellulose chains of the complex. By contrast, the hydrophilic face of cellulose exhibits progressive rearrangement over the course of MD simulations, as it seeks to present its hydrophobic face, with disrupted intra- and interchain hydrogen bonding; residue twisting to form CH-π interactions with graphene; and partial permeation of water. This transition is also accompanied by a more favorable cellulose-graphene adhesion energy as predicted at the PM6-DH2 level of theory. The stability of the cellulose-graphene hydrophobic interface in water exemplifies the amphiphilicity of cellulose and provides insight into favored interactions within graphene-cellulose materials. Furthermore, partial permeation of water between exterior cellulose chains may indicate potential in addressing cellulose recalcitrance.

Exploring protein kinase conformation using swarm-enhanced sampling molecular dynamics.

Protein plasticity, while often linked to biological function, also provides opportunities for rational design of selective and potent inhibitors of their function. The application of computational methods to the prediction of concealed protein concavities is challenging, as the motions involved can be significant and occur over long time scales. Here we introduce the swarm-enhanced sampling molecular dynamics (sesMD) method as a tool to improve sampling of conformational landscapes. In this approach, a swarm of replica simulations interact cooperatively via a set of pairwise potentials incorporating attractive and repulsive components. We apply the sesMD approach to explore the conformations of the DFG motif in the protein p38α mitogen-activated protein kinase. In contrast to multiple MD simulations, sesMD trajectories sample a range of DFG conformations, some of which map onto existing crystal structures. Simulated structures intermediate between the DFG-in and DFG-out conformations are predicted to have druggable pockets of interest for structure-based ligand design.

Brazilin is a natural product inhibitor of the NLRP3 inflammasome.

Excessive or aberrant NLRP3 inflammasome activation has been implicated in the progression and initiation of many inflammatory conditions; however, currently no NLRP3 inflammasome inhibitors have been approved for therapeutic use in the clinic. Here we have identified that the natural product brazilin effectively inhibits both priming and activation of the NLRP3 inflammasome in cultured murine macrophages, a human iPSC microglial cell line and in a mouse model of acute peritoneal inflammation. Through computational modeling, we predict that brazilin can adopt a favorable binding pose within a site of the NLRP3 protein which is essential for its conformational activation. Our results not only encourage further evaluation of brazilin as a therapeutic agent for NLRP3-related inflammatory diseases, but also introduce this small-molecule as a promising scaffold structure for the development of derivative NLRP3 inhibitor compounds.

Insights into the activity and specificity of Trypanosoma cruzi trans-sialidase from molecular dynamics simulations.

Trypanosoma cruzitrans-sialidase (TcTS), which catalyzes the transfer or hydrolysis of terminal sialic acid residues, is crucial to the development and proliferation of the T. cruzi parasite and thus has emerged as a potential drug target for the treatment of Chagas disease. We here probe the origin of the observed preference for the transfer reaction over hydrolysis where the substrate for TcTS is the natural sialyl donor (represented in this work by sialyllactose). Thus, acceptor lactose preferentially attacks the sialyl-enyzme intermediate rather than water. We compare this with the weaker preference for such transfer shown by a synthetic donor substrate, 4-methylumbelliferyl α-d-acetylneuraminide. For this reason, we conducted molecular dynamics simulations of TcTS following its sialylation by the substrate to examine the behavior of the asialyl leaving group by the protein. These simulations indicate that, where lactose is released, this leaving group samples well-defined interactions in the acceptor site, some of which are mediated by localized water molecules; also, the extent of the opening of the acceptor site to solvent is reduced as compared with those of unliganded forms of TcTS. However, where there is release of 4-methylumbelliferone, this leaving group explores a range of transient poses; surrounding active site water is also more disordered. The acceptor site explores more open conformations, similar to the case in which the 4-methylumbelliferone is absent. Thus, the predicted solvent accessibility of sialylated TcTS is increased when 4-methylumbelliferyl α-d-acetylneuraminide is the substrate compared to sialyllactose; this in turn is likely to contribute to a greater propensity for hydrolysis of the covalent intermediate. These computational simulations, which suggest that protein flexibility has a role in the transferase/sialidase activity of TcTS, have the potential to aid in the design of anti-Chagas inhibitors effective against this neglected tropical disease.

Cellulose Iß microfibril interaction with pristine graphene in water: Effects of amphiphilicity by molecular simulation.

Graphene-cellulose interactions have considerable potential in the development of new materials. In previous computational work (Biomacromolecules2016, 16, 1771), we predicted that the model 100 hydrophobic surface of cellulose interacted favourably with pristine graphene in aqueous solution molecular dynamics simulations; conversely, a model of the hydrophilic 010 surface of cellulose exhibited progressive rearrangement to present a more hydrophobic face with the graphene, with weakened hydrogen bonds between cellulose chains and partial permeation of water. Here, we extend this work by simulating the interaction in aqueous solution of the amphiphilic 110 surface of a cellulose Iß microfibril model, comprising 36 chains of 40 glucosyl residues, with an infinite sheet of pristine graphene. This face of the microfibril is of intermediate hydrophilicity and progressively associates with graphene over replicate simulations. As cellulose chains adhere to the graphene surface, forming interactions via its CH and OH groups, we observe a degree of local and global untwisting of the microfibril. Complementary rippling of the graphene surface is also observed, as it adapts to interaction with the microfibril. This adsorption process is accompanied by increased exclusion of water between cellulose and graphene although some water localises between chains at the immediate interface. The predicted propensity of a cellulose microfibril to adsorb spontaneously on the graphene surface, with mutual structural accommodation, highlights the amphiphilic nature of cellulose and the types of interactions that can be harnessed to design new graphene-carbohydrate biopolymer materials.

Probing the effect of NEK7 and cofactor interactions on dynamics of NLRP3 monomer using molecular simulation.

The NLRP3 inflammasome is a cytoplasmic complex that regulates the activation of inflammatory cytokines and, given its implication in a range of diseases, is an important therapeutic target. The cofactor ATP and the centrosomal kinase NEK7 are important for NLRP3 activation. Here we have constructed and simulated computational models of full-length monomeric NLRP3 to shed light on the importance of NEK7 and cofactor interactions for its conformation and dynamics in aqueous solution. We find that molecular dynamics simulation reproduces well the features of the recently published cryo-EM structure of the ADP-bound NLRP3-NEK7 complex; on the removal of NEK7, the NLRP3 molecule adopts a more compact closed form during simulations. Replacement of ADP by ATP promotes a rearrangement of hydrogen-bonding interactions, domain interfaces, and a degree of opening of the NLRP3 conformation. We also examine the dynamics of an acidic loop of the LRR domain of NLRP3, which samples in a region observed in the NEK7-bound cryo-EM structure but not in an oligomeric form of inactive NLRP3. During the molecular dynamics simulations of NLRP3, we find some plasticity in its topology that suggests access routes for ATP to the cofactor pocket not immediately evident from the existing NEK7-bound cryo-EM structure. These computed dynamical trajectories of NLRP3 provide insight into coordinates of deformation that may be key for cofactor binding and inflammasome activation.