'Big data' is ususally characterized by 3V (Volume, Velocity, and Variety), 4V (+ Veracity), or 5V (+Value). However, it is more reasonable to understand that it symbolizes a paradigm shift from mission-driven R&Ds to data-driven R&Ds. This talk will clarify why this shift is triggered, what are the potential application areas in this shift, how is the state-of-the-art of its enabling technologiesis, what kind of technologies are still missing, and what issues are left to be solved.

View full abstract

Many software programs visualizing organic molecules such as GaussView can easily represent chirality by one-button operation. However, there is no program to describe the conformations of organic molecules such as trans and +gauche automatically. For the determination of absolute configuration of chiral organic molecules using vibrational circular dichroism (VCD) spectroscopy, the accurate conformational information is necessary, and especially in the case of pharmaceutical candidates, a lot of conformations must be arranged to narrow down the stable conformers. The program of conformational code for organic molecules (CCOM) was applied to VCD conformational analysis of pravastatin sodium. The output of CCOM was arranged as the sdf file format containing the information of conformational codes and Gibbs free energies (if calculated). The auto-selection of the Gaussian outputs using conformational code to exclude the same optimized geometries for the population-weighted VCD spectra was also carried out by using this program. Further, the characteristic conformational 3D fragments of mevastatin in X-ray structure can be easily retrieved together with other molecules in combination with Simplified Molecular Input Line Entry Specification syntax (SMILES) and SMiles ARbitrary Target Specification (SMARTS) strings.

View full abstract

To realize quantum-chemical calculations with chemical accuracy (> 1 kcal/mol), efficient estimations of accurate electron correlation energies in a complete basis set (CBS) limit are important. Several extrapolation schemes to CBS limit, and composite schemes in combination with energies calculated with some levels of theories and basis sets have been proposed. This study proposes efficient estimation schemes of electron correlation energies in a CBS limit at the CCSD(T) level utilizing informatics technique.

View full abstract

To observe molecular motion in the liquid state, we performed molecular dynamics simulation on the unit cell of the face-centered cubic lattice by Lennard-Jones potential. The results show that the vibrational motion and diffusional one are observed in the small cell with the only 4 molecules in the basic cell as in the large cell with the 864 molecules.

View full abstract

Accuracy and applicability domains (ADs) of regression models are discussed in our presentation. Generally, we construct a regression model so as to prevent overfitting to training data and to have highly predictive performance for diverse compounds. However, an overfitted model must have highly predictive ability only within an AD, which is narrowly limited. In this study, the aqueous solubility data set was analyzed to compare performance of regression models while considering their ADs. Support vector regression (SVR) was used as a regression analysis method and hyperparameters of SVR changed. The ADs were set based on data density. There existed two types of SVR models. One is well-constructed SVR models that could predict solubility values for diverse compounds. The other is overfitted SVR models that seemed to have bad predictive ability but provided better prediction results for compounds within the ADs than the other type of SVR models. It was confirmed that overfitting itself was not a problem and we could operate overfitted models by setting their ADs appropriately.

View full abstract

Protein–ligand docking is an optimization problem, which aims to identify the binding pose of a ligand with the lowest energy in the active site of a target protein. Energy landscapes of the scoring functions are usually complicated and exhibit a rugged funnel shape. Hence, successful docking simulations require an efficient optimization algorithm. Various optimization algorithms have been developed for the protein–ligand docking. Genetic algorithm (GA) based approaches are the most general. On the other hand, some variants of particle swarm optimization (PSO) are reported that improve docking accuracy over GA based approaches. In this study, we attempted to apply a novel optimization algorithm, called fitness learning-based artificial bee colony with proximity stimuli (FlABCps), to the protein–ligand docking. The artificial bee colony (ABC) algorithm is a simple and powerful optimization algorithm for the multi-dimensional and multi-modal functions, inspired from intelligent behaviors of honey bee swarm. It has been reported that the ABC based algorithms give better results for various optimization problems than the conventional algorithms. Simulation results revealed that FlABCps improved the success rate of docking, compared to four state-of-the-art algorithms. The present results also showed superior docking performance of FlABCps, in particular for dealing with highly flexible ligands with a number of rotational bonds.

