TorsionNet: A Deep Neural Network to Rapidly Predict Small-Molecule Torsional Energy Profiles with the Accuracy of Quantum Mechanics.

Fast and accurate assessment of small-molecule dihedral energetics is crucial for molecular design and optimization in medicinal chemistry. Yet, accurate prediction of torsion energy profiles remains challenging as the current molecular mechanics (MM) methods are limited by insufficient coverage of drug-like chemical space and accurate quantum mechanical (QM) methods are too expensive. To address this limitation, we introduce TorsionNet, a deep neural network (DNN) model specifically developed to predict small-molecule torsion energy profiles with QM-level accuracy. We applied active learning to identify nearly 50k fragments (with elements H, C, N, O, F, S, and Cl) that maximized the coverage of our corporate compound library and leveraged massively parallel cloud computing resources for density functional theory (DFT) torsion scans of these fragments, generating a training data set of 1.2 million DFT energies. After training TorsionNet on this data set, we obtain a model that can rapidly predict the torsion energy profile of typical drug-like fragments with DFT-level accuracy. Importantly, our method also provides an uncertainty estimate for the predicted profiles without any additional calculations. In this report, we show that TorsionNet can accurately identify the preferred dihedral geometries observed in crystal structures. Our TorsionNet-based analysis of a diverse set of protein-ligand complexes with measured binding affinity shows a strong association between high ligand strain and low potency. We also present practical applications of TorsionNet that demonstrate how consideration of DNN-based strain energy leads to substantial improvement in existing lead discovery and design workflows. TorsionNet500, a benchmark data set comprising 500 chemically diverse fragments with DFT torsion profiles (12k MM- and DFT-optimized geometries and energies), has been created and is made publicly available.

Coupling Mixed Mode Chromatography/ESI Negative MS Detection with Message-Passing Neural Network Modeling for Enhanced Metabolome Coverage and Structural Identification.

A key unmet need in metabolomics continues to be the specific, selective, accurate detection of traditionally difficult to retain molecules including simple sugars, sugar phosphates, carboxylic acids, and related amino acids. Designed to retain the metabolites of central carbon metabolism, this Mixed Mode (MM) chromatography applies varied pH, salt concentration and organic content to a positively charged quaternary amine polyvinyl alcohol stationary phase. This MM method is capable of separating glucose from fructose, and four hexose monophosphates a single chromatographic run. Coupled to a QExactive Orbitrap Mass Spectrometer with negative ESI, linearity, LLOD, %CV, and mass accuracy were assessed using 33 metabolite standards. The standards were linear on average >3 orders of magnitude (R2 > 0.98 for 30/33) with LLOD < 1 pmole (26/33), median CV of 12% over two weeks, and median mass accuracy of 0.49 ppm. To assess the breadth of metabolome coverage and better define the structural elements dictating elution, we injected 607 unique metabolites and determined that 398 are well retained. We then split the dataset of 398 documented RTs into training and test sets and trained a message-passing neural network (MPNN) to predict RT from a featurized heavy atom connectivity graph. Unlike traditional QSAR methods that utilize hand-crafted descriptors or pre-defined structural keys, the MPNN aggregates atomic features across the molecular graph and learns to identify molecular subgraphs that are correlated with variations in RTs. For sugars, sugar phosphates, carboxylic acids, and isomers, the model achieves a predictive RT error of <2 min on 91%, 50%, 77%, and 72% of held-out compounds from these subsets, with overall root mean square errors of 0.11, 0.34, 0.18, and 0.53 min, respectively. The model was then applied to rank order metabolite IDs for molecular features altered by GLS2 knockout in mouse primary hepatocytes.

Actuation of Janus Emulsion Droplets via Optothermally Induced Marangoni Forces.

Microscale Janus emulsions represent a versatile material platform for dynamic refractive, reflective, and light-emitting optical components. Here, we present a mechanism for droplet actuation that exploits thermocapillarity. Using optically induced thermal gradients, an interfacial tension differential is generated across the surfactant-free internal capillary interface of Janus droplets. The interfacial tension differential causes droplet-internal Marangoni flows and a net torque, resulting in a predictable and controllable reorientation of the droplets. The effect can be quantitatively described with a simple model that balances gravitational and thermal torques. Occurring in small thermal gradients, these optothermally induced Marangoni dynamics represent a promising mechanism for controlling droplet-based micro-optical components.

Are 2D fingerprints still valuable for drug discovery?

