Deep generative design with 3D pharmacophoric constraints

Generative models have increasingly been proposed as a solution to the molecular design problem. However, it has proved challenging to control the design process or incorporate prior knowledge, limiting their practical use in drug discovery. In particular, generative methods have made limited use of three-dimensional (3D) structural information even though this is critical to binding. This work describes a method to incorporate such information and demonstrates the benefit of doing so. We combine an existing graph-based deep generative model, DeLinker, with a convolutional neural network to utilise physically-meaningful 3D representations of molecules and target pharmacophores. We apply our model, DEVELOP, to both linker and R-group design, demonstrating its suitability for both hit-to-lead and lead optimisation. The 3D pharmacophoric information results in improved generation and allows greater control of the design process. In multiple large-scale evaluations, we show that including 3D pharmacophoric constraints results in substantial improvements in the quality of generated molecules. On a challenging test set derived from PDBbind, our model improves the proportion of generated molecules with high 3D similarity to the original molecule by over 300%. In addition, DEVELOP recovers 10× more of the original molecules compared to the baseline DeLinker method. Our approach is general-purpose, readily modifiable to alternate 3D representations, and can be incorporated into other generative frameworks. Code is available at https://github.com/oxpig/DEVELOP.

A novel deep generative model combines convolution and graph neural networks to allow 3D-aware molecular design. We show how 3D pharmacophoric information can be incorporated into generative models and apply our model to both linker and R-group design.     

Structure-based de novo drug design using 3D deep generative models

Deep generative models are attracting much attention in the field of de novo molecule design. Compared to traditional methods, deep generative models can be trained in a fully data-driven way with little requirement for expert knowledge. Although many models have been developed to generate 1D and 2D molecular structures, 3D molecule generation is less explored, and the direct design of drug-like molecules inside target binding sites remains challenging. In this work, we introduce DeepLigBuilder, a novel deep learning-based method for de novo drug design that generates 3D molecular structures in the binding sites of target proteins. We first developed Ligand Neural Network (L-Net), a novel graph generative model for the end-to-end design of chemically and conformationally valid 3D molecules with high drug-likeness. Then, we combined L-Net with Monte Carlo tree search to perform structure-based de novo drug design tasks. In the case study of inhibitor design for the main protease of SARS-CoV-2, DeepLigBuilder suggested a list of drug-like compounds with novel chemical structures, high predicted affinity, and similar binding features to those of known inhibitors. The current version of L-Net was trained on drug-like compounds from ChEMBL, which could be easily extended to other molecular datasets with desired properties based on users'' demands and applied in functional molecule generation. Merging deep generative models with atomic-level interaction evaluation, DeepLigBuilder provides a state-of-the-art model for structure-based de novo drug design and lead optimization.

DeepLigBuilder, a novel deep generative model for structure-based de novo drug design, directly generates 3D structures of drug-like compounds in the target binding site.     

Exploration of 3D activity cliffs on the basis of compound binding modes and comparison of 2D and 3D cliffs

Activity cliffs are formed by pairs or groups of structurally similar compounds having large differences in potency and are focal points of structure-activity relationship (SAR) analysis. The choice of molecular representations is a critically important aspect of activity cliffs analysis. Thus far, activity cliffs have predominantly been defined on the basis of molecular graph or fingerprint representations. Herein we introduce 3D activity cliffs derived from comparisons of experimentally determined compound binding modes. The analysis of 3D activity cliffs is generally applicable to target proteins for which structures of multiple ligand complexes are available. For two popular targets, β-secretase 1 (BACE1) and factor Xa (FXa), public domain X-ray structures with bound inhibitors were collected. Crystallographic binding modes of inhibitors were systematically compared using a 3D similarity method taking conformational, positional, and atomic property differences into account. In addition, standard 2D similarity relationships were also determined. SAR information associated with individual compounds substantially changed when either bioactive conformations or 2D molecular graphs were used for similarity evaluation. 3D activity cliffs were identified for BACE1 and FXa inhibitor sets and systematically compared to 2D cliffs. It was found that less than 40% of 3D activity cliffs were conserved when 2D similarity was applied. The limited conservation of 3D and 2D cliffs provides further evidence for the strong molecule representation dependence of activity cliffs. Moreover, 3D cliffs represent a new class of activity cliffs that convey SAR information in ways that differ from graph-based similarity measures. In cases where sufficient structural information is available, the comparison of 3D and 2D cliffs is expected to aid in SAR analysis and mapping of critical binding determinants.     

