Development of solid-state catalysts has been highly difficult and needed long period researches because the catalysts are intrinsically complicated materials. Development of next solid-state catalysts for demanded but undeveloped catalytic reactions will be even more difficult because necessary catalysts will be more complex in terms of functionality as well as constituent number. One of the most important methodology for solving this situation should be catalysis informatics. For accelerating the catalysis informatics, a few practical examples of solid-state catalyst development will be introduced.

View full abstract

Volatile organic compounds (VOCs) are small molecules exhibit high vapor pressure under ambient conditions. VOCs are produced naturally by living organisms and have important role in chemical ecology and human health. Information on volatile emission from various organisms is scattered in the literature, however, there is still no available database describing VOCs and their biological activities. To attain this, we have developed Metabolite Ecology Database, which contains the information on the relationships between VOCs, emission organisms and their biological activities. In this presentation, we discuss VOC database development and chemometrics methods for analyzing the relationships between chemical structures of VOCs and their corresponding biological functions.

View full abstract

In the field of chemoinformatics, computer-aided reaction prediction systems have been developed. However, the systems have not been widely used by experimental chemists because of its low accuracy for predicting chemical reactions. Recently, a reaction prediction scheme, which is based on the machine learning method and utilizes topological information of molecules as descriptors, has shown high effectiveness for organic reactions. Nevertheless, the application has been limited due to lack of steric and electronic information of molecules. The present study has developed a novel scheme, which uses the machine learning with descriptors obtained from quantum chemical calculation, in order to realize an accurate prediction for reactions including arbitrary chemical species such as organometallic and ionic reactions. The accuracy of present system has been clarified to be close to that of the previous system for basic polar and radical organic chemical reactions. In this presentation, we will show the details of the system and the dependencies on conditions of quantum chemical calculations.

View full abstract

We have been developing QMRDB (Quantum Mechanical calculation Results Data Base) to reuse results of theoretical calculations for further investigations and educations in molecular chemistry. QMRDB uses the PostgreSQL program for data handling and the Open Babel program for molecular structure retrieving. By use of information in QMRDB, we are constructing another data base named TSDB (TS Data Base) which is intended to use for in silico screening. In order to analyze chemical reaction mechanisms efficiently, we also developed “the substitution method for locating transition states (TSs). We have to distinguish the differences of substituents and their positions between for the method. We implemented a function which specifies the maximum structure component (MSC) between a reference and a target product. We used the jar of Chemistry Development Kit (CDK) to extract the MSC fragment and display its structure in the iStructure program. The function is closely related to a synthesis route proposition system using only theoretical calculations which we are now developing.

View full abstract

Although bioactive 2-azaspiro compounds were reported in recent years, there is a few numbers of effective synthetic methods. We found the synthesis of 2-azaspiro compound with the hypervalent iodine reagent in previous report. In this reaction, however, there is a significant substituent effect on the nitrogen atom of amide group. The reaction mechanism is investigated by the stable structures by using the Hamiltonian algorithm.

View full abstract

Water clusters with a polyhedral framework have many isomers with different hydrogen bond (HB) networks. Although the number of HB networks of various polyhedral water clusters in which every water molecule forms three HBs was reported, little work has been done on the HB networks of water clusters with molecules having four HBs. In this study, we discuss the HB networks of dicubic water dodecamers. The dicube is formed by joining two cubes face to face. All topologically-distinct HB networks of dicubic water dodecamer are presented with a digraph representation. The HB networks are classified into 13 groups based on the numbers of 9 types of HBs formed in the dicubic water dodecamers. Structures of dicubic water dodecamers were produced from the HB networks and were optimized at the MP2/aug-cc-pVDZ level of theory. Although most optimized structures retained initial dicubic HB networks, single HB was broken during the geometry optimization in some structures. We found that there was a common local HB network around the broken HB.