Recently, molecular fingerprints extracted from three-dimensional (3D) structures using advanced mathematics, such as algebraic topology, differential geometry, and graph theory have been paired with efficient machine learning, especially deep learning algorithms to outperform other methods in drug discovery applications and competitions. This raises the question of whether classical 2D fingerprints are still valuable in computer-aided drug discovery. This work considers 23 datasets associated with four typical problems, namely protein-ligand binding, toxicity, solubility and partition coefficient to assess the performance of eight 2D fingerprints. Advanced machine learning algorithms including random forest, gradient boosted decision tree, single-task deep neural network and multitask deep neural network are employed to construct efficient 2D-fingerprint based models. Additionally, appropriate consensus models are built to further enhance the performance of 2D-fingerprint-based methods. It is demonstrated that 2D-fingerprint-based models perform as well as the state-of-the-art 3D structure-based models for the predictions of toxicity, solubility, partition coefficient and protein-ligand binding affinity based on only ligand information. However, 3D structure-based models outperform 2D fingerprint-based methods in complex-based protein-ligand binding affinity predictions.

Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein-Ligand Complexes using ab Initio Calculations.

The energetics of rotation around single bonds (torsions) is a key determinant of the three-dimensional shape that druglike molecules adopt in solution, the solid state, and in different biological environments, which in turn defines their unique physical and pharmacological properties. Therefore, accurate characterization of torsion angle preference and energetics is essential for the success of computational drug discovery and design. Here, we analyze torsional strain in crystal structures of druglike molecules in Cambridge structure database (CSD) and bioactive ligand conformations in protein data bank (PDB), expressing the total strain energy as a sum of strain energy from constituent rotatable bonds. We utilized cloud computing to generate torsion scan profiles of a very large collection of chemically diverse neutral fragments at DFT(B3LYP)/6-31G*//6-31G** or DFT(B3LYP)/6-31+G*//6-31+G** (for sulfur-containing molecule). With the data generated from these ab initio calculations, we performed rigorous analysis of strain due to deviation of observed torsion angles relative to their ideal gas-phase geometries. Contrary to the previous studies based on molecular mechanics, we find that in the crystalline state, molecules generally adopt low-strain conformations, with median per-torsion strain energy in CSD and PDB under one-tenth and one-third of a kcal/mol, respectively. However, for a small fraction (<5%) of motifs, external effects such as steric hindrance and hydrogen bonds result in strain penalty exceeding 2.5 kcal/mol. We find that due to poor quality of PDB structures in general, bioactive structures tend to have higher torsional strain compared to small-molecule crystal conformations. However, in the absence of structural fitting artifacts in PDB structures, protein-induced strain in bioactive conformations is quantitatively similar to those due to the packing forces in small-molecule crystal structures. This analysis allows us to establish strain energy thresholds to help identify biologically relevant conformers in a given ensemble. The work presented here is the most comprehensive study to date that demonstrates the utility and feasibility of gas-phase quantum mechanics (QM) calculations to study conformational preference and energetics of drug-size molecules. Potential applications of this study in computational lead discovery and structure-based design are discussed.

Molecular Transformer unifies reaction prediction and retrosynthesis across pharma chemical space.

Predicting how a complex molecule reacts with different reagents, and how to synthesise complex molecules from simpler starting materials, are fundamental to organic chemistry. We show that an attention-based machine translation model - Molecular Transformer - tackles both reaction prediction and retrosynthesis by learning from the same dataset. Reagents, reactants and products are represented as SMILES text strings. For reaction prediction, the model "translates" the SMILES of reactants and reagents to product SMILES, and the converse for retrosynthesis. Moreover, a model trained on publicly available data is able to make accurate predictions on proprietary molecules extracted from pharma electronic lab notebooks, demonstrating generalisability across chemical space. We expect our versatile framework to be broadly applicable to problems such as reaction condition prediction, reagent prediction and yield prediction.

Predicting Surface Tensions of Surfactant Solutions from Statistical Mechanics.