Calcium croconate and calcium oxalato-croconate complexes with 2D and 3D crystal structures

Two novel complexes of calcium croconate Ca(μ₃-C₅O₅)(H₂O)₃ (I) and calcium oxalato-croconate Ca₂(μ₄-C₅O₅)(μ₆-C₂O₄)(H₂O) (II) have been synthesized by soft chemical routes and their crystal structures have been solved from single-crystal data. Croconates I and II crystallize in orthorhombic space groups Pbca and Pnma, respectively. Their crystal structures evidence the versatile character of the croconate anion in the framework design. The calcium croconate I displays a 2D structure with corrugated sheets formed by croconate dianions and eightfold coordinated calcium atoms within [Ca₂O₁₄] dimers. The sheets develop eight-member apertures made of four dimers and four croconate dianions. The structure of II displays a 3D open-framework built from layers connected to each other by croconate dianions. This arrangement generates rectangular cavities free of water molecules. Seven- and eightfold coordinated calcium atoms with oxalate dianions constitute the layers. In both structures, the connecting modes of croconates are unusual. The thermal behaviour study has demonstrated that II is stable until 380 °C.     

Collaborative annotation of 3D crystallographic models

This paper describes the AnnoCryst system-a tool that was designed to enable authenticated collaborators to share online discussions about 3D crystallographic structures through the asynchronous attachment, storage, and retrieval of annotations. Annotations are personal comments, interpretations, questions, assessments, or references that can be attached to files, data, digital objects, or Web pages. The AnnoCryst system enables annotations to be attached to 3D crystallographic models retrieved from either private local repositories (e.g., Fedora) or public online databases (e.g., Protein Data Bank or Inorganic Crystal Structure Database) via a Web browser. The system uses the Jmol plugin for viewing and manipulating the 3D crystal structures but extends Jmol by providing an additional interface through which annotations can be created, attached, stored, searched, browsed, and retrieved. The annotations are stored on a standardized Web annotation server (Annotea), which has been extended to support 3D macromolecular structures. Finally, the system is embedded within a security framework that is capable of authenticating users and restricting access only to trusted colleagues.     

中药有效成分三维结构数据库的开发和研究

介绍了北京大学中药有效成分三维结构数据库软件系统的结构、功能及开发步骤。该数据库系统不仅仅提供6500个中草药有效成分的二维和三维结构以及其它各类相关信息,同时拥有功能强大的数据库查询、维护及分子表达系统。在该系统中,用户可以交互式地实现多种分子特征的查询以及二维子结构的查询。查询得到的分子可以直接在北京大学药物设计系统(PKUDDS)中进行三维结构的显示和分析。该数据库系统和我们科研组开发的北京大学药物设计系统以及中草药信息系统构成了完整的基于中药的药物设计系统。该系统已经用于NS3-NS4A蛋白酶抑制剂以及其它体系的研究并取得了很好的结果。     

Gas-phase H/D exchange of sodiated glycine oligomers with ND3: exchange kinetics do not reflect parent ion structures