View full abstract

Group 6 elements, chromium (Cr), molybdenum (Mo), and tungsten (W), are important trace elements for geochemistry since they are redox sensitive and have several stable isotopes. In particular, Mo has drawn attention in paleoceanography because its geochemical behavior sharply changes depending on the redox conditions (Anbar, 2004). A fractionation for Mo was observed during the adsorption to manganese oxides, i.e. δ^97/95Mo = -1.8‰ (Wasylenk et al., 2011). On the other hand, Cr did not show an isotope fractionation during the adsorption to Fe-(oxyhydr)oxides (Ellis et al., 2002). Kashiwabara et al. (2011) suggested that the Mo isotope fractionation at water/metal oxides interface is caused by symmetrical change of molybdate from tetrahedral (Td) to octahedral (Oh) structure during adsorption by XAFS analysis. However, this fact was not confirmed well. In this study, we tried to understand the isotope fractionation of group 6 elements accompanying with adsorption on Fe/Mn-(oxyhydr)oxides by means of density functional theory (DFT) calculations and information of adsorption structure obtained by XAFS.

View full abstract

Trehalose is a non-reducing disaccharide composed of two α-(D)-glucose units linked together by a 1,1-glycosidic linkage. Trehalose has a number of conformers because it has 12 independent dihedral angles, which include the conformations of 8 OH groups, of the glycosidic linkage, and of CH₂OH groups. The relative stabilities of conformations of the disaccharide depend on the intramolecular hydrogen bond networks. In this work, we clarify stable conformers of trehalose using MP2 level of ab initio MO theory. We find out that the relative stabilities of conformers of trehalose depend on the conformation of glycosidic linkage. The most stable conformer of trehalose is composed of the most stable conformer of α-glucose held together by a 1,1-glycosidic linkage.

View full abstract

Generating virtual chemical structures by combining building blocks is frequently conducted by chemists with the help of computer programs. This simple task usually suffers from generating duplicates, leading to inefficiency of the structure generation. On top of that, diversity-oriented structure generation is required when exhaustive generation is intractable. Here, we have proposed a structure generation algorithm by combining ring systems and atom fragments in a tree-like way without generating duplicates. The algorithm makes use of reduced graphs, which retain the same topology as that of the corresponding ring systems but have less vertices than in the corresponding ring systems. For diversity-oriented structure generation, we have proposed an algorithm that generate a chemical graph per atom-fragment-based framework. One of the biggest advantages of this algorithm is that it takes diversity into account during structure generation instead of sampling diverse structures after structure generation. The features and efficiencies of the proposed two algorithms were demonstrated with simple structure generation case studies.

View full abstract

An NMR database for natural organic compounds ”CH-NMR-NP” was opened as a freely accessible system from JEOL Resonance Homepage. Total number of the compounds are about 30,000 compiled mainly from the published papers from 2000 to 2014. Items for a compound are Name, Molecular formula, Chemical structure with assigment, the values of ¹³C and ¹H shifts, ¹H-coupling information,solvent, reference and others

View full abstract

A chemical structure database is an essential tool to store and retrieve chemical information. For a long term, enormous cost for purchasing and developing prevent academic organization from utilizing chemical structure database systems in house. Although an academic licensing of commercial software and development of open-source software for chemical structure database partially solve the difficulties, problems with the developing cost still remains. In this decade, agile software development grabs attention in the software development community. To develop chemical structure database system in an agile way, we try to develop the system with Django, a Python-based web framework and RDKit database cartridge, an add-on of PostgreSQL for chemical structure search. Two undergraduate students in school of pharmacy with no programming experience can develop a chemical structure retrieve web service within several months. It shows this combination is a useful tool for developing chemical structure database systems.

View full abstract

In fragment-based drug design (FBDD), active fragments are identified using experimental methods and medicinal chemists combine then to design new compounds to synthesis. Recently, in silico methods are also proposed for fragments identification and following compound design. In silico fragment identification methods may find a number of active fragment candidates on the active site, resulting in the combinatorial explosion of fragment lists that are combined to design new compounds. Therefore computationally inexpensive ways to select fragment lists are required. To find an appropriate threshold for selecting, we try to analyze all the triplets from a fragment library for in silico FBDD against an existing compounds DB and a drug-like compound DB. It shows whether each combination exists in known compounds or compounds with a specified physicochemical property. The result shows that the hit count of substructure search of fragment list against a compound DB may be used as a criterion for fragment list selection.

View full abstract

Molecular shape and size can be important factors in the control of organic reactions. In this presentation, we will report a statistical method to extract important information of optimal molecular shape and size for a reaction. We employ finely divided molecular structures (0,1 vectors calculated from 3D molecular structures) as steric descriptors. By using regularized regression (LASSO/Elastic Net), we perform regression analysis between reaction outcomes and the steric descriptors. Based on the coefficients of the resulting regression models, we can visualize important partial structures for the reaction.