The importance of surfactants to various industries necessitates a predictive understanding of their surface tension and adsorption behavior in terms of molecular characteristics. Previous models are highly empirical, require fitting parameters, and have limited applicability at various temperatures. Here, we provide a surface tension model based on statistical mechanics that (1) is thermodynamically consistent, (2) provides a higher predictive power, wherein surface tension can be calculated for any tail length, concentration, and temperature from molecular parameters, and (3) provides a physical understanding of the important molecular interactions at play. This model is applicable to both nonionic and ionic surfactants, where the effects of the electric double layer have been taken into account in the latter case. For nonionic surfactants, we were able to extend our model to predict dynamic surface tension as well. We have validated our model with tensiometry experiments for various surfactants, concentrations, and temperatures. In addition, we have validated our model with a diverse set of literature data, wherein agreement within a few mN M-1 and a correct prediction of phase change behavior is shown. The model could enable a more informed design of surfactant systems and serve as the theoretical basis for theory on more complex surfactant systems such as mixtures.

Combined Molecular Dynamics Simulation-Molecular-Thermodynamic Theory Framework for Predicting Surface Tensions.

A molecular modeling approach is presented with a focus on quantitative predictions of the surface tension of aqueous surfactant solutions. The approach combines classical Molecular Dynamics (MD) simulations with a molecular-thermodynamic theory (MTT) [ Y. J. Nikas, S. Puvvada, D. Blankschtein, Langmuir 1992 , 8 , 2680 ]. The MD component is used to calculate thermodynamic and molecular parameters that are needed in the MTT model to determine the surface tension isotherm. The MD/MTT approach provides the important link between the surfactant bulk concentration, the experimental control parameter, and the surfactant surface concentration, the MD control parameter. We demonstrate the capability of the MD/MTT modeling approach on nonionic alkyl polyethylene glycol surfactants at the air-water interface and observe reasonable agreement of the predicted surface tensions and the experimental surface tension data over a wide range of surfactant concentrations below the critical micelle concentration. Our modeling approach can be extended to ionic surfactants and their mixtures with both ionic and nonionic surfactants at liquid-liquid interfaces.

CO2-Reactive Ionic Liquid Surfactants for the Control of Colloidal Morphology.

This article reports on a new class of stimuli-responsive surfactant generated from commercially available amphiphiles such as dodecyltrimethylammmonium bromide (DTAB) by substitution of the halide counterion with counterions such as 2-cyanopyrrolide, 1,2,3-triazolide, and L-proline that complex reversibly with CO2. Through a combination of small-angle neutron scattering (SANS), electrical conductivity measurements, thermal gravimetric analysis, and molecular dynamics simulations, we show how small changes in charge reorganization and counterion shape and size induced by complexation with CO2 allow for fine-tunability of surfactant properties. We then use these findings to demonstrate a range of potential practical uses, from manipulating microemulsion droplet morphology to controlling micellar and vesicular aggregation. In particular, we focus on the binding of these surfactants to DNA and the reversible compaction of surfactant-DNA complexes upon alternate bubbling of the solution with CO2 and N2.

Reconfigurable and responsive droplet-based compound micro-lenses.

Micro-scale optical components play a crucial role in imaging and display technology, biosensing, beam shaping, optical switching, wavefront-analysis, and device miniaturization. Herein, we demonstrate liquid compound micro-lenses with dynamically tunable focal lengths. We employ bi-phase emulsion droplets fabricated from immiscible hydrocarbon and fluorocarbon liquids to form responsive micro-lenses that can be reconfigured to focus or scatter light, form real or virtual images, and display variable focal lengths. Experimental demonstrations of dynamic refractive control are complemented by theoretical analysis and wave-optical modelling. Additionally, we provide evidence of the micro-lenses' functionality for two potential applications-integral micro-scale imaging devices and light field display technology-thereby demonstrating both the fundamental characteristics and the promising opportunities for fluid-based dynamic refractive micro-scale compound lenses.

Understanding the colloidal dispersion stability of 1D and 2D materials: Perspectives from molecular simulations and theoretical modeling.

The colloidal dispersion stability of 1D and 2D materials in the liquid phase is critical for scalable nano-manufacturing, chemical modification, composites production, and deployment as conductive inks or nanofluids. Here, we review recent computational and theoretical studies carried out by our group to model the dispersion stability of 1D and 2D materials, including single-walled carbon nanotubes, graphene, and graphene oxide in aqueous surfactant solutions or organic solvents. All-atomistic (AA) molecular dynamics (MD) simulations can probe the molecular level details of the adsorption morphology of surfactants and solvents around these materials, as well as quantify the interaction energy between the nanomaterials mediated by surfactants or solvents. Utilizing concepts from reaction kinetics and diffusion, one can directly predict the rate constants for the aggregation kinetics and dispersion life times using MD outputs. Furthermore, the use of coarse-grained (CG) MD simulations allows quantitative prediction of surfactant adsorption isotherms. Combined with the Poisson-Boltzmann equation, the Langmuir isotherm, and the DLVO theory, one can directly use CGMD outputs to: (i) predict electrostatic potentials around the nanomaterial, (ii) correlate surfactant surface coverages with surfactant concentrations in the bulk dispersion medium, and (iii) determine energy barriers against coagulation. Finally, we discuss challenges associated with studying emerging 2D materials, such as, hexagonal boron nitride (h-BN), phosphorene, and transition metal dichalcogenides (TMDCs), including molybdenum disulfide (MoS2). An outlook is provided to address these challenges with plans to develop force-field parameters for MD simulations to enable predictive modeling of emerging 2D materials in the liquid phase.

