We introduce a brief overview of recent similarity searches for large molecular databases.

View full abstract

Semantic Web technology can aid in the elucidation of glycan function by linking life science data. We are developing the GlyCosmos Portal as a portal site of glycoscience and aim to integrate life science data. We have collaborated with domestic and foreign researchers to develop GlycoRDF which is an ontology of glycan information, WURCS which is a carbohydrate structure notation, and GlyTouCan which is an international glycan structure repository. This portal consists of repositories and databases. The repositories include one for glycoconjugates in addition to GlyTouCan. Moreover, we are developing the GlyCosmos database which will store data of glycan-related pathways and glycan structure.

View full abstract

In Fragment-Based Drug Design (FBDD), active fragments are identified using experimental methods and then medicinal chemists combine them to design new compounds to be synthesized. In silico fragment identification methods may find a number of fragment ligands on the active site, resulting in the combinatorial explosion of fragment sets that are combined to design new compounds. Previously, we proposed the knowledge-based method to prioritize such fragment sets based on the number of hits on the substructure search against a reference database. Although it is possible for the method to select hopeful fragment lists, the calculation will be too expensive if the number of compounds in the reference compound library is very large. In this study, we try to construct a subset library (informative compound library) enough representative of the original reference library by clustering and filtering the compounds contained in the database. It can be expected to improve the performance of our fragment-combination selection and also may be used with other FBDD methods.

View full abstract

This talk addresses decision tree and random forest. Using decision tree, I tried to extract effective descriptor and branch rules automatically. Furthermore, I used random forest in order to get better prediction performance.

View full abstract

Solvatochromism of p-nitroaniline (PNA) dissolved in super critical CO₂ was investigated by fragment based quantum chemical (QM) and molecular dynamics (MD) approaches, i.e., effective fragment potential (EFP) and fragment molecule orbital (FMO) methods. It was found that CO₂ molecules cause not only red-shift but also blue-shift of electronic transition of PNA. It was also clarified that the effect of each CO₂ molecule is additive.

View full abstract

The ferulic acid is known to have strong antioxidant properties. In the present study, we have investigate the electronic structures of the ferulic acid and its radical extracting the hydrogen atom from its phenolic hydroxyl group. We have discussed the relation of the results with the radical scavenging activity with the DPPH reagent measured by Sakamoto et al..

View full abstract

It is useful to obtain theoretical prediction data for decomposition reactions in the stability test of drugs. In this study, theoretical analyses and reaction kinetics simulations using free energy were performed for the hydrolysis reaction of aspirin. Reaction analysis of B3LYP/6-311+G(d,p)//B3LYP/6- 31G(d) level of theory estimated ΔG‡=27.2 kcal/mol for decomposition of aspirin with two water molecules transferring protons. Kinetics simulations using this result computed a calculated value of 0.091% which agreed with an experimental value of 0.083% of the decomposition rate in 6 months under the condition of 37 degrees Celsius temperature and 42% humidity. We also conducted a wide range of simulations at a humidity of 30 to 100% and obtained very good agreement with experimental values even under conditions of 60 degrees Celsius temperature and 30% humidity.

View full abstract

Recent remarkable development in computational chemistry has made computer use common practice in the development of novel devices. We have been developing organic-semiconductor-based solar cells using phthalocyanine compounds for p-type and fullerene for n-type as raw materials. However, the physical properties of the structure of the present system are too complicated to have been elucidated in detail yet. We have conducted evaluation of the electronic properties of the phthalocyanine compounds - fullerene C60 system both in the ground state and in an excited state using the density functional method and Hartree-Fock method.

View full abstract

Nuclear Magnetic Resonance(NMR) is measured in the gas phase, solid phase and liquid phase. It is known that NMR shielding constants are influenced bythe surrounding environment. We can predict the NMR shielding constantsin the gas phase, while we cannot predict the NMR shielding constantsonly by using high level calculationin liquid phase. Here we obtainthe solutionstructureusing QM/MM method. We show that structural changes of solute moleculesby surrounding solventsshow marginal effect on NMR shielding constants and taking the solvent into account has an impact on NMR shielding constants.