View full abstract

Biosynthesis of Wax-esters (WS) in Euglena gracilis is associated with the quantity of ATP. In the present study, we measured the quantities of several metabolites in several conditions. One of the key results is that he quantity of WE is recovered when the level of ATP is recovered in an anaerobic condition. Thus biosynthesis of ATP is highly related with that in WEs. We tried to clarify the relation of quantities between ATP and WS taking TCA cycle into consideration.

View full abstract

We can recognize the diverged ring structures in secondary metabolites. In the present study, we used 48,681 metabolites accumulated in KNApSAcK DB to estimate variation of cyclic structures based on simple cyclic graph and obtained 5,905,755 types of ring systems. We carried out the significant ring systems associated with metabolic pathways derived from those diverged ring systems.

View full abstract

We have designed healthy cuisine DB associated with function based on scientific evidence and accumulated 1,539 recipes from cooking books authorized by medical and pharmaceutical researchers. In the present study we systematized the relationship between foods and functional property in human health care.

View full abstract

In order to prevent three major diseases in Japan, we constructed food healthcare database. We have accumulated recipes consisting of 221 efficacies for 468 foods.In the present study, we introduce designe of cuisine databse focussed on the three major diseases.

View full abstract

The one topic is on the fragmentation library of ESI-MS/MS data. MassBank, a mass spectral database, collects high resolution mass spectra of ESI-MS/MS that analyzed biochemical compounds (‹ 1200 u) on various analytical conditions. The molecular formula of fragment ions, the chemical bonds cleaved, and their degradation sequences were exactly assigned and were collected on the MassBank fragmentation library. An application of the library will be discussed. The other is on the hashing of spectral data. These days several mass spectral databases are available. It is, however, hard to search mass spectra through the databases because each database adopts its own record formats different from others. This makes their users hard to search the target data.　We developed a hashing method, SPLASH, of mass spectra by an international collaboration. More than 500 thousands data have implemented the SPLASH data currently. The hashing algorithm will be shown.

View full abstract

In the realm of data visualization, reducing complex data sets to 2-D maps is a common practice used for reconciling information and for identifying meaningful patterns and clusters. Within the plethora of different methodologies and criteria for defining optimal maps, however, one must be aware on how to assess the performance of such visualization. This work focuses on Generative Topographic Mapping, a well-known nonlinear visualization methodology, to investigate and propose different indexes used for defining optimal latent 2-D maps. More specifically, this work focuses on criteria used for obtaining optimal hyperparameters used for map training. Common criterion such as RMSE is used, but also a new strategy relying on Root Mean Square Error of Midpoint (RMSEM) and its association with Nearest Neighbors (NN) is proposed. In order to evaluate their performance, an artificial data set and Tennessee Eastman Process (TEP) were used as case studies, highlighting the potential of the proposed criterion for defining more reliable and meaningful data visualization.

View full abstract

Because catalytic activities of metal nano clusters depend on the composition, size, environment, and structural isomers of the cluster, it has been difficult to elucidate the primary factors in their catalytic activities. In this study, we attempted to extract the factors in the catalytic activity using the sparse modeling techniques and the systematic quantum chemical calculations assisted by the automatic search of reaction pathways. In particular, the transition state energies for NO dissociation on Cu13 clusters were modeled with the orbital energies and local indices by the LASSO, SCAD, and MC+ regressions. It was found that the transition state energy negatively correlates with the LUMO energy. The SCAD and MC+ regressions could generate more compact and better models with higher correlation factors than the LASSO regression.

View full abstract

In investigations of crime scenes, scientific analyses of biological evidences are conducted in order to identify suspects. In particular, body fluid samples are frequently collected and discrimination of body fluids is essential in criminal investigations. Recently, spectroscopic analysis methods of body fluids have been actively developed in forensic field because they are non-invasive and practically easy compared to biochemical methods currently in use. However, establishments of precise and reliable analysis methods of such complicated spectra have been challenging issues. In this study, we developed an analysis method for discrimination of antemortem and postmortem blood samples using FT-IR ATR spectra and Partial Least Squares-Discriminant Analysis (PLS-DA). In addition, we tried to discriminate blood stains using Multivariate Curve Resolution method for decomposition of substrate signals. By extracting blood signals appropriately and selection of component number in PLS-DA model, we achieved a good discrimination precision of antemortem and postmortem blood stains.