View full abstract

We aim at developing an analytical method for data mining which would be useful for molecular design and discovery in Chemistry and Pharmaceutical Sciences. We have developed a program for electronic-structure data mining to automatically discover a mathematical correlation function between an electronic descriptor and a chemical property. This program uses the electronic factors obtained by electronic-structure calculation, by performing a simple linear regression analysis and multivariate analysis is to explore the regularity of chemistry. Electronic factors to be used for analysis are orbital energy, excitation energy, oscillator strength, transition dipole moment, dipole moment, ionization energy, and the electron affinity. Using this electronic-structure data mining, an analysis was carried out for pyrazinamide derivatives that are anti-tuberculosis drugs. A large number of function fittings, it has been shown that the best fitting is obtained when the orbital energy of the molecular orbital localized on the benzene moiety is chosen as a descriptor.

View full abstract

Inverse-QSPR is a method to propose chemical structures having desired properties by inversely analyzing regression models. In general, QSPR models are neither surjective nor injective, so it is difficult to define the pre-image of these models. The authors once proposed to solve this problem by using probability distribution. In that method, Bayesian theorem played a crucial role to retrieve posterior distribution of independent variables given an objective variable value. Although that method works well for some case studies, regression models must be constructed using Multiple Linear Regression (MLR). This premise, however, does not fit many cases. To overcome this limitation, herewith we have developed two methodologies for inverse-QSPR. One is using different MLR models for each cluster, defined by Gaussian mixture models. The other is using Gaussian mixture regression. Both of them can analytically define the posterior probability distribution of independent variables for inverse analysis. We investigated how both of them capture features of data with nonlinearity and showed they worked well at least for a simulation dataset.

View full abstract

We improved a method of molecular-shape analysis developed in our group. We tried to automate a procedure consisting of several independent tasks. In addition, the computational cost was largely reduced by restricting the sampling points within a sphere centered on each atom. The efficiency was confirmed through applications to water molecule: the number of sampling points was reduced to a quarter of the initial set. By this improvement, application to large molecules has become possible at low cost. The improved algorithm was applied to investigate molecular similarity among bioisosteres suggested empirically. The similarity was numerically evaluated by referring to interaction between an ion probe and a target molecule.

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. The DB shows all the information in Web browser such as FireFox, Chrome and so on. In this data base, we are also gathering information of chemical reactions such as reaction equations and optimized coordinates of transition states (TS) and others. We can use a TS coordinate in searching those of similar reactions by use of our substitution method. It is true that quantum chemical calculations can be applied to find TSs for synthetic routes created by SRDSs (Synthesis route Development Systems) or organic chemists before starting experimental works. This procedure, theoretical studies prior to synthesis experiments, is called “in silico screening”. 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 and to realize “Quantum Chemistry-Assisted Synthesis Route Development”.

View full abstract

As the number of potential compound-protein interactions (CPIs) that could be assayed is essentially infinite, brute-force experimental screening for CPIs is highly wasteful. Attention has thus been given to CPI prediction models that can guide researchers to fast lanes for hit discovery. Existing CPI prediction models have mostly used a curated database of interactions for building a single fixed model, with the Support Vector Machine (SVM) often used for model construction. On a dataset of 100,000 CPIs, the SVM can train a model in less than one day. Yet the size of available datasets can be in the millions, and since SVMs require an exponential increase in resources, model construction on such datasets is infeasible. We investigated the ability of Deep Learning to handle large volumes of CPIs that cannot be processed by SVMs. Deep learning does not require learning on all input data at once as in the standard SVM, but rather the model is iteratively tuned over the course of data input. We evaluate the learning error rate as a function of the number of learning iterations, our method based on Deep learning outperformed the method based on SVM.

View full abstract

In the early stage of drug discovery, it is important that the virtual compound library used for computational compound screening contains 1) lots of novel scaffolds in synthetically accessible compounds and 2) useful information for chemical synthesis. We have generated 1,700 million of non-redundant chemical structures from ten million initial chemical structures by applying the structural transformation rules mainly based on named reactions. As all the chemical structures in our virtual library have the synthetic route information, they are instrumental in planning of chemical synthesis strategy. Besides the information of chemical structures and synthetic route information, the physical properties based on molecular descriptors and the fingerprint for the compounds are added to the library: these information are available for searches on substructure, similarity, and druggability. We will develop the highly effective virtual compound library and additionally the sophisticated visualization technology for understanding the chemical space of our virtual compound library.