H/D exchange is a method commonly used to probe molecular structure. The majority of studies in the gas phase have involved protonated molecular ions. The present study gives attention to molecular ions formed by coordination with a sodium ion. In particular, ND(3) is reacted with sodiated glycine oligomers, Gly(n)(), where n = 1-5, and the results are interpreted using density functional calculations. Experimentally, Gly(1)Na(+), Gly(4)Na(+), and Gly(5)Na(+) all undergo three fast exchanges with ND(3), while Gly(2)Na(+) and Gly(3)Na(+) undergo one fast and two slow exchanges with ND(3). The methyl esters Gly(3)OMeNa(+) and Gly(5)OMeNa(+) do not exchange with ND(3). In agreement with earlier experimental studies, theoretical calculations show that the lowest-energy conformers of the sodiated glycine oligomers are charge-solvated structures. Calculations further indicate that, in the process of H/D exchange with ND(3), sodiated monoglycine and tetraglycine adopt zwitterionic structures, sodiated diglycine adopts a salt-bridge form, and sodiated triglycine takes on an ion-stabilized ion pair form. Sodiated monoglycine and diglycine exchange via an onium-ion mechanism. The proposed exchange mechanisms require a carboxylic acid hydrogen to complete the exchange, which is in agreement with the experimental results showing that no exchange occurs with methyl ester glycine oligomers. These studies clearly demonstrate that, in the process of H/D exchange, noncovalent complexation of the exchange reagent provides the energy required to access intermediates structurally distinct from the parent ions. H/D exchange is facile for these intermediates. Contrary to the assumption often expressed in earlier studies, H/D exchange kinetics may not directly reflect ion structures.     

Supramolecular architecture of crystals of perfluorinated 3-alkylphthalides

The molecular and crystal structures of perfluoro-3-alkylphthalides (3-hydroxyperfluoro-3-methylphthalide and its hydrate, two polymorphs of 3-hydroxyperfluoro-3-ethylphthalide and 3-hydroxyperfluoro-3-isopropylphthalide) are determined by single crystal X-ray diffraction. In the crystals, the supramolecular O–Н…O=C synthon occurs leading, except the crystallohydrate, to the formation of C₁¹ (6) hydrogen bonded chains (supramolecular 1D motifs). According to the DFT/M06-2X/TZV calculations, the interaction energy of hydrogen bonded molecular pairs increases in this series, which can be explained by additional C=O…π, O…π, and C–F…π interactions.     

Dimension Engineering of Stimuli-Responsive 1D Molecular Crystals into Unusual 2D and 3D Zigzag Waveguides

We demonstrate an innovative technique to achieve organic 2D and 3D waveguides with peculiar shapes from an acicular, stimuli-responsive molecular crystal, (2Z,2′Z)-3,3′-(anthracene-9,10-diyl)bis(2-(3,5-bis(trifluoromethyl)phenylacrylonitrile), Ant-CF₃. The greenish-yellow fluorescent (FL) Ant-CF₃ molecular crystals exhibit laser power-dependent permanent mechanical bending in 2D and 3D. Investigation of a single-crystal using spatially-resolved Raman/FL/electron microscopy, and theoretical calculations revealed photothermal (Z,E)/(E,E) isomerization-assisted transition from crystalline to amorphous phase at the laser-exposed regions. This phenomenon facilitates the dimension engineering of a 1D crystal waveguide into 2D waveguide on a substrate or a 3D waveguide in free space. The bends can be used as interconnection points to couple different optical elements. The presented technique has broader implications in organic photonics and other crystal-related photonic technologies.     

ATS drugs molecular structure representation using refined 3D geometric moment invariants

The campaign against drug abuse is fought by all countries, most notably on ATS drugs. The identification process of ATS drugs depends heavily on its molecular structure. However, the process becomes more unreliable due to the introduction of new, sophisticated, and increasingly complex ATS molecular structures. Therefore, distinctive features of ATS drug molecular structure need to be accurately obtained. In this paper, two variants of refined 3D geometric moment invariants for ATS drug molecular structure representation are discussed. This paper is also meant for comparing the performance of these two variants. The comparison was conducted using drug chemical structures obtained from Isomer Design’s PiHKaL.info database for the ATS drugs, while non-ATS drugs are obtained randomly from ChemSpider database. The assessment highlights the best technique which is suitable to be further explored and improved in the future studies so that it is wholly attuned with ATS drug molecular similarity search domain.     

Molecular geometry directed Kagomé and honeycomb networks: toward two-dimensional crystal engineering

We present here the formation of a molecular Kagomé network within a two-dimensional (2D) crystal on a surface. This system provides a clear example of how, by design, molecular geometry can be expressed at the level of the 2D crystal lattice, leading to the formation of open networks. Key elements to control molecular network formation are core symmetry, location and orientation of interacting and connecting substituents, as well as symmetry matching between the networks and the surface.     