View full abstract

For the safe and stable operation of industrial and chemical plants, it is necessary to monitor and control their operating conditions. Because of the huge amount of operating data in plants, data-based process control systems have received considerable attention in recent years. Controlling each process variable independently is inefficient, because there are many process variables that must be controlled. One practical solution is multivariate statistical process monitoring (MSPM) which monitors multiple process variables and their relationships simultaneously. Principal component analysis is widely used as an MSPM method. However, it cannot consider nonlinearities between process variables and multimodal data distributions. In addition, although process faults can be detected, it is difficult to estimate process states in detail. Therefore we developed a new MSPM method to detect process faults and to estimate each process state in industrial plants simultaneously by combining PCA and ensemble learning. Many PCA models, each of which represents local process state, are prepared using initial database. Fault detection and process state estimation are performed by checking similarity between a query and each PCA model. We demonstrate the effectiveness of the proposed method using numerical simulation data in which actual industrial process is simulated.

View full abstract

Basic process of drug bioactivity that is pharmaceutical effects and side effects is as follows. Compound interacts with target proteins. The activated proteins transmit signals into the cell, that is called signaling pathways. The signaling pathway results in cellular phenotype changes such as cell death, proliferation, differentiation, and so on. Our motivation for rational drug design is to understand molecular mechanism from drugs to phenotypes, involving interactions between drugs, target proteins, pathways, and phenotypes. In this study, we constructed a multi-layer model for predicting target proteins using chemical and biological data such as drugs, proteins, and phenotypes. We proposed a machine learning method that predicted associations between drugs and phenotypes, and so target proteins involved in the phenotypic changes.

View full abstract

Water plays a significant role in binding process between a protein and a ligand. However, water molecules are often underappreciated or even ignored in protein-ligand docking. Usually, free energies of active-site water molecules are substantially different from those of waters in the bulk region. The binding of a ligand to a protein causes displacement of these waters from an active site to bulk, and this displacement process contributes to the free energy change of protein-ligand binding significantly. For instance, water molecules in a hydrophobic region of protein which cannot make appropriate hydrogen bonds are energetically unfavorable, and the displacement of such water molecules into bulk region earns a substantial contribution in binding free energy. The free energy of active-site water molecules can be calculated by grid inhomogeneous solvation theory (GIST), using molecular dynamics trajectory (MD) of a target protein and water. In this work, we combined GIST-based desolvation energy with the scoring function of AutoDock4 and discussed the accuracies of scoring and binding pose prediction.

View full abstract

Recently, many guidances about drug development and applications have been published by authorities in Japan, USA, and EU. The guidance published in 2012 by FDA attracts industrial attentions so much. The experimental methods and evaluations for DDIs in detail, for example CYP inhibition and induction, are described there. Then, the pharmaceutical company researchers should not only obey with this guidance but also try to avoid DDI beforehand prior to drug development stage. Here, we summarize the problems to be cleared in DDI, and the method for the prediction of DDI using molecular informatics. Then, we shows 3D-QSAR model for pregnane X receptor (PXR) as an example. The PXR that induces the expression of xenobiotic metabolism and transporter genes such as cytochrome P450 (CYP) 3A4 and P-glycoprotein (P-gp) is a nuclear receptor activated by agonists. Novel drug candidates that are PXR agonists may cause severe DDI and should be eliminated from the drug discovery process. In this study, we used MD simulation, MM-GB/SA, and CoMFA to predict the potency of drug induction by PXR quantitatively.

View full abstract

Descriptors, calculated properties of compounds, are generally used as features for prediction models (such as Multiple Regression Analysis models) in Quantitative Structure-Activity Relationship (QSAR) studies. Plenty of software packages regardless of commercial or non-commercial were developed to calculate such descriptors. PaDEL-descriptor, a well-known free software, which is referenced more than 300 times, can calculate numerous kinds of descriptors and is widely used. However, we found that there are several problems within it. To overcome its disadvantage and provide correct calculations, we developed a novel software package named mordred, which is implemented in Python language. It can be used as a module of Python2 or 3. Python is increasingly used in machine learning, especially in neural networks nowadays. Therefore, the Python coded mordred is easy to be used in constructing machine learning models. At current time, more than 2,000 descriptors including 2D and 3D-descriptors can be calculated in Command line interface (CLI), Web Application, and Python module. Moreover, mordred with its documentation can be obtained from github (https://github.com/mordred-descriptor), are released under the BSD3 license, and can be freely used including commercial purposes and modifications.