View full abstract

Soft sensors have been widely used in chemical processes to predict values of difficult-to-measure process variables online. If the relationship between explanatory variables X and an objective variable y is changed by catalyst deterioration, change of product and so on, prediction accuracy of a soft sensor is reduced. This problem is called degradation of a soft sensor model. To overcome the degradation, many adaptive soft sensors have been proposed. In this paper, we aim to improve prediction accuracy of just-in-time (JIT) models. JIT models are constructed with only data close to a query or with all data having weights according to similarity with a query. If the type of degradation is shift of y-value, prediction accuracy of JIT models is reduced since data with similar X-values but different y-values are mixed in database and the relationship between X and y is not consistent. To resolve this problem, we propose a management method of database, called JIT database. JIT database were constructed only the latest data in all areas of X. By constructing JIT model from JIT database, improvement of prediction accuracy was achieved for simulation data.

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 of allyl-group. The reaction mechanism is investigated by the stable structures of the phenoxenium ccation as a reaction intermediate by using the Hamiltonian algorithm.

View full abstract

We developed an automatic modeling system of the reaction mechanisms in chemical vapor deposition processes by implementing the bio-inspired algorithms to the inference engine of the system. In addition, we evaluated the algorithms using both the median and the quartile deviation of the convergence values of the fitness function and calculation time. CMA-ES showed the best performance among the algorithms from the aspects of accuracy, convergence stability and calculation time.

View full abstract

We propose a straightforward method to extract the essential information regarding secondary structure interconversions from molecular dynamics simulations of proteins. Based on the method, named secondary structure principal component analysis (SSPCA), we have been studying about intrinsic disorder proteins, structurally ambivalent peptides, and chameleon sequences, which lack a strong intrinsic secondary structure, thus promoting alpha-sheet/beta-strand conformational conversions. We applied the SSPCA method to prion protein (PrP). In the early stage of prion diseases, secondary structure conversion in PrP causing beta-sheet expansion facilitates the formation of a pathogenic isoform with a high content of beta-sheets and strong aggregation tendency to form amyloid fibrils. The definite existence of a PrP isoform with an increased beta-sheet structure was confirmed in a free-energy landscape constructed by mapping protein structural data into a reduced space according to the principal components determined by the SSPCA.

View full abstract

Polyene chemistry has attracted little attention of organic chemists due to the scarcity of known compounds and lack of novel properties. The concept of cross-conjugation was also overlooked by them. However, recently a number of new compounds and such novel properties have been discovered that attract the interest of physicists. “Cross-conjugation” has become one of the most important keywords in these areas. We have already noticed the importance of cross-conjugation and began to reconstruct the polyene chemistry from theoretical stand-point. There a cross-conjugated hydrocarbon is a non-linear acyclic and monocyclic conjugated hydrocarbon constructed by fusing two or more linear polyene networks, all having one Kekulé structure. Then radialenes, fulvenes, and quinoides are included into this category. The stability of these molecules is highly correlated with the topological index Z, and also well correlated with the mean length of conjugation L. These discussions are extended to the theoretical justification and indication of the coverage of “organic electron theory”.

View full abstract

A new chemical structure elucidator, called CAST/CNMR Structure Elucidator, has been developed. Since the new elucidator uses a structure-spectrum database, its capability depends on the quality and quantity of the database; however, the elucidator produces a correct structure if all fragments are included in the database. As is known, computer-assisted chemical structure elucidation (CASE) has been intensively studied since the first use of computers in chemistry in the 1960's. Nonetheless, there still remain challenges as to how to deal with large-scale databases that are being continuously updated and how to obtain candidate/correct structures as fast as possible. The new elucidator utilizes practically efficient graph algorithms, which in combination of the progress of computers provides a fast elucidation of chemical structures from an NMR spectrum. Its application to the analysis of organic compounds, e.g., structure revision, has been already started.

View full abstract

Interactions between nucleobases in DNA-duplex play an important role in biological systems. In particular, the stacking interactions and many-body effects of interactions between nucleobases depend greatly on the base sequences. Guanine (G) changes into 8-oxoguanine (8OG) after the oxidation. 8OG can take either keto or enol forms by keto-enol tautomerization. The hydrogen bonding pattern of 8OG(enol) is different from that of 8OG(keto). In other words, 8OG(enol) can pair with Thymine (T). Thus, 8OG can form 8OG(keto)-C pair and 8OG(enol)-T pair by keto-enol tautomerization. Experimentally, it has been reported that the probability of point mutation from G to A becomes higher when G changes into 8OG. In this study, we construct B-DNA model including not only Watson-Crick base pairs but also 8OG(keto)-C and 8OG(enol)-T pairs. Then, we discuss many-body effects of the interactions between nucleobases in B-DNA and the induction of point mutation from G to A.