3D打印生物医用材料研究进展

3D打印(亦称增材制造)技术因其独特的材料成型优势,在组织工程、航空航天、汽车制造、以及电子工业等众多领域显示出巨大的应用潜力。然而,在实际生物医学应用中,3D打印生物器件和组织器官除了要求具有复杂的结构和优异的生物学性能外,其打印结构的表面性质也需满足某些特定的要求,如3D打印组织骨架和器官必须具有生物相容性、抗菌性及细胞粘附性等。因此,将3D打印与传统表面修饰技术相结合,在不改变材料三维结构的基础上调控其表面生物化学性质,从而赋予3D打印生物骨架器官多功能化,可实现更为广泛的应用。本文以3D打印生物骨架及器官的表面修饰为主要内容对就近年来3D打印生物医用材料的最新研究进展进行了综述。     

Crystal Structures of Actinomycin D and Actinomycin Z3

Untwinned single crystals of the actinomycins D and Z₃ that diffracted to atomic resolution could be obtained for the first time. Low-temperature data collection and a new ab initio method for solving the structures led to precise crystal structures which showed, for example, that the unit cell of actinomycin D contains three molecules, two of which are present in the form of a hydrogen-bridged dimer related by a pseudo-twofold axis (see picture).     

Intermolecular interaction-induced hierarchical transformation in 1D nanohybrids: analysis of conformational changes by 2D correlation spectroscopy

We have examined both self-assembly and confinement effect in room-temperature ionic liquid (RTIL)-aluminum hydroxide hybrids (RAHs) to attain a fundamental understanding of special phenomena in nanoscale spaces as well as to design functional nanomaterials for practical applications. Phase-controlled one-dimensional (1D) RAHs were synthesized through a simple ionothermal process. The RAHs were hierarchically transformed in terms of the molecular structures, morphologies, and phases of the materials during the ionothermal process with respect to the concentration of RTIL. In addition to the hierarchical transformation, the RTIL/aluminum hydroxide nanohybrids revealed unexpected physical behaviors, including thermal transition variation of the RTIL in confined environments and a phase transition from nanosolid to nanoliquid affected by changes of the melting points. More importantly, intermolecular interaction induced-self-assembly and confinement effect of RTILs inside an integrated hybrid system, which have not been clearly explained to date, were analyzed by 2D infrared correlation spectroscopy (2D IR COS); dynamic behaviors of RTILs, i.e., sequentially spatial reorientation and kinetically conformational changes, were attributed to the interactions between RTILs and aluminum hydroxides. 2D IR COS offers a new way to interpret highly complex, veiled systems such as the formation mechanism of nanoparticles, biomineralization, self/supramolecular assembly, and nanoconfinement.     

Interpretation of scoring functions using 3D molecular fields. Mapping the diacyl-hydrazine-binding pocket of an insect ecdysone receptor

A method is presented for the interpretation of receptor docking score values (rough measures of binding affinities) of ligands in terms of 3D molecular field interaction contributions. The FlexX and FlexX-Pharm methods were used to dock the structures of designed sets of ligands into the ligand-binding pocket of a selected receptor. In the next step the relationship was investigated between the FlexX and CScore scores and 3D molecular fields obtained for the docked conformations of the ligands, using the CoMFA (Comparative Molecular Field Analysis) and CoMSIA (Comparative Molecular Similarity Indices Analysis) methods. The approach yielded highly significant CoMFA and CoMSIA models demonstrating that a high portion of the variance in the docking score values of the ligands can be explained by steric, electrostatic, hydrophobic, and hydrogen bond donor and acceptor molecular field interaction contributions. The approach was exemplified by using the crystal structure of the ligand-binding domain of the ecdysone receptor (EcR) of the moth Heliotis virescens as well as virtual molecule libraries of analogues of known diacyl-hydrazine (DAH) type ecdysteroid agonists. By docking appropriately designed virtual compound libraries into the DAH binding pocket of EcR followed by CoMFA and CoMSIA of the docked conformations, hitherto unexplored regions of the receptor cavity could be mapped. By mapping the significant molecular field interaction contributions onto the model of the receptor-ligand complex, important receptor-ligand interactions could be highlighted that may help the design of novel highly scored receptor ligands. An advantage of the method is that no experimental biological activity data are required to exhaustively map the receptor-binding site.     