View full abstract

Phenylketonuria (PKU) is an inborn error of phenylalanine metabolism due to the reduction in enzymatic activity of phenylalanine hydroxylase (PAH) caused by mutations. In recent years, small compounds that recover the enzymatic activity of mutant PAHs, pharmacological chaperone (PhC), have been reported. These compounds are expected to have potential for novel therapeutic agent for PKU treatment. For the molecular design of more effective PhC drugs, it is important to understand the mechanisms of the reduction in enzymatic activity by mutation and its restoration by PhC binding. In the present study, these mechanisms were investigated using molecular dynamics simulation. As a results, it was shown that the thermal motion of the PAH active site is increased by the mutation. The results suggest that the reduction in the enzymatic activity is caused by the increment of the thermal motion of the active site due to mutation.　 In addition, it was shown that the enzymatic activity of mutant PAH is recovered by stabilization of the thermal motion of the active site by PhC binding.

View full abstract

Hydrogen evolution reaction (HER) on Ag electrode in neutral and alkaline solution is significantly accelerated by the addition of 4,4’-bipyridine (BiPy). The mechanism of the HER acceleration is discussed at a molecular level by using surface-enhanced infrared absorption spectroscopy (SEIRAS) combined with electrochemical techniques and density functional theory (DFT) calculations. Simultaneous SEIRAS and electrochemical measurements revealed that the onset potential of the accelerated HER coincides with the 1e^- + 1H⁺ reduction potential of adsorbed BiPy to yield monohydro BiPy (BiPy-H^*) and that both BiPy and BiPy-H^* are oriented perpendicular to the electrode via one N-end. The detailed electrochemical data indicates that BiPy-H^* is further reduced via an additional 1e^- +1H⁺ step. DFT calculation showed that generated N,N-dihydro BiPy (BiPy-H₂) is unstable and easily decomposes to BiPy and H₂. Since adsorbed BiPy is readily reduced again to BiPy-H^* in the potential range of HER, the reaction process is catalytically cycled.

View full abstract

Bioisosteres are partial structures that can be replaced without altering the primary activity of a medicine. Bioisosteres are most often classified by chemical structure. Since a chemical structure depends on its electronic structure, it is expected possible to classify the bioisosteres to electronic-structure calculations. In this study, we attempt to classify bioisosteres based on similarity of electronic structures represented by an expanded set of electronic descriptors.

View full abstract

We evaluated the similarity of carotenoids derived from marine organisms using descriptors obtained from the results of electronic state calculations and classified them using the obtained similarities. These classifications were then compared to the biological taxonomy of the carotenoids. As a result of a degree of correlation was found between the obtained classifications and the biological taxonomy of the carotenoids.

View full abstract

We study the spatial distribution of amino acid residues that strongly interact with a ligand bound to the binding pocket of a protein. This analysis is based on the fragment molecular orbital (FMO) calculations. Taking the binding pocket of the protein as the center, we develop a method where the binding pocket was numerically evaluated on the basis of results from FMO calculations. By using the newly developed method, homology of different ligands and binding pockets are analyzed for characterization and classification.

View full abstract

For the purpose of materials optimization and cost reduction in hole transport materials for perovskite solar cells, we carried out electronic-structure informatics research based on quantum chemical calculations. We performed a systematic analysis on hole transport materials used in perovskite solar cells based on the results of electronic-state data-mining and evaluations of electronic similarity to elucidate structure-related correlations.

View full abstract

Molecules which selectively interact with biogenic amine receptors are investigated by using the electronic-structure informatics method. Through molecular similarity search on the basis of electronic-structure calculations, classification of agonists and antagonists was carried out. In this presentation, we will introduce more electronic descriptors than those defined in our previous presentation.

View full abstract

We investigated the stable structure of graphene oxide (GO) which were generated by randomly selecting the C-C bond in graphene models. Throughout the calculations, the most stable isomer for each stoichiometry was predicted. At the same time, the averaged energy among the isomers was evaluated. For systems having the lease number of oxygen atoms, we investigated 150 isomers. In this calculation we observed that the calculated energy was almost convergent. In GO models including up to 96 carbon atoms with 60 % oxygen content (for the number of carbon atoms), we investigated 30 isomers for each system. By calculating oxygen binding energy, we found that GO becomes relatively stable when the oxygen content is about 40-54 %, which is in reasonable agreement with the experimental value.