View full abstract

Beckmann rearrangement of cyclohexanone oxime is widely used for the synthes is of ε-caprolactam. Studies of Beckmann rearrangement using various catalysts so far have also found that cyanuric chloride also acts as a catalyst. In this study, based on the reaction mechanism reported by Deng et al., We investigated whether to proceed with similar reaction mechanism using cyanuric chloride and elucidated the most suitable reaction mechanism using cyanuric chloride.

View full abstract

We have been investigating a method which evaluates potential synthesis routes using transition state (TS) structures for similar mechanisms accumulated in the transition state data base. We are participating in Funatsu project concerning with usage of big data for drug development. Our subproject includes constructing two data bases and accumulating their data. The data bases use two types of search engines. One is the Open babel program calculating Tanimoto coefficients. The other developed by Tabei is the gWT program which evaluate cosine similarities of molecular fingerprints. We developed a program which utilizes coordinates in TSDB as templates for finding TS structures of target compounds. Three atoms in a reference TS structure are used for constructing a candidate structure for TS optimization. This is the key procedure to make good candidate coordinates and to shorten CPU times evaluating possibilities of synthesis routes from synthesis route development programs. In the present study, we report a new program, the achievement of the two data bases at present.

View full abstract

Solubility is one of the important parameters for designing new compounds. However, it is difficult to get solubility of all candidate compound, so quantitative structure-property relationships (QSPR) is widely used. In this study, we use the deep learning, which is popular in the field of informatics. Deep learning model can represent complex non-linear relationship between the descriptor and the property. To simulate the chemical phenomena, we divided the model in two parts. One is feature extraction part, which converts descriptors into solubility-based information. The other one is interaction representation part, which represents the interaction between the solvent and the solute. The proposed method showed high accuracy. For chemical interpretation, we visualized the intermediate output of the model using Isomap. On the visualization map, the similarity of solubility behavior between structures is expressed as distance. This result is useful in the following cases: searching for an alternative solvent or choosing solvent pairs with different solubility behaviors.

View full abstract

Liquid crystals contain rigid mesogenic parts (mesogen) and flexible alkyl parts, which induce the liquid crystalline (LC) state. The aim of this study is to develop models for the prediction of LC behavior applied on a large dataset of rod-like aromatic organic compounds using a QSPR approach. The prediction models are performed using ensemble learning methods with a series of molecular descriptors and chemical fingerprints considering mesogenic parts and flexible alkyl parts of LC structures. This work demonstrates that the complex phenomena of LC phase formation by large variety of mesogens can be effectively modelled using ensemble learning. The best of these models showed high accuracy and F1 score. (90% and 93%) The best model allowing experimentalists to seek the synthesis of predicted molecule that would exhibit the desire LC properties to accelerate the progress in the discovery of new LC materials.

View full abstract

In order to automate the research and development process for chemical vapor deposition (CVD) processes, we developed an automatic experimental design system. The system proposed the optimal experimental conditions using the CMA-ES (Covariance Matrix Adaptation Evolution Strategy) algorithm for 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 system using demonstrations by a synthetic CVD process.

View full abstract

Adaptive models are the techniques of updating the statistical models in a soft sensor, which enable a soft sensor to follow time changes of the relationships between the process data and improve its estimation accuracy. Conventionally, researches about the soft sensor design with adaptive models are developed in the situation that one soft sensor has one adaptive model, though it is, to be considered, difficult to maintain the estimation accuracy in all phases. Then, the previous researches developed the methodical technique to select adaptive models by applying the ensemble learning method to an adaptive model of one side and judging whether to use it or not. In that technique, however, it is a problem that an adaptive model not applied the ensemble learning method could be selected improperly in a scene with inferior estimation accuracy than the other one. Therefore, in this research, the selection technique that evaluates all adaptive models to use based on process state indexes is developed. Through a numerical simulation dataset analysis, the proposed method was verified.