View full abstract

This presentation describes the applications of fragment molecular orbital (FMO) calculations to the elucidation of the mechanism for enzyme-catalyzed reactions in organic synthesis and the computational molecular design of pharmaceuticals. The molecular interactions, based on FMO2-MP2/6-31G calculations, between each of three different lipases, Burkholderia cepacia lipase (BCL), Candida antarctica lipase typeB (CALB), and Candida antarctica lipase typeA (CALA), and the enantiomers of various organic compounds indicated that particular amino acid residues in the respective lipases, such as HIS286 in BCL, THR40 in CALB, and ASP95 in CALA, can play an important role in the chiral recognition of substrate enantiomers. A group of computational life science researchers has established FMO Drug Design(FMODD) consortium in November 2014 that aims to demonstrate the availability of the FMO method as a novel technique for in silico drug design. The activities of FMODD and our FMO calculations on a target protein, renin, with the K computer will also be presented.

View full abstract

Cp^*Ir (Cp*=η⁵C₅Me₅) complexes catalyze efficiently both dehydrogenation from alcohols and hydrogenation to ketones. For example, the dehydrogenation reaction of 1,2-benzenedimethanol using a Cp^*Ir complex proceeds to form 2-(hydroxymethyl)benzaldehyde as a product in aqueous solution. It is possible to consider two types of catalyst cycles, which include Cp^*Ir complex of six and five coordination environment. In the present study, we investigated theoretically the mechanism of the catalytic cycle. All the structures were optimized and characterized as minima or transition states at the B3LYP/BSI level of theory (BSI designates the basis set combination of 541/541/111/1 for Ir and 6-311+g(2d,2p) for all nonmetal atoms). The free energy profiles including solvent effects were calculated to analyze the reaction mechanisms in detail. The SMD continuum model was used to include implicit solvent effects. For the catalytic cycles using Cp^*Ir complex with the six co ordination environment, the barrier heights for the hydrogen transfer was calculated to be 33.1 [kcal/mol].

View full abstract

Chagas’ disease, African trypanosomiasis, and leishmaniasis are some examples of neglected tropical diseases (NTDs) caused by the parasite Trypanosoma. While nifurtimox and benznidazole are currently available for treatment of Chagas' disease, they have major drawbacks, such as side effects and their insufficient effectiveness in the chronic stage. To develop a novel anti-trypanosomiasis drug, we performed virtual screening for 4.8 million small molecules of spermidine synthase as the target protein and conducted an in vitro enzyme assay to determine IC50 value. As a result, we identified hit compounds that inhibit T. cruzi SRM (TcSRM). We also determined the TcSRM-ligand complex structure using X-ray crystallography.

View full abstract

De novo drug design has been an active research area over the past decades. Recent advancement of chemoinformatics technology enabled us to propose various novel approaches for de novo drug design. Many successful examples of practical applications of de novo design in combination with synthetic chemistry and biological assays have been reported as journal papers. Recently, the author and co-workers proposed a novel approach to design drug-like molecules by using molecular fragments and an evolutionary algorithm. Our methodology and the results obtained by computational experiments will be shown in the presentation.

View full abstract

The contribution of cheminformatics for life sciences are mostly pharmaceutical studies, but it is expected that integration of cheminformatics and bioinformatics will lead wider applications. My main research interest is so-called "omics", where we need to integrate big data on genomics and metabolomics. Here I would like to introduce a series of community ecology analysis of chemical interaction networks between insects and plants.