Halide anion-templated assembly of di- and triiodoperfluorobenzenes into 2D and 3D supramolecular networks

The single crystal structures of five co-crystals formed by the reaction of different iodide and bromide salts with di- and triiodoperfluorobenzenes (I-ArF) are reported. All of these perfluorocarbon-hydrocarbon systems are heteromeric three-component systems, wherein the weakly coordinating cations favour the formation of naked halides, which function as electron-donors towards the I-ArF modules. The analysis of the crystal structures shows that I⁻?I-ArF, and Br⁻?I-ArF halogen bonds (XBs) control the self-assembly of the obtained supramolecular architectures. 2D and 3D supramolecular networks have been obtained, wherein naked iodide and bromide anions act as tri-, tetra-, or pentadentate nodes. The selected examples demonstrate that I-ArF modules can be particularly robust and reliable tectons for XB-based coordination of halide ions and afford supramolecular architectures in a rational and predictable way.     

Surface-aligned ion-molecule reaction on the surface of a molecular crystal CD3+ + CD3I --> C2D5+ + DI

An ion-molecule reaction has been observed from a condensed molecular crystal of CD(3)I using the time-of-flight electron-stimulated desorption ion angular distribution technique. The CD(3)I multilayer is produced by growth on an ordered substrate. The reaction occurs between CD(3)(+) ions produced by electron-stimulated desorption and neighbor CD(3)I molecules in the topmost layer of the molecular crystal of CD(3)I, forming product C(2)D(5)(+) ions whose desorption dynamics have been measured. The normal momentum of the product ion is close to that of the reactant ion, suggesting that the reaction is dominated by a two-body collision, i.e., the momentum of the reactant CD(3)(+) ion governs the momentum of the product C(2)D(5)(+) ion. The ion-molecule reaction is of high cross section since the C(2)D(5)(+) yield is comparable to the CD(3)(+) yield. It is found that the yield and directionality of the emission of the C(2)D(5)(+) product ion is governed by the molecular order that is characteristic of the molecular crystal of CD(3)I. Destroying or modifying this order by using a spacer layer of H(2)O diminishes the C(2)D(5)(+) product ion yield relative to the reactant CD(3)(+) yield and broadens the ion emission directions.     

A 3D AgCl Hierarchical Superstructure Synthesized by a Wet Chemical Oxidation Method

A novel 3D AgCl hierarchical superstructure, with fast growth along the 〈111〉 directions of cubic seeds, is synthesized by using a wet chemical oxidation method. The morphological structures and the growth process are investigated by scanning electron microscopy and X‐ray diffraction. The crystal structures are analyzed by their crystallographic orientations. The surface energy of AgCl facets {100}, {110}, and {111} with absorbance of Cl^? ions is studied by density functional theory calculations. Based on the experimental and computational results, a plausible mechanism is proposed to illustrate the formation of the 3D AgCl hierarchical superstructures. With more active sites, the photocatalytic activity of the 3D AgCl hierarchical superstructures is better than those of concave and cubic ones in oxygen evolution under irradiation by visible light.     

Comparison between the Test and Simulation Results for PLA Structures 3D Printed,Bending Stressed

The additive manufacturing process is one of the technical domains that has had a sustained development in recent decades. The designers’ attention to equipment and materials for 3D printing has been focused on this type of process. The paper presents a comparison between the results of the bending tests and those of the simulation of the same type of stress applied on 3D-printed PLA and PLA–glass structures. The comparison of the results shows that they are close, and the simulation process can be applied with confidence for the streamline of filament consumption, with direct consequences on the volume and weight of additive manufactured structures. The paper determines whether the theories and concepts valid in the strength of materials can be applied to the additive manufacturing pieces. Thus, the study shows that the geometry of the cross-section, by its shape (circular or elliptical) and type (solid or ring shaped), influences the strength properties of 3D-printed structures. The use of simulation will allow a significant shortening of the design time of the new structures. Moreover, the simulation process was applied with good results on 3D-printed structures in which two types of filaments were used for a single piece (structure).