View full abstract

Molecular mechanics (MM) calculations in such a systematic sampling approach reported from Merck & Co., Inc. have usually been 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. A ring structure-enabled auto-generation program of initial structures for DFT calculations (RingFragGeneration) using chemical structure search for three dimensional (3D) fragments represented as maximal common substructures (MCSs) was constructed. Initial structures of rapamycin for DFT calculations were automatically generated from the database of op timized population-rich conformations of fragment molecules by using this program for VCD analysis of the macrolide ring with 49 rotatable single bonds.

View full abstract

In metabolomics, unknown peaks that chemical structures have not been determined are sometimes selected as metabolomic biomarker candidates in disease associated biomarker discovery. Recently, some chemoinformatics approaches using MS/MS spectra with the objective with structural elucidation of unknown peaks have been reported. In this study, we aimed to predict fragmentation by using neutral loss-fragment pairs.

View full abstract

Even if the set of compounds has a common scaffold which is responsible for their bioactivity, most of known chemical similarity measures which represent global similarity can show lower similarity value due to their large residue size. This report describes an overview of newly developed local similarity measure based on subgraph isomorphism and its application to the network-based structure-activity relationship analysis method known as Chemical space network (CSN).

View full abstract

The chemical substance names described in the patent publications have various descriptions and the description of the name depends on the author. Such variation causes hindering information sharing. Auto-extraction of chemical substance names is useful for information sharing. In order to grasp the current situation, we examined and analyzed the occurrence frequency of chemical substance names and description method in patent publication (chemical field). Chemical substances were tagged by type, such as single substance, mixture, polymer, etc., and the frequency of chemical substance names by type in each document was compared. As an attempt to extract chemical substance names, we used morphological analysis for detecting the names, and used the unique characters of the names for selection the names.

View full abstract

ELNES is the powerful method with high temporal and spatial resolution to analyze local atomic and electron structures. However, extracting that information from the experimental spectra requires some difficulties. It requires the theoretical calculations and expert knowledge and cannot deal with big data. Here, we developed the new approach to reconstruct and interpret ELNES spectra. Our method is based on hierarchical clustering and decision tree method, and enable to tie the structure information with the spectrum shape. Moreover, it is noted that our method is not applicable only to spectrum data, for instance, XAS and DOS, but also image data in principle.

View full abstract

We study protein-ligand interaction 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, PDBID:4YDP.

View full abstract

We developed a 3D-QSAR model for CDK2 inhibitors using protein-ligand docking, fragment molecular orbital (FMO) method and PLS regression. In order to obtain reliable complex structures, each inhibitor with known activity was docked to the protein structure prepared from the X-ray structure of protein complexed with the most similar ligand. Then, fragment binding energies upon ligand binding were calculated using FMO method. Finally, using the fragment binding energies as descriptors, a PLS regression model for predicting inhibitory activity was developed and validated. The 3D-QSAR model was able to predict the experimental activity for a variety of CDK2 inhibitors with broad activity.

View full abstract

Amyotrophic lateral sclerosis (ALS), one of intractable neurological diseases, occurs by progressive disorder of specific motor neurons. Although the pathogenic mechanism of ALS has not been elucidated, 5~10 % of all cases have family medical history, which is called as hereditary ALS. Recently, genome analysis for hereditary ALS revealed that aldo-keto reductase (AKR) 1C3 has a low-active variant, the C154Y mutation. AKR1C3 is a reductase metabolizing ketone and aldose using NADP as a coenzyme. Though the C154Y mutation is located at the α4 helix and away from the active site, the enzymatic activity for the C154Y variant decreased to nearly half of that for the wild type. In order to investigate the cause of the decreased activity in the C154Y mutant from the viewpoint of the structure, we performed molecular dynamics (MD) simulations for wild-type AKR1C3 and the C154Y mutant and analyzed these solution structures. As a result, α helix conformation was partially unfolded in the α3 helix of the C154Y mutant. Moreover, hydrogen bond network around NADP was weaker in the C154Y mutant compared with the wild type. This result suggested that the C154Y mutant could not stably bind with NADP and could not exhibit fully the enzymatic activity.

