Accurate prediction of the energetics of weakly bound complexes using the machine learning method kriging

Structural Chemistry - Here, we extend the system energy prediction approach used in the force field FFLUX (Maxwell et al. Theor Chem Acc 135:195, 2016) to complexes bound by weak intermolecular...     

The effect of atomic and extra-atomic relaxation on atomic binding energies

An equivalent-cores-relaxation model is given for calculating atomic binding energies from orbital energies using only ground-state atomic properties. The agreement with experiment is excellent for the noble gases. On the basis of present knowledge of atomic relaxation, the phenomenon of “extra-atomic relaxation”, in which electronic charge is attracted toward a hole-state atom, is shown to have an important effect in lowering atomic core-level binding energies in condensed phases. This will affect the interpretation of most core-level binding energies measured to date.     

On the perception of molecules from 3D atomic coordinates

A method is presented for perceiving chemical types of atoms in molecules given 3D atomic coordinates and element identities. The method assigns hybridizations, bond orders, and formal charges for structures whether hydrogen atoms are present. The Maximum Weighted Matching algorithm for nonbipartite graphs is used to assign bond orders with weights derived from statistics of a large collection of organic molecules. Results form tests on a collection of functional groups, heterocycles, entries from the Protein Data Bank, and Cambridge Structural Database as well as a comparison to other methods, are presented and discussed.     

The cross-species prediction of bacterial promoters using a support vector machine

Due to degeneracy of the observed binding sites, the in silico prediction of bacterial sigma(70)-like promoters remains a challenging problem. A large number of sigma(70)-like promoters has been biologically identified in only two species, Escherichia coli and Bacillus subtilis. In this paper we investigate the issues that arise when searching for promoters in other species using an ensemble of SVM classifiers trained on E. coli promoters. DNA sequences are represented using a tagged mismatch string kernel. The major benefit of our approach is that it does not require a prior definition of the typical -35 and -10 hexamers. This gives the SVM classifiers the freedom to discover other features relevant to the prediction of promoters. We use our approach to predict sigma(A) promoters in B. subtilis and sigma(66) promoters in Chlamydia trachomatis. We extended the analysis to identify specific regulatory features of gene sets in C. trachomatis having different expression profiles. We found a strong -35 hexamer and TGN/-10 associated with a set of early expressed genes. Our analysis highlights the advantage of using TSS-PREDICT as a starting point for predicting promoters in species where few are known.     

Core ionization energies of atoms and ions calculated using the generalized Sturmian method

The physical event of the umbrella inversion of ammonia has been studied in detail by application of the formalisms of frontier orbital theory, the density functional theory, the localized molecular orbital method, and the energy partitioning analysis. An intuitive structure for the transition state and dynamics of the physical process of structural reorganization prior to inversion have been suggested. The computation starts with the CNDO/2 equilibrium geometry, and thereafter the cycle proceeds through all the conformations of ammonia obtained by varying the ∠HNH angle in steps of 2° from its equilibrium value up to the transition state. The geometry of each conformation is optimized with respect to the length of the N–H bond. The glimpses of the charge density reorganization during the movement of the molecule from equilibrium conformation toward the transition state is computed in terms of dipole moment and the quantum mechanical hybridizations of bond pair and lone pair of N atom through the localized molecular orbitals (LMOs) of all the conformations. Results demonstrate that as the geometry of the molecule begins to evolve through the reorganization of structure, the N–H bond length and the dipole moment begin to decrease, and the trend continues up to the transition state. The dipole moment of the molecule at the suggested transition state is zero. The computed nature of quantum mechanical hybridization of bond pair and lone pair of the N atom as a function of reaction coordinates of the ∠HNH angles reveals that the percentage of s character of the lone pair hybrid decreases and that of the bond pair hybrid forming the σ(N–H) bond increases during the process of geometry reorganization from the equilibrium shape to the transition state. The rationale of the zero dipole moment of the transition state for inversion is not straightforward from its point‐group symmetry, but is self‐evident from its electronic structure drawn in terms of LMOs. The electronic structure of the transition state, which may be drawn in terms of the LMOs, seems to closely reproduce its suggested intuitive structure. The pattern of variation of dipole moment and nature of the changes of the percentage of the s character in the lone pair hybrid creating dipole with the evolution of geometry during the physical process of structural reorganization for the inversion are found to be nicely correlated according to the suggestion of Coulson. The profiles of the increasing strength of the N–H bond and the increasing percentage of s character of the bond pair hybrid of N atom forming this bond as a function of reaction coordinates are also found to be correlated in accordance with the suggestion of Coulson. The profile of global hardness as a function of reaction coordinate seems to demonstrate that the dynamics of the evolution of the molecular structure from equilibrium shape to the transition state following the course of suggested mode of structural reorganization conforms to the principle of maximum hardness (PMH). The profiles of parameters like the energies of highest occupied and lowest unoccupied molecular orbital (HOMO and LUMO), the gap in energy between HOMO and LUMO, the global hardness, the global softness, and chemical potential as a function of reaction coodinates of a continuous structural evolution from equilibrium shape to the transition state mimic the potential energy diagram of the total energy. Both the frontier orbital parameters and the density functional quantities are found to be equally effective and reliable to monitor the process of necessary structural reorganization prior to the inversion of mofecules. An energy partitioning analysis demonstrates that the origin of barrier has no unique single source rather as many as four mutually exclusive but interacting one‐ and two‐center energy terms within the molecule entail the origin and the height of the barrier. From a close analysis of the results, it seems highly probable that the necessary structural reorganization prior to umbrella inversion of ammonia most realistically occurs following the course of normal modes of vibration of the molecule. © 2000 John Wiley & Sons, Inc. Int J Quant Chem 80: 1–26, 2000     