View full abstract

In drug discovery process, in-silico computational methods to efficiently explore and optimize drug candidates among huge amount of organic compounds are strongly required. In this study, we aim to construct a novel computational workflow on “K-computer” that overcomes fundamental problems inherent in calculation accuracy and computational cost. Also, the workflow is constructed reflecting assessments by reseachers in pharmaceutical companies because it should be easy to handle and useful in practical drug development process. Based on these concepts, we have implemented CGBVS (Chemical Genomics-based Virtual Screening method) and MP-CAFEE (Massively Parallel Computation of Absolute binding Free Energy method) on “K-computer”. CGBVS is a virtual screening method based on big data analysis and enabled ultrafast prediction of binding of 18,930,000,000 protein-compound pairs (631 kinds of kinases and GPCRs x 30,000,000 compounds). In contrast, because MP-CAFEE is based on molecular dynamics simulation including water molecules, the method successfully predicted the protein-compound binding free energy (dG) for five sets of inhibitors targeting CHK1, CDK2, ERK2 kinases, urokinase, and GPCR. In this talk, I will report activities of the project and their outcomes up to the present.

View full abstract

Studies about validity and utility of classical trajectories in time propagation of nonadiabatic systems were conducted. At first, classical trajectories which can describe nonadiabatic transitions collectedly are mathematically derived via formulation of nonadiabatic path integral for the time propagation kernel and semiclassical approximations to them. Then, it is found that energy spectra of nonadiabatic systems can be described in the terms of the classical trajectories.

View full abstract

Metal nanoparticles show very interesting properties in solid solution metal systems. For example, Pd/Pt solid solution system can absorb larger amount of hydrogen compared with pure Pd nanoparticles, although Pd and Pt are phase separation system in bulk. However, the mechanism of phase change between core/shell and solid solution of PdPt nanoparticle is still unclear. However, it is difficult to solve above mechanism in metal nanoparticles by using only conventional first-principles approaches. In this study, we calculated electronic structure and optimized geometry of PdPt nanoparticles by using density functional theory.

View full abstract

The adsorption behavior of hazardous compounds on minerals controls their migration process in environments. Thus, it is important to investigate the adsorption behavior of the arsenic compounds on minerals to understand the migration process and to predict future fate of them in environment. In this study, we conducted adsorption experiments, As K-edge X-ray absorption fine structure (XAFS) measurements and DFT calculations for organoarsenic compounds adsorbed on ferrihydrite to understand the effects of substitution of organic functional groups on their adsorption behavior.

View full abstract

It is known that a large number of chemical substances which cause the liver hypertrophy shows the variety of structurally diverse. Therefore, the comprehensive understanding of the toxicological significance of liver hypertrophy has not been elucidated. In this study, we have developed the predicting methods for the liver hypertrophy based on the chemical structure. We made the toxicological database from the risk assessment reports of pesticides, food additives, and veterinary medicinal products that were published by Food Safety Commission of Japan at first. Then, we constructed the prediction Quantitative Structure Activity Relationship (QSAR) model of liver hypertrophy action, based on chemical substance’s descriptors by Deep Learning from this toxicological database. In addition, we compared other machine learning methods, Random forest, and Support vector machines. As a result, the deep learning shows higher performance liver hypertrophy prediction QSAR model compared with other machine learning methods which had the nearly 80% prediction accuracy.

View full abstract

Academic detailing is personalized support for improving both knowledge and clinical decision-making by the latest non-commercial evidence-based data. Doctors select mainly best medicine through viewpoint of experiences and clinical guidelines to prescribe patients. Pharmacists select mainly best medicine through viewpoint of drug characteristic such as pharmacological action, physical chemistry characteristic and metabolite mechanism. Furthermore, pharmacists also require providing medical teams with the latest information including molecular pharmacology, genome information and others. Our university is traditionally fulfilling basic pharmaceutical sciences. However, our pharmacist education program utilizing basic pharmaceutical sciences for patients is definitely not enough to be a highly qualified clinical pharmacist. Eight essential fields which are necessary to support Doctor’s prescription are Biology, Chemistry, Physics, Pharmacology, Pharmaceutics, Pharmacotherapy, Clinical Guideline and Drug Adverse Reaction. Our aim is to integrate these data in order to develop original interface with LSI. Then we would be able to propose a doctor the most appropriate medicine for a patient by using it.

View full abstract

Correlation between electronic similarity and medicinal effect in bioamine and related molecules showing similar medicinal action was investigated. This study was carried out to reveal important factors in determining interaction between a bioamine (e.g. dopamine and octopamine) and its receptor.

View full abstract

We developed an electronic-structure database in which numerical data of electronic structure calculations of medicinal biomolecules were stored. Some computational tools for applications were developed and implemented as a set of programs. The database was applied to evaluate molecular similarity. In the present study, several electronic descriptors were newly introduced. Through the applications, usefulness and improvements of the database and analysis tools are discussed.