Dominance of Dispersion Interactions and Entropy over Electrostatics in Determining the Wettability and Friction of Two-Dimensional MoS2 Surfaces.

The existence of partially ionic bonds in molybdenum disulfide (MoS2), as opposed to covalent bonds in graphene, suggests that polar (electrostatic) interactions should influence the interfacial behavior of two-dimensional MoS2 surfaces. In this work, using molecular dynamics simulations, we show that electrostatic interactions play a negligible role in determining not only the equilibrium contact angle on the MoS2 basal plane, which depends solely on the total interaction energy between the surface and the liquid, but also the friction coefficient and the slip length, which depend on the spatial variations in the interaction energy. While the former is found to result from the exponential decay of the electric potential above the MoS2 surface, the latter results from the trilayered sandwich structure of the MoS2 monolayer, which causes the spatial variations in dispersion interactions in the lateral direction to dominate over those in electrostatic interactions in the lateral direction. Further, we show that the nonpolarity of MoS2 is specific to the two-dimensional basal plane of MoS2 and that other planes (e.g., the zigzag plane) in MoS2 are polar with respect to interactions with water, thereby illustrating the role of edge effects, which could be important in systems involving vacancies or nanopores in MoS2. Finally, we simulate the temperature dependence of the water contact angle on MoS2 to show that the inclusion of entropy, which has been neglected in recent mean-field theories, is essential in determining the wettability of MoS2. Our findings reveal that the basal planes in graphene and MoS2 are unexpectedly similar in terms of their interfacial behavior.

Liquid-Phase Exfoliation of Phosphorene: Design Rules from Molecular Dynamics Simulations.

The liquid-phase exfoliation of phosphorene, the two-dimensional derivative of black phosphorus, in the solvents dimethyl sulfoxide (DMSO), dimethylformamide (DMF), isopropyl alcohol, N-methyl-2-pyrrolidone, and N-cyclohexyl-2-pyrrolidone is investigated using three molecular-scale "computer experiments". We modeled solvent-phosphorene interactions using an atomistic force field, based on ab initio calculations and lattice dynamics, that accurately reproduces experimental mechanical properties. We probed solvent molecule ordering at phosphorene/solvent interfaces and discovered that planar molecules such as N-methyl-2-pyrrolidone preferentially orient parallel to the interface. We subsequently measured the energy required to peel a single phosphorene monolayer from a stack of black phosphorus and analyzed the role of "wedges" of solvent molecules intercalating between phosphorene sheets in initiating exfoliation. The exfoliation efficacy of a solvent is enhanced when either molecular planarity "sharpens" this molecular wedge or strong phosphorene-solvent adhesion stabilizes the newly exposed phosphorene surfaces. Finally, we examined the colloidal stability of exfoliated flakes by simulating their aggregation and showed that dispersion is favored when the cohesive energy between the molecules in the solvent monolayer confined between the phosphorene sheets is high (as with DMSO) and is hindered when the adhesion between these molecules and phosphorene is strong; the molecular planarity in solvents like DMF enhances the cohesive energy. Our results are consistent with, and provide a molecular context for, experimental exfoliation studies of phosphorene and other layered solids, and our molecular insights into the significant role of solvent molecular geometry and ordering should complement prevalent solubility-parameter-based approaches in establishing design rules for effective nanomaterial exfoliation media.

Understanding Miltefosine-Membrane Interactions Using Molecular Dynamics Simulations.