Development of a sugar-binding residue prediction system from protein sequences using support vector machine

Several methods have been proposed for protein–sugar binding site prediction using machine learning algorithms. However, they are not effective to learn various properties of binding site residues caused by various interactions between proteins and sugars. In this study, we classified sugars into acidic and nonacidic sugars and showed that their binding sites have different amino acid occurrence frequencies. By using this result, we developed sugar-binding residue predictors dedicated to the two classes of sugars: an acid sugar binding predictor and a nonacidic sugar binding predictor. We also developed a combination predictor which combines the results of the two predictors. We showed that when a sugar is known to be an acidic sugar, the acidic sugar binding predictor achieves the best performance, and showed that when a sugar is known to be a nonacidic sugar or is not known to be either of the two classes, the combination predictor achieves the best performance. Our method uses only amino acid sequences for prediction. Support vector machine was used as a machine learning algorithm and the position-specific scoring matrix created by the position-specific iterative basic local alignment search tool was used as the feature vector. We evaluated the performance of the predictors using five-fold cross-validation. We have launched our system, as an open source freeware tool on the GitHub repository (https://doi.org/10.5281/zenodo.61513).     

Multi-objective optimization of detergent pre-formulations using machine learning techniques

The right combination of surfactants and stabilizers in the detergent formulations plays a significant role in their cleaning performance. However, it becomes a complex optimization problem when the formulation is composed of multiple ingredients and the solution has to be optimized for competing performance metrics. In recent times, machine learning techniques have been used extensively to study such processes. In this research, a detergent pre-formulation has been designed using an aqueous solution of Tween-20, Ethanol and 1-Octanol. To determine the optimal values of the ingredients of the formulations, supervised machine learning models were developed and optimized for the Ross Miles Index 30 ml (RMI 30) and cleaning time (CT). A full factorial experimental design was performed and three regression models based on linear, 2FI and Quadratic designs were developed respectively for RMI30 and CT. ANOVA analysis of trained models reported an optimal p-value of 0.0018 for RMI 30 and less than 0.0001 for CT. The optimal values for RMI30 and CT obtained through regression models are 72.32 ml and 17.67 s. For multi-objective optimization, grey relational analysis was performed. Two pairs of optimal values corresponding to Rank 1 were recorded as 88.9 ml, 20 s (RMI30, CT); and 81.2 ml, 14 s (RMI30, CT) respectively. As a result, the optimal combination of Tween-20, Ethanol and 1-Octanol for maximizing the RMI30 and minimizing the CT are reported. The obtained optimal values were experimentally validated.     

eFindSite: Improved prediction of ligand binding sites in protein models using meta-threading, machine learning and auxiliary ligands

Molecular structures and functions of the majority of proteins across different species are yet to be identified. Much needed functional annotation of these gene products often benefits from the knowledge of protein–ligand interactions. Towards this goal, we developed eFindSite, an improved version of FINDSITE, designed to more efficiently identify ligand binding sites and residues using only weakly homologous templates. It employs a collection of effective algorithms, including highly sensitive meta-threading approaches, improved clustering techniques, advanced machine learning methods and reliable confidence estimation systems. Depending on the quality of target protein structures, eFindSite outperforms geometric pocket detection algorithms by 15–40 % in binding site detection and by 5–35 % in binding residue prediction. Moreover, compared to FINDSITE, it identifies 14 % more binding residues in the most difficult cases. When multiple putative binding pockets are identified, the ranking accuracy is 75–78 %, which can be further improved by 3–4 % by including auxiliary information on binding ligands extracted from biomedical literature. As a first across-genome application, we describe structure modeling and binding site prediction for the entire proteome of Escherichia coli. Carefully calibrated confidence estimates strongly indicate that highly reliable ligand binding predictions are made for the majority of gene products, thus eFindSite holds a significant promise for large-scale genome annotation and drug development projects. eFindSite is freely available to the academic community at http://www.brylinski.org/efindsite.     

