Variable selection for QSAR by artificial ant colony systems

Derivation of quantitative structure-activity relationships (QSAR) usually involves computational models that relate a set of input variables describing the structural properties of the molecules for which the activity has been measured to the output variable representing activity. Many of the input variables may be correlated, and it is therefore often desirable to select an optimal subset of the input variables that results in the most predictive model. In this paper we describe an optimization technique for variable selection based on artificial ant colony systems. The algorithm is inspired by the behavior of real ants, which are able to find the shortest path between a food source and their nest using deposits of pheromone as a communication agent. The underlying basic self-organizing principle is exploited for the construction of parsimonious QSAR models based on neural networks for several classical QSAR data sets.     

Machine Learning Informs RNA-Binding Chemical Space**

Small molecule targeting of RNA has emerged as a new frontier in medicinal chemistry, but compared to the protein targeting literature our understanding of chemical matter that binds to RNA is limited. In this study, we reported R epository O f BI nders to N ucleic acids (ROBIN), a new library of nucleic acid binders identified by small molecule microarray (SMM) screening. The complete results of 36 individual nucleic acid SMM screens against a library of 24 572 small molecules were reported (including a total of 1 627 072 interactions assayed). A set of 2 003 RNA-binding small molecules was identified, representing the largest fully public, experimentally derived library of its kind to date. Machine learning was used to develop highly predictive and interpretable models to characterize RNA-binding molecules. This work demonstrates that machine learning algorithms applied to experimentally derived sets of RNA binders are a powerful method to inform RNA-targeted chemical space.     

机器学习方法用于二氢叶酸还原酶抑制剂的活性预测

分别采用支持向量学习机、人工神经网络、调节性逻辑回归和K-最临近等机器学习方法对761个二氢叶酸还原酶抑制剂建立了其活性分类预测模型. 采用组成描述符和拓扑描述符表征抑制剂的分子结构及物理化学性质, 使用Kennard-Stone方法进行训练集的设计, 并用Metropolis Monte Carlo模拟退火方法作变量选择. 结果表明, 支持向量学习机优于其它机器学习方法, 所得到的最优模型具有较好的预测结果, 其预测正确率为91.62%. 说明通过合适的训练集设计及变量选择, 支持向量学习机方法可以很好地用于二氢叶酸还原酶抑制剂的活性分类预测.     

Artificial intelligence and machine learning for targeted energy storage solutions

With the application of machine learning to large-material data sets, models are being developed that allow us to better predict novel materials with designed properties. Advances in artificial intelligence and its subclasses, as well as compute infrastructure, are making it possible to rapidly compute material properties, to access time/length scales and chemical spaces beyond the current capabilities of density functional theory and to outperform humans in interpretation and characterization of the data. This review highlights the latest developments in the field with special interest to energy storage materials.     

Atomistic Modeling of Silica Based Sol-Gel Processes

Density Functional Theory is used to study water, methanol, ethanol, TMOS, and TEOS molecules and the most important silica clusters participating in sol-gel processes. Calculated bond lengths, bond angles and electric dipole moments compare well with experimental data. The energy of these molecules is reported and used to discuss the energetics of the hydrolysis and condensation reactions. Molecular Dynamics is employed to simulate liquid water, methanol, ethanol, TMOS, TEOS and experimental sol-gel solutions. Calculated densities and enthalpies of vaporisation compare well with experimental data. Preliminary results are presented for MD simulations of sol-gel solutions.     

Atomistic Simulations of Hydrated Sulfonated Polybenzophenone Block Copolymer Membranes

We present a classical molecular dynamics simulations study on the nanostructures of the sulfonated polybenzophenone (SPK) block copolymer membranes at 300 K and 353 K. The results of the radial distribution function (RDF) show that the interactions of the sulfonate groups of the membrane with the hydronium ions are more significant than those of water due to the strong electrostatic attraction over the hydrogen bonding. However, the effect of temperatures on the RDF profile seems insignificant. Furthermore, the spatial distribution function (SDF) portrays that the sulfonate groups of the hydrophilic components are preferential binding sites for hydronium ions against the hydrophobic counterpart of the SPK membrane. The mobility of the H₃O⁺ ions at 300 K and 353 K is two (or three) times lower than that of Nafion/Aciplex. However, the diffusion coefficients for water molecules closely agree with Nafion/Aciplex. This study suggests that water clusters are more localized around the sulfonate groups in the SPK membranes. Thus, the molecular modeling study of SPK block copolymer membranes is warranted to design better-performing membrane electrolytes.     

Data-Driven Machine Learning for Understanding Surface Structures of Heterogeneous Catalysts

The design of heterogeneous catalysts is necessarily surface-focused, generally achieved via optimization of adsorption energy and microkinetic modelling. A prerequisite is to ensure the adsorption energy is physically meaningful is the stable existence of the conceived active-site structure on the surface. The development of improved understanding of the catalyst surface, however, is challenging practically because of the complex nature of dynamic surface formation and evolution under in-situ reactions. We propose therefore data-driven machine-learning (ML) approaches as a solution. In this Minireview we summarize recent progress in using machine-learning to search and predict (meta)stable structures, assist operando simulation under reaction conditions and micro-environments, and critically analyze experimental characterization data. We conclude that ML will become the new norm to lower costs associated with discovery and design of optimal heterogeneous catalysts.     