Coarse-grained molecular dynamics simulations are used to calculate the free energies of transfer of miltefosine, an alkylphosphocholine anticancer agent, from water to lipid bilayers to study its mechanism of interaction with biological membranes. We consider bilayers containing lipids with different degrees of unsaturation: dipalmitoylphosphatidylcholine (DPPC, saturated, containing 0%, 10%, and 30% cholesterol), dioleoylphosphatidylcholine (DOPC, diunsaturated), palmitoyloleoylphosphatidylcholine (POPC, monounsaturated), diarachidonoylphosphatidylcholine (DAPC, polyunsaturated), and dilinoleylphosphatidylcholine (DUPC, polyunsaturated). These free energies, calculated using umbrella sampling, were used to compute the partition coefficients (K) of miltefosine between water and the lipid bilayers. The K values for the bilayers relative to that of pure DPPC were found to be 5.3 (DOPC), 7.0 (POPC), 1.0 (DAPC), 2.2 (DUPC), 14.9 (10% cholesterol), and 76.2 (30% cholesterol). Additionally, we calculated the free energy of formation of miltefosine-cholesterol complexes by pulling the surfactant laterally in the DPPC + 30% cholesterol system. The free energy profile that we obtained provides further evidence that miltefosine tends to associate with cholesterol and has a propensity to partition into lipid rafts. We also quantified the kinetics of the transport of miltefosine through the various bilayers by computing permeance values. The highest permeance was observed in DUPC bilayers (2.28 × 10(-2) m/s) and the lowest permeance in the DPPC bilayer with 30% cholesterol (1.10 × 10(-7) m/s). Our simulation results show that miltefosine does indeed interact with lipid rafts, has a higher permeability in polyunsaturated, loosely organized bilayers, and has higher flip-flop rates in specific regions of cellular membranes.

Dynamically reconfigurable complex emulsions via tunable interfacial tensions.

Emulsification is a powerful, well-known technique for mixing and dispersing immiscible components within a continuous liquid phase. Consequently, emulsions are central components of medicine, food and performance materials. Complex emulsions, including Janus droplets (that is, droplets with faces of differing chemistries) and multiple emulsions, are of increasing importance in pharmaceuticals and medical diagnostics, in the fabrication of microparticles and capsules for food, in chemical separations, in cosmetics, and in dynamic optics. Because complex emulsion properties and functions are related to the droplet geometry and composition, the development of rapid, simple fabrication approaches allowing precise control over the droplets' physical and chemical characteristics is critical. Significant advances in the fabrication of complex emulsions have been made using a number of procedures, ranging from large-scale, less precise techniques that give compositional heterogeneity using high-shear mixers and membranes, to small-volume but more precise microfluidic methods. However, such approaches have yet to create droplet morphologies that can be controllably altered after emulsification. Reconfigurable complex liquids potentially have great utility as dynamically tunable materials. Here we describe an approach to the one-step fabrication of three- and four-phase complex emulsions with highly controllable and reconfigurable morphologies. The fabrication makes use of the temperature-sensitive miscibility of hydrocarbon, silicone and fluorocarbon liquids, and is applied to both the microfluidic and the scalable batch production of complex droplets. We demonstrate that droplet geometries can be alternated between encapsulated and Janus configurations by varying the interfacial tensions using hydrocarbon and fluorinated surfactants including stimuli-responsive and cleavable surfactants. This yields a generalizable strategy for the fabrication of multiphase emulsions with controllably reconfigurable morphologies and the potential to create a wide range of responsive materials.

2D equation-of-state model for corona phase molecular recognition on single-walled carbon nanotube and graphene surfaces.

Corona phase molecular recognition (CoPhMoRe) has been recently introduced as a means of generating synthetic molecular recognition sites on nanoparticle surfaces. A synthetic heteropolymer is adsorbed and confined to the surface of a nanoparticle, forming a corona phase capable of highly selective molecular recognition due to the conformational imposition of the particle surface on the polymer. In this work, we develop a computationally predictive model for analytes adsorbing onto one type of polymer corona phase composed of hydrophobic anchors on hydrophilic loops around a single-walled carbon nanotube (SWCNT) surface using a 2D equation of state that takes into consideration the analyte-polymer, analyte-nanoparticle, and polymer-nanoparticle interactions using parameters determined independently from molecular simulation. The SWCNT curvature is found to contribute weakly to the overall interaction energy, exhibiting no correlation for three of the corona phases considered, and differences of less than 5% and 20% over a larger curvature range for two other corona phases, respectively. Overall, the resulting model for this anchor-loop CoPhMoRe is able to correctly predict 83% of an experimental 374 analyte-polymer library, generating experimental fluorescence responses within 20% error of the experimental values. The modeling framework presented here represents an important step forward in the design of suitable polymers to target specific analytes.