View full abstract

Comparison of protein structures is important to understand and predict their functions. The protein structure is complicated, so comparing is not straightforward. In general, the similarity is evaluated by amino acid sequence analyzed by the homology search. In this study, we develop a method that evaluates similarity of protein structure reflecting interactions with surrounding amino acids. For easy comparison of protein structures, we propose a method of sonification which expresses three-dimensional structural features. In the method, surrounding amino acids for one target amino acid are expressed as overlap of sounds generated from each amino acid. In the overall sound, the distance information is taken into account in the sonification. By this method, we can quantitatively evaluate similarity/difference of 3D protein structures in a data in a lower dimension.

View full abstract

This study aims at developing electronic-structure database with knowledge information. The knowledge was designed to be obtained from Web pages on the Internet. For this purpose, we developed some software for Web crawling and Web scraping.

View full abstract

We study rigidity of soft matter by analyzing the time series data from molecular dynamics simulation using the random matrix theory. We construct the time-dependent variance-covariance matrix and analyze the eigensystem. As an example, we present a result for a protein lysozyme, PDBID:1AKI. We find that there are three different time scales involved in the coupling formation of correlated sectors of atoms and two different time scales for the size of the correlated sectors. These five time scales coexist simultaneously.

View full abstract

In pharmaceutical process, process state is monitored and managed by online and non-destructive spectroscopy testing and this methodology is focused on as Real Time Release Testing. Prediction of mole fractions of pure components in mixtures is an important issue for proper control in blending process. In order to predict mole fractions of pure components in mixtures with high accuracy, statistical models must be built from much amount of training data. However, only little amount of data is available, because taking much amount of data costs much. In this study, we proposed a calibration-minimum method that enables to predict mole fractions of pure components in non-ideal mixtures with high accuracy by expressing molecular interaction effect on a mixture spectrum as a function of mole fractions. The parameters in the proposed model equation can be inferred from little amount of data. The molecular interaction effect obtained from the proposed model is supposed to enhance further understanding of molecular interaction in complex mixtures.

View full abstract

For drug development, methods of constructing a library consisting of diverse ligand candidates are required. We focused on de novo design algorithm for exploring chemical space (DAECS) which is the method generating many structures in a selected target area on the chemical space, and improved it. DAECS can generate only structures that exist in a specific area on a subspace set by using ligands data. But it is impossible to consider properties other than the activity and ensure the diversity of generated structures in DAECS. In this study, we introduce an area selection method with the visualization of drug-likeness distribution of the chemical space and a structural conversion method using substructures for solving the problem. To confirm superiority of our methods over the prior study, we performed a case study using a data set of ligands for human adrenergic alpha2A receptors from GVK database and showed the proposed methods can generate more diverse structures on a selected area and generate structures considering their drug-likeness.

View full abstract

In this study, we attempted to construct a web server that would compile libraries of chemical compounds with properties for different purposes. In response to the compound data input, this web server simulates virtual chemical reactions accordingly to generate digital data of new chemical compounds, thereby compiling virtual compound libraries. By imparting objectives to the virtual chemical reactions here, the server completes the compilation of libraries comprising compounds with properties for different purposes.

View full abstract

The first-principles electronic structure calculations were carried out in order to search new catalyst for hydrogen carrier generation. In this work, we focus on MgO-supported Ru catalyst for ammonia synthesis from hydrogen and nitrogen gases. Some first-principles molecular dynamics calculations were implemented. Based on the computational results, simple molecular models were developed for the detailed mechanistic analysis.

View full abstract

In this study, we propose the method for making the classical potential based on the ab initio molecular dynamics. The potential made with our method can describe the anharmonicity in internal coordinate system. We applied the method for the organic inorganic systems, methylammonium lead perovskites.

View full abstract

This poster draws a self-organized map for simple molecules and discusses to apply auto-classification of simulation result.

View full abstract

In previous studies, we obtained pH profile of pH indicators and physicochemical parameters of time-dependent crystallization of drug in solid dispersion by applying chemometric methods using singular value decomposition to Ultraviolet–visible spectroscopy, Terahertz time-domain spectroscopy, and X-ray Powder Diffraction. Pharmaceutical cocrystals are new solid forms with the potential to enhance pharmaceutical properties. To elucidate the enhancement of bioavailability of Indomethacin by formation of cocrystal with Saccarin, we applied our methods to Indomethacin-Saccarin mixed system.