Integrating Computation,Experiment, and Machine Learning in the Design of Peptide-Based Supramolecular Materials and Systems

Interest in peptide-based supramolecular materials has grown extensively since the 1980s and the application of computational methods has paralleled this. These methods contribute to the understanding of experimental observations based on interactions and inform the design of new supramolecular systems. They are also used to virtually screen and navigate these very large design spaces. Increasingly, the use of artificial intelligence is employed to screen far more candidates than traditional methods. Based on a brief history of computational and experimentally integrated investigations of peptide structures, we explore recent impactful examples of computationally driven investigation into peptide self-assembly, focusing on recent advances in methodology development. It is clear that the integration between experiment and computation to understand and design new systems is becoming near seamless in this growing field.     

Understanding Alkaline Hydrogen Oxidation Reaction on PdNiRuIrRh High-Entropy-Alloy by Machine Learning Potential

High-entropy alloy (HEA) catalysts have been widely studied in electrocatalysis. However, identifying atomic structure of HEA with complex atomic arrangement is challenging, which seriously hinders the fundamental understanding of catalytic mechanism. Here, we report a HEA-PdNiRuIrRh catalyst with remarkable mass activity of 3.25 mA μg⁻¹ for alkaline hydrogen oxidation reaction (HOR), which is 8-fold enhancement compared to that of commercial Pt/C. Through machine learning potential-based Monte Carlo simulation, we reveal that the dominant Pd−Pd−Ni/Pd−Pd−Pd bonding environments and Ni/Ru oxophilic sites on HEA surface are beneficial to the optimized adsorption/desorption of *H and enhanced *OH adsorption, contributing to the excellent HOR activity and stability. This work provides significant insights into atomic structure and catalytic mechanism for HEA and offers novel prospects for developing advanced HOR electrocatalysts.     

Two-Dimensional Supramolecular Polymerization of DNA Amphiphiles is Driven by Sequence-Dependent DNA-Chromophore Interactions

Two-dimensional (2D) assemblies of water-soluble block copolymers have been limited by a dearth of systematic studies that relate polymer structure to pathway mechanism and supramolecular morphology. Here, we employ sequence-defined triblock DNA amphiphiles for the supramolecular polymerization of free-standing DNA nanosheets in water. Our systematic modulation of amphiphile sequence shows the alkyl chain core forming a cell membrane-like structure and the distal π-stacking chromophore block folding back to interact with the hydrophilic DNA block on the nanosheet surface. This interaction is crucial to sheet formation, marked by a chiral “signature”, and sensitive to DNA sequence, where nanosheets form with a mixed sequence, but not with a homogeneous poly(thymine) sequence. This work opens the possibility of forming well-ordered, bilayer-like assemblies using a single DNA amphiphile for applications in cell sensing, nucleic acid therapeutic delivery and enzyme arrays.     

Frustration in Super-Ionic Conductors Unraveled by the Density of Atomistic States

The frustration in super-ionic conductors enables their exceptionally high ionic conductivities, which are desired for many technological applications including batteries and fuel cells. A key challenge in the study of frustration is the difficulties in analyzing a large number of disordered atomistic configurations. Using lithium super-ionic conductors as model systems, we propose and demonstrate the density of atomistic states (DOAS) analytics to quantitatively characterize the onset and degree of disordering, reveal the energetics of local disorder, and elucidate how the frustration enhances diffusion through the broadening and overlapping of the energy levels of atomistic states. Furthermore, material design strategies aided by the DOAS are devised and demonstrated for new super-ionic conductors. The DOAS is generally applicable analytics for unraveling fundamental mechanisms in complex atomistic systems and guiding material design.     

基于机器学习方法的激素敏感脂肪酶抑制剂活性预测(英文)

脂肪组织中,激素敏感脂肪酶(HSL)被认为是调节脂肪酸代谢的关键限速酶.HSL在糖尿病的发病过程中起重要作用,抑制HSL活性有助于糖尿病的治疗,因此探索新颖的HSL抑制剂成为当前研究的热门.在激素敏感脂肪酶的作用机制和三维结构缺乏的情况下,需要发展预测HSL抑制剂的方法.本文采用几种机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5DT))对已知的HSL抑制剂与非抑制剂建立分类预测模型.252个结构多样性化合物(123个HSL抑制剂与129个HSL非抑制剂)被用于测试分类预测系统,并用递归变量消除法选择与HSL抑制剂相关的性质描述符以提高预测精度.本研究对独立验证集的总预测精度为75.0%-80.0%,HSL抑制剂的预测精度为85.7%-90.5%,非HSL抑制剂的预测精度为63.2%-68.4%.支持向量机方法给出最好的总预测精度(80.0%).本研究表明支持向量机等机器学习方法可以有效预测未知数据集中潜在的HSL抑制剂,并有助于发现与其相关的分子描述符.     