View full abstract

This poster draws a decision tree for a trial to extract important descriptors in order to predict the ignition point.

View full abstract

In the molecular dynamics simulation of protein, the RESP charge determined for each residue is used as the atomic charge. We have concluded that more reasonable atomic charges including the interaction in protein conformation were obtained by using the result of canonical molecular orbital calculation in the whole protein. In this study, by the penalized regression method, such as the Ridge regression and Lasso regression, the charges of the buried atom were mathematically evaluated.

View full abstract

In photo-excitation and -deexcitation of a molecule, it is very important to consider the change of occupied molecular orbitals between ground and excited states. Meanwhile, the vibrational structure is another important factor. In the current study, we calculated both molecular orbitals for electrons and nuclear vibrational spectra based on Franck-Condon principle using density functional method. In Franck-Condon calculation, we used zeroth order approximation of electronic transition dipole moment in excited states. The target of current study is an aromatic molecule which has a fused ring system containing nitrogen and quinone structure. We resolved absorption and emission spectra using molecular orbital functions including vibrational structures between ground and excited states. In our works, we calculated fluorescence spectrum in emission process. And in comparison of computed and experimental results, good reproducibilities of the waveform in both spectra were obtained. In addition, we obtained high reproducibility in emission intensity.

View full abstract

Real-time propagation (RT) of time-dependent theories, such as time-dependent Hartree-Fock (TDHF) method and time-dependent density functional theory (TDDFT), have been applied to theoretically describing electron dynamics. However, RT calculations are computationally demanding, because of evaluation of time-evolution operator by conventional numerical integration such as the Runge-Kutta method. In this study, we developed the three-term recurrence-relation (3TRR) method as an efficient time-evolution method for electron dynamics, being inspired by the real-wave-packet method for nuclear wave packet dynamics with time-dependent Schrödinger equation. The basic formula of this approach was derived by introducing a transformation of the operator using the arcsine function. Since this operator transformation causes transformation of time, we derived the relation between original and transformed time. We applied this 3TRR method to equation of motion for density matrix in RT-THDF/TDDFT. 3TRR method achieved about four times faster RT-TDHF calculation than conventional fourth-order Runge-Kutta method.

View full abstract

A lot of proteins retained activity in vivo forms quaternary structures. In the protein-protein docking to predict the quaternary structure, fast-evaluation of the protein-protein interactions is important. In our previous works on the development of a fast search method for protein-peptide ligand complex structures (peptide-ligand docking), we applied two statistical coarse-grained potentials, Krishnamoorthy’s original and our improvement, as the score evaluation function for quaternary structures of a protein. In this work, to review whether these potentials can be employed as the stability evaluation of protein quaternary structures, many polymorphic structures of the phosphofructokinase tetramer are generated and compared by using AMBER force field and their statistical coarse-grained potentials.

View full abstract

The carbon dioxide capture and storage (CCS), which is a technique to reduce an amount of emission of CO₂, has many attentions. The chemical absorption method, which is one scheme of the CCS technique, consists of two processes: a separation process for absorbing CO₂ in a basic absorbent and a collection process for emitting high concentrate CO₂ by high temperature. Many amines have been studied and developed to reduce the energy in the CO₂ absorption. To explore an amine with a high efficiency in the CO₂ absorption, the time courses in concentrations of chemical species for each amine give important information. In this study, we develop a reaction predictor to reproduce the experimental data, which uses elementary processes in a CO₂
-amine reaction and the swarm intelligence. Furthermore, equilibrium constants for elementary processes are estimated from reaction Gibbs energies obtained by quantum chemical calculations in order to predict loading dependencies of chemical species in any amines.

View full abstract

In precision agriculture, near-infrared (NIR) spectroscopy is a useful tool to predict soil properties. When applying statistical learning methods to near-infrared spectroscopy, wavelength selection becomes an inevitable issue due to the wide range of wavelengths measured. By comparing linear regression method and nonlinear regression method combined with variable region selection, it could be confirmed that some variables in NIR spectroscopy are nonlinearly related to soil properties.　Additionally, soil properties are quite different for each area, so this large inter-area variability makes the prediction of new areas difficult. From this premise, multiple Bayesian ensemble regression is proposed for solving this applicability domain problem. For NIR spectroscopy data coming from different sources, prediction results of the proposed method were shown to be significantly superior to traditional modeling methods.

