In order to apply the ensemble learning method to the regression analysis, Arcing_RA was proposed. This procedure decreased the average of absolute deviation of both training and test data about 20%.

View full abstract

While we have several statistics to select prediction variables in parametric regression analyses such as ordinal multiple regression, we have no suitable statistics in nonparametric regression analyses such as generalized additive model. Thus, at first, the authors took notice of AIC, Cross Validation (CV), and Generalized Cross Validation (GCV) methods. AIC is the criterion based on the maximum likelihood function and the entropy theorem of information. Although GCV is created to remove the leverage which is inevitable in Cross Validation method, it is very hard to compute. Therefore we devised some transformations of CV so as to hold the concept of GCV and get the simplicity of CV. We adapted these statistics to analyze some artificial data, and compared the ability of them for selecting the prediction variables. GCV gave a little better results than the others. On the contrary to our expectation, GCV was not always able to give stable results. It shows us the difficulty of prediction test in nonparametric regression analyses. The authors will continue to examine and improve them using other artificial data.

View full abstract

Previously, solubility parameter (SP) values were calculated by the group contribution method, the precision and the reliability of which are not so enough for the QSAR study. On the purpose of improvement of SP estimation method, correlation analysis among the observed value of SP and the other physicochemical properties of liquid compounds was studied. In the case of non-associated compounds, very distinct linear relationship between the energy of vaporization and the square of normal boiling point was found. New estimation method, which calculate SP values from those of the normal boiling point and the density of liquids utilizing this linear relationship, will be presented.

View full abstract

It was described in the book of George Gamow that Mr. Tompkins explores the atom. We attempted to express this description using CAVE. The CAVE (CAVE Automatic Virtual Environment) is a projection-based virtual reality system that surrounds the viewer with 4 screens. Isosurfaces of an atomic orbital are drawn using a software called AVS (Application Visualization System) .

View full abstract

Three-dimensional hybridized atomic orbital wave functions with one s-orbital and three p-orbitals on rectangular coordinate were visualized using isosurface technique of a software called AVS (Application Visualization System). Three-dimensional hybridized atomic orbital visualization results showed normalized orthogonality. Four-dimensional hybridized atomic orbital wave functions with one s-orbital and four p-orbitals were given by normalized orthogonality method. Four-dimensional hybridized atomic orbital wave functions were visualized using projection on three-dimensional space. We obtained the general expressions for D-dimensional hybridized atomic orbital wave functions with one s-orbital and D p-orbitals.

View full abstract

QSAR modeling is useful for us to predict some property of unknown chemical compound. But the problems are often in requiring nonlinear modeling. Since artificial neural network is very capable of dealing with nonlinear problems, it can be significant to apply the neural network to QSAR studies. We have developed a software tool for QSAR(Quantitative Structure Activity Relationships) modeling by artificial neural network, called NNQSAR(Neural Network QSAR). NNQSAR can perform learning and predicting the structure-activity data on the same user interface. And NNQSAR has features as follows: (1) two types of modeling, pattern classification and multiple regression (MR) modeling, can be treated. (2) A learning curve can be displayed in a synchronous manner. It makes us easy to visually check the state of convergence and divergence of the learning. (3) The animation of changing synapse weights in the learning process of the network is also available. Thus the changing of the weights can be checked by graphics. (4) In the MR type modeling, the learning and the prediction(evaluation) of a data set of multiplex signals can be performed. It means that the system can treat two or more activities in the learning process (or prediction process) simultaneously. The detail of the system will be discussed with a practical example.

View full abstract

Organic reaction types having six nodes are enumurated algorithmically, where each enumurated reaction type is represented by an ITS (Imaginary-Transition-Structure) conection table and characterized by the corresponding canonical name. Then the duplication of equivalent reaction types is omitted by the comparison between the generated canonical names. The resulting non-redundant set of reaction types is drawn by converting the canonical names into EPS files.

View full abstract

One of the author has proposed ITS (Imaginary Transition Structure) as a representation of organic reaction. ITS has an ability to represent starting-stage and product-stage molecules at a time so that it is suitable for search keys. On the other hand, conventional databases of organic reactions treat a producing reaction as a relation between starting-stage and product-stage. In this paper, we report ITS Producer that produces ITS from such organic reaction databases, as well as a search system that uses ITS Producer.

View full abstract

The authors have been developing a 1H-NMR spectral database management system in the internet-intranet environments. With this system, a user does not need to obtain any specific client program. He can access the database by a WWW browser with Java Plug-in. Since