Determination of one-centre core integrals from the average energies of atomic configurations

One-center core integrals for valence orbitals are determined from the experimental average energies of neutral atomic configurations from Li through Zn. These values are compared with those estimated from CNDO /1, “INDO /1”, CNDO /2, “INDO /2” and with theoretical values calculated from a pseudo-potential method. The agreement is good between values obtained from neutral atoms and from the psuedo-potential calculation except for the 3d orbitals of the transition elements where the theoretically calculated integrals over single ξ functions are not realistic. These two methods reproduce both term and average configuration energies for the first two rows of atoms; the semiempirical method reliably reproduces them for the third row. The CNDO /1 and INDO /1 methods underestimate atomic energies, while the CNDO /2 and INDO /2 procedures fail rather poorly. The propriety of using core integrals estimated semiempirically in molecular orbital calculations is discussed.     

Ionization energy of atoms obtained from GW self-energy or from random phase approximation total energies

A systematic evaluation of the ionization energy within the GW approximation is carried out for the first row atoms, from H to Ar. We describe a Gaussian basis implementation of the GW approximation, which does not resort to any further technical approximation, besides the choice of the basis set for the electronic wavefunctions. Different approaches to the GW approximation have been implemented and tested, for example, the standard perturbative approach based on a prior mean-field calculation (Hartree-Fock GW@HF or density-functional theory GW@DFT) or the recently developed quasiparticle self-consistent method (QSGW). The highest occupied molecular orbital energies of atoms obtained from both GW@HF and QSGW are in excellent agreement with the experimental ionization energy. The lowest unoccupied molecular orbital energies of the singly charged cation yield a noticeably worse estimate of the ionization energy. The best agreement with respect to experiment is obtained from the total energy differences within the random phase approximation functional, which is the total energy corresponding to the GW self-energy. We conclude with a discussion about the slight concave behavior upon number electron change of the GW approximation and its consequences upon the quality of the orbital energies.     

The use of interparticle coordinates in electronic energy calculations for atoms and molecules

The use of interparticle coordinates in atomic and molecular calculations is reviewed and emphasized. For states where the wave function is dependent only upon interparticle coordinates the Hamiltonian is a relatively simple expression. If the wave function is a linear combination of products of functions of single interparticle coordinates of the exponential power form the formulas for local energy are particularly simple.

---

Zusammenfassung Es wird über Teilchenabständc als Koordinaten in Atom- und Molekülproblemen referiert und auf ihren Wert hingewiesen. In Fällen, wo die Wellenfunktion nur von Teilchenabständen abhängt, wird der Hamilton-Operator verhältnismäßig einfach. Ist die Wellenfunktion eine Linearkombination von Produkten von Potenz-Exponential-Funktionen einzelner Teilchenabstände, so werden die Ausdrücke für die lokale Energie besonders einfach.

---

Résumé L'auteur fait une revue et souligne l'importance de l'emploi des distances relatives des particules comme coordonnées pour les calculs atomiques et moléculaires. Pour des états dont les fonctions d'onde ne dépendent que de ces distances le l'Hamiltonien devient relativement simple. Si la fonction d'onde s'écrit comme une combinaison linéaire de produits de fonctions exponentielles (à puissances) ne dépendant que d'une seule coordonnée interparticulaire, les formules pour l'énergie locale se simplifient particulièrement.

     

Blind prediction of solvation free energies from the SAMPL4 challenge

Here, we give an overview of the small molecule hydration portion of the SAMPL4 challenge, which focused on predicting hydration free energies for a series of 47 small molecules. These gas-to-water transfer free energies have in the past proven a valuable test of a variety of computational methods and force fields. Here, in contrast to some previous SAMPL challenges, we find a relatively wide range of methods perform quite well on this test set, with RMS errors in the 1.2 kcal/mol range for several of the best performing methods. Top-performers included a quantum mechanical approach with continuum solvent models and functional group corrections, alchemical molecular dynamics simulations with a classical all-atom force field, and a single-conformation Poisson–Boltzmann approach. While 1.2 kcal/mol is still a significant error, experimental hydration free energies covered a range of nearly 20 kcal/mol, so methods typically showed substantial predictive power. Here, a substantial new focus was on evaluation of error estimates, as predicting when a computational prediction is reliable versus unreliable has considerable practical value. We found, however, that in many cases errors are substantially underestimated, and that typically little effort has been invested in estimating likely error. We believe this is an important area for further research.     