View full abstract

Trimethylamine N-oxide is well known as anosmolyte that is in marine organisms, for example, sharks.However,the mechanism is not completely unraveled even now. In previous studies, it was observed that TMAO had strong interaction with water molecules around it andgot a specific solvation structure. In this study, we treat dichloromethane and water as solvent molecules in order to find out whether or not there are the specific properties in solvation structure of TMAO in other solvent.We created some cluster structures with TMAO and solvent molecules, and optimized these cluster structure. The local minimum structure was confirmedby means of the normal mode analysis. It was found that TMAO interacted strongly with solvent molecules, which coordinated with an oxygen atom of TMAO when solvent was dichloromethane and water.Furthermore, the electron density onbond critical pointof the O…Hwas equal to that ofhydrogen bond inwater dimer.

View full abstract

There has been reported several works about in silico comprehensive predictions of compound-target interactions across a variety of target proteins and ligands based on descriptors of proteins and ligands. Here we propose a new protein descriptors based on the 3D grid points around proteins in order to take account of 3D shapes and/or physicochemical properties of proteins. Essentially proteins should be aligned before calculating grid point data and the conformational change of protein should be take into account. It makes difficult to use 3D grid point data for comprehensive activity prediction. In order to solve this problem, we discussed to use the principal component score as the descriptor after performing principal component analysis on a set of 3D grid point created from structures with different orientations, and utilize wider grid point spacing than used for common docking or CoMFA in order to reduce sensitivity to conformational changes. In our work, we performed two validation to confirm these.

View full abstract

We carried out electronic structure calculations on molecules having scent, and evaluated quantitative indices (electronic descriptors). Based on the calculated electronic descriptors, a correlation between electronic similarities and scents of the molecules was investigated. Through this analysis, we found that a good correlation exists between these characteristics. By using the same analysis method, we tried to predict a scent of one target molecule. It is shown that our prediction is consistent with the linguistic expression of the molecule reported in the literature.

View full abstract

Three-dimensional electronic topologies like molecular orbitals, electrostatic potential (ESP), and electronic density obtained by electronic-structure calculations are important to investigate chemical characteristics molecules. In this work, we develop a computer program to quantitatively evaluate ESP similarity of various molecules.

View full abstract

Electronic-structure calculations were performed for 33 molecules known as fatty acid synthase (FAS)inhibitors, and numerical value were evaluated to see correlation with half inhibitory concentration (IC50). Structural similarity evaluated using the fingerprint method was also carried out for the same purpose. The correlation coefficient of electronic similarity was found to be 0.79, while that of structural similarity was 0.47. Therefore, we conclude that the electronic similarity would be the better measure of pharmacological activity.

View full abstract

We are developing electronic-structure informatics to evaluate molecular similarity by using electronic structure calculation. In this presentation, we report classification of organic molecules and investigation on scaffold hopping by using the present method.

View full abstract

A database that accumulates information on molecular structures and electronic structures is expected to be useful to search for functional molecules, but it seems very difficult to establish a strategy of material development with only numerical data. To overcome this issue, knowledge information about molecules, which would be expressed with language, is considered to be important. In this presentation, we report our development of electronicstructure database of molecules which also accumulates knowledge information described with English words.

View full abstract

It has been clarified that biogenic amine receptors can be target molecules for the development of new pesticides because they have been known to control feeding behavior of insects. In this study, 17 types of parameters obtained from the electronic state calculation are taken into account as electronic descriptors of molecules. By using these descriptors, we create a regression model that explains or predicts experimental values representing biological activity. We showed that there is a good correlation between the experimental and predicted values by the regression model.: the coefficient ofdetermination is 0.737. Explanatory variables giving a major influence on the model are found to be the descriptors representing the solvation energy and molecular volume.

View full abstract