View full abstract

In order to obtain novel compound for drug discovery, it is important that predicting pharmacophore which is common or specific characteristics among active compounds. When target protein-ligand complex structure is available, molecular mechanics method is often used to analyze the interaction between target protein and ligand. However, that method is not applicable to all ligand because there is a limit to the determination of molecular potentials based on atom type, especially of quantum chemical elements such as π electrons. In this study, we focused on dihydroorotate dehydrogenase (DHODH) of Trypanosoma cruzi, a target protein of Chagas disease. We identified pharmacophores using the Fragment Molecular Orbital (FMO) method, which employs ab initio quantum mechanical calculations. We analyzed interaction energy between TcDHODH and orotate, oxonate as a competitive inhibitor of TcDHODH, and 43 orotate derivatives. As a result, 4 common pharmacophore and 1 specific pharmacophore were obtained. Furthermore, compounds that met the specific pharmacophore was suggested to bind to TcDHODH selectively, rather than Human DHODH.

View full abstract

In order to automate the research and development process for chemical vapor deposition (CVD) processes, we developed and proposed a concept of planning experimental design automatically. The concept was applied to identifying the reaction mechanism (reaction model) that indicates the reaction paths both in a gas-phase and on a surface from the reactant (source gas) to the product (film). In addition, we evaluated the validity of the concept which we proposed using the demonstrations by a synthetic CVD process.

View full abstract

We developed a calculation method to reproduce the deposition profiles in chemical vapor deposition reactors using bio-inspired algorithms. In addition, we evaluated the algorithms using both the median and the quartile deviation of the convergence values of the fitness function and average calculation time. REX-star+JGG showed overall the best performance among the algorithms from the aspects of accuracy, convergence stability and calculation time.

View full abstract

Since predicting hydrolysis rates is necessary for estimating fates of chemical compounds in environment, computational or chemometrical procedures of predicting such velocities have been expected to construct. Recently, some category approaches were published in practical use. However, some kinds of chemical compounds which are difficult to classify into the one category in the case of those approaches. In addition, there are many compounds which cannot be classified into a certain category in the case. Thus, we thought a novel model, which uses no or a few categories of chemicals, was needed for estimating hydrolysis rate using many descriptors instead of category approaches. In order to construct such a model, before, “Chance Correlation Problem” was unavoidable. In this study, using Lasso and other our techniques, we tried to alleviate the problem and obtain more predictive models of hydrolysis rates. We used two criteria, Cross-Validation and Basian Information Criteria (BIC) for Lasso regularization. Relatively, BIC provided more predictive results. And when the chemicals were classified into a category, esters, better results were obtained.　Now, better models are being under construction for eliminating more chance correlation descriptors.

View full abstract

Due to development of Computational Chemistry, rich information on various molecules can be easily obtained by electronic-structure calculations. In order to use the computed results efficiently, it is necessary to store the data in well-organized structure. In the present study, we make a well-organized electronic-structure database by which one can search molecules on the basis of their electronic characteristics. The database is also designed to accumulate the knowledge information about molecules that are taken from the Internet and electronic data files from scientific literatures. In this work, we show demonstration of using our database system to classify cancer-related molecules stored in the database. In this application, quantum chemical similarity is referred to for classification, and a clustering method in statistics is applied. Herein we apply eight electronic descriptors such as UV/vis spectrum, IR spectrum, the HOMO-LUMO gap and so on. It is shown that the cancer-related molecules including anticancer agents and carcinogens are reasonably classified on the basis of the electronic descriptors.

View full abstract

Lithium ion batteries (LIB) are widely used in recent years, for example, a mobile PC, a smartphone, and an electric car. There are many design parameters for LIB such as size and material of components. Those parameters should be optimized depending on a use of LIB, which is a very difficult task because the number of combinations of design parameters is numerous and many properties of LIB should be considered simultaneously. We propose an optimization method based on an adaptive experimental design method. The probability P that a result of a simulation on a combination of design parameters will improve on the previous optimal value of an optimizing property and meet requirements of other properties is calculated in the proposed method. The combination of design parameters with the highest P-value is simulated next. The superiority of the proposed method was verified through comparison with a traditional method, and the optimized LIB could be proposed.

View full abstract