机器学习方法用于选择性环氧化酶-2抑制剂活性预测模型的建立

与传统的非甾体类消炎药相比,选择性环氧化酶-2抑制剂具有无胃肠道粘膜损伤,溃疡和肾功能障碍等严重的副作用,设计选择性环氧化酶-2抑制剂具有重要意义。本文用支持矢量学习机和神经网络两种机器学习方法建立选择性环氧化酶-2抑制剂的活性预测模型,以期为选择性环氧化酶-2抑制剂药物的合成提供先导化合物。我们将467个环氧化酶-2抑制剂用Kennard-Stone方法分为训练集,验证集和独立测试集,对每一抑制剂分子我们计算了463个包含组成描述符和拓扑描述符的分子描述符来表征其分子结构,并通过F-Score方法选取最重要的分子描述符用于分类模型的建立。结果表明,SVM方法通过变量筛选后具有很好的预测能力,其预测正确率达到93.30%。     

Data-driven machine learning models for quick prediction of thermal stability properties of OLED materials

Organic light-emitting diode (OLED) materials have exhibited a wide range of applications. However, the further development and commercialization of OLEDs requires higher quality OLED materials, including materials with a high thermal stability. Thermal stability is associated with the glass transition temperature (T_g) and decomposition temperature (T_d), but experimental determinations of these two important properties generally involve a time-consuming and laborious process. Thus, the development of a quick and accurate prediction tool is highly desirable. Motivated by the challenge, we explored machine learning (ML) by constructing a new dataset with more than 1,000 samples collected from a wide range of literature, through which ensemble learning models were explored. Models trained with the LightGBM algorithm exhibited the best prediction performance, where the values of mean absolute error, root mean squared error, and R² were 17.15 K, 24.63 K, and 0.77 for T_g prediction and 24.91 K, 33.88 K, and 0.78 for T_d prediction. The prediction performance and the generalization of the ML models were further tested by two applications, which also exhibited satisfactory results. Experimental validation further demonstrated the reliability and the practical potential of the ML-based models. In order to extend the practical application of the ML-based models, an online prediction platform was constructed. This platform includes the optimal prediction models and all the thermal stability data under study, and it is freely available at http://www.oledtppxmpugroup.com. We expect that this platform will become a useful tool for experimental investigation of T_g and T_d, accelerating the design of OLED materials with desired properties.     

Large-Z and -N dependence of atomic energies from renormalization of the large-dimension limit

By combining Hartree–Fock results for nonrelativistic ground-state energies of N-electron atoms with analytic expressions for the large-dimension limit, we have obtained a simple renormalization procedure. For neutral atoms, this yields energies typically threefold more accurate than the Hartree–Fock approximation. Here, we examine the dependence on Z and N of the renormalized energies E(N, Z) for atoms and cations over the range Z, N = 2 → 290. We find that this gives for large Z = N an expansion of the same form as the Thomas–Fermi statistical model, E → Z^7/2(C₀ + C₁Z^?1/3 + C₂Z^?2/3 + C₃Z^?3/3 + ?), with similar values of the coefficients for the three leading terms. Use of the renormalized large-D limit enables us to derive three further terms. This provides an analogous expansion for the correlation energy of the form δE δZ^4/3(δC₃ + δC₅Z^?2/3 + δC₆Z^?3/3 + ?); comparison with accurate values of δE available for the range Z ? 36 indicates the mean error is only about 10%. Oscillatory terms in E and δE are also evaluated. © 1994 John Wiley & Sons, Inc.     

Estimation of O-H bond dissociation energies in alcohols and acids from kinetic data

The O-H bond dissociation energies (D _O-H) in five alcohols and six acids have been determined from experimental data (rate constants of radical reactions). The ratio of the rate constants of the reactions R¹O˙+RH→R¹OH+R˙ and R ⁱ O˙+RH→R ⁱ OH+R˙ and the intersecting parabolas method are used in the estimation procedure. The D _O-H values are used to calculate the activation energies and rate constants for hydrogen abstraction from 2-methylbutane, butene-1, and cumene by alkoxyl and carboxyl radicals. The geometric parameters of the transition state are calculated for these reactions.