High-Performance Chemical Information Database towards Accelerating Discovery of Metal-Organic Frameworks for Gas Adsorption with Machine Learning

Chemical structure searching based on databases and machine learning has attracted great attention recently for fast screening materials with target functionalities. To this end, we established a high-performance chemical structure database based on MYSQL engines, named MYDB. More than 160000 metal-organic frameworks (MOFs) have been collected and stored by using new retrieval algorithms for efficient searching and recommendation. The evaluations results show that MYDB could realize fast and efficient keyword searching against millions of records and provide real-time recommendations for similar structures. Combining machine learning method and materials database, we developed an adsorption model to determine the adsorption capacitor of metal-organic frameworks toward argon and hydrogen under certain conditions. We expect that MYDB together with the developed machine learning techniques could support large-scale, low-cost, and highly convenient structural research towards accelerating discovery of materials with target functionalities in the field of computational materials research.     

Pure Isotropic Proton NMR Spectra in Solids using Deep Learning

The resolution of proton solid-state NMR spectra is usually limited by broadening arising from dipolar interactions between spins. Magic-angle spinning alleviates this broadening by inducing coherent averaging. However, even the highest spinning rates experimentally accessible today are not able to completely remove dipolar interactions. Here, we introduce a deep learning approach to determine pure isotropic proton spectra from a two-dimensional set of magic-angle spinning spectra acquired at different spinning rates. Applying the model to 8 organic solids yields high-resolution ¹H solid-state NMR spectra with isotropic linewidths in the 50–400 Hz range.     

基于机器学习方法的丙型肝炎病毒非结构蛋白5B聚合酶抑制剂活性预测

在丙型肝炎病毒(HCV)的基因复制和蛋白质成熟的过程中, 非结构蛋白5B(NS5B)作为RNA依赖的RNA聚合酶起到了重要的作用. 抑制NS5B聚合酶可以阻止丙型肝炎病毒的RNA复制, 因此成为一种治疗丙型肝炎的有效方法. 通过计算机方法进行虚拟筛选和预测NS5B聚合酶抑制剂已经变得越来越重要. 本文主要采用机器学习方法(支持向量机(SVM)、k-最近相邻法(k-NN)和C4.5决策树(C4.5 DT))对已知的丙型肝炎病毒NS5B蛋白酶抑制剂与非抑制剂建立分类预测模型. 1248个结构多样性化合物(552个NS5B抑制剂与696个非NS5B抑制剂)被用于测试分类预测系统, 并用递归变量消除法选择与NS5B抑制剂相关的性质描述符以提高预测精度. 独立验证集的总预测精度为84.1%-85.0%, NS5B抑制剂的预测精度为81.4%-91.7%, 非NS5B抑制剂的预测精度为78.2%-87.2%. 其中支持向量机给出最好的NS5B抑制剂预测精度(91.7%); C4.5决策树给出最好的非NS5B抑制剂预测精度(87.2%); k-最近相邻法给出最好的总预测精度(85.0%). 研究表明机器学习方法可以有效预测未知数据集中潜在的NS5B抑制剂, 并有助于发现与其相关的分子描述符.     

机器学习设计单步逆向合成反应的研究进展

在逆向合成分析的过程中,对特定的目标分子设计单步逆向合成反应是探寻最优有机合成路线的关键环节。随着机器学习(Machine learning)研究的兴起,很多研究者开始尝试利用机器学习方法设计单步逆向合成反应。相关研究主要集中在两方面：(1)研究化合物分子输入方法;(2)基于特定的分子输入,研究各类单步逆向合成反应预测模型的构建方法。本文首先综述了分子输入的三种主流方法;然后分别分析了基于这三种分子输入方法构建的单步逆向合成反应预测模型的研究实例;之后,总结了当前机器学习方法设计单步逆向合成反应研究中存在的问题,并给出了解决问题的思路;最后,对机器学习设计单步逆向合成反应的前景作出展望。     

Predicting of Covalent Organic Frameworks for Membrane-based Isobutene/1,3-Butadiene Separation: Combining Molecular Simulation and Machine Learning

Efficient separation of C4 olefins is of critical importance and a challenging task in petrochemical industry. Covalent organic frameworks(COFs) could be used as promising candidates for membrane-based isobutene/1,3-butadiene(i-C₄H₈/C₄H₆) separation. Owing to large amounts of COFs appearing, however, the rapid prediction of optimal COFs is imperative before experimental efforts. In this work, we combine molecular simulation and machine learning to study COF membranes for efficient isolation of i-C₄H₈ over C₄H₆. Using molecular simulation, four potential COF membranes, which possess both high membrane performance score (MPS) value and moderate membrane selectivity were screened out and the mechanism of membrane separation further revealed is an adsorption dominated process. Further, random forest(RF) model with high prediction accuracy(R²>0.84) was obtained and used for elucidating key factors in controlling the membrane selectivity and i-C₄H₈ permeability. Ultimately, the optimal COF features were obtained through structure-performance relationship study. Our results may trigger experimental efforts to accelerate the design of novel COFs with better i-C₄H₈/C₄H₆separation performance.