View full abstract

Molecular mechanics (MM) calculations such as a systematic sampling approach reported from Merck & Co., Inc. are usually used for conformational search. However, in a collaborative study with a Japanese pharmaceutical company, we found that the predicted vibrational circular dichroism (VCD) spectra of more than 30% of pharmaceutical candidates, where the spectra were predicted using routine procedures, could not reproduce the observed spectra; in addition, several population-rich conformations predicted using density functional theory (DFT) calculations were missed by routine conformational searches with MM calculations. An auto-generation program of initial structures for DFT calculations using chemical structure search for common 3D fragments with SMARTS descriptions (ConfFragGeneration) was constructed. The initial structures of cholesteryl pelargonate for DFT calculations were automatically generated from the database of optimized population-rich conformations of cyclohexyl nonanoate and cholesteryl chloride by using the ConfFragGeneration program for the VCD analysis of the cholesteric liquid-crystal phase.

View full abstract

If the molecular energy obtained from the quantum chemical calculation can be modeled with local explanatory variables such as atomic charges, it is possible to drastically reduce the computational cost. We attempted to model the total energies of C₃H₆O₂ isomers by using Mulliken, Lowdin, natural, and ESP charges as the explanatory variables. By applying the MC+ regression, we could obtain a model showing good agreement with the calculated energy with a couple of explanatory variables. Applying this modeling scheme to the isomers of C₈H₁₀O₂, C₉H₁₀O₂, C₉H₁₀O₃, and C₁₀H₁₂O₂, however, resulted in failing to obtain good models. By adding the numbers of 3 to 8 membered rings as the explanatory variable, we could obtain good sparse models. In addition, it was found that all these sparse models contain the largest natural charge and the number of 6-membered rings as the explanatory variable.

View full abstract

It is significant to understand the physical properties and dynamics of chemicals in the environment. However, experimenting these properties for such huge amounts of compounds is almost impossible since it takes too much cost and time. Hence, we often use the QSPR (Quantitative Structure- Property Relationship) as an alternative approach. The purpose of current study is to construct a prediction model for the photolytic half-lives and meanwhile to estimate some physicochemical properties which are important to predict the photolytic half-lives. We used the Random Forest, a widely used machine learning algorithm. As a result, HOMO_LUMO gap, charge-related descriptors, and some other descriptors were selected to be significant. The accuracies of the model constructed with the ten most important descriptors were approximately 66% and 73% for training and test sets respectively.

View full abstract

Leveraging rapid pervading smartphones and tablets equipped with camera and internet communication can be expected to contribute various advantageous effects on the chemical research and education. A molecular visualization application for such mobile Android devices, "TUTmol", can display the 3D molecular structure with various drawing methods and is possible to operate the scaling and rotation by touching on the screen directly. In this study, as one of the TUTmol enhancements, a character recognition technique using the augmented reality (AR) technology has been introduced. By this new function, TUTmol can recognize PDBID specified by four alphanumeric AR markers, automatically download the corresponding PDB file via Internet, and then visualize the molecules with the PDBID. The management function with respect to the lists of many molecular structure files downloaded has been introduced.

View full abstract

Root of Angelica acutiloba is assigned as an important Kampo medicine. We compared the constituent difference of leaves between feeding damage by insects and control. In the present study, we clarifies significant differences in quantities of metabolites such as gamma-Terpinene, Falcarindiol, Vitamin E 1-tetradecanol, Ligustilide, L-Tetradecyl acetate based on principal component analysis. Based on the results, we also propose differences in metabolic pathways of secondary metabolites between the feeding damage and control.

View full abstract

In the present study, we examined relationships between chemical structures and biological functions. Initially we have extracted 17,717 pairs in the level of 95%-structure similarity pairs based on COMPLIG algorithm from 50,037 metabolites. Secondly we classified metabolites by graph clustering algorithm (DPClus) using 17,717 pairs and obtained significant relationships between chemical structures and biological activities.

View full abstract

It is well known that Euglena gracilis　produces wax-esters in anaerobic condition, but the regulation system in transcription level does not completely clarified. In the present study, we applied Bayes ANOVA to understand to discuss relationship between primary metabolic pathway and accumulation of wax esters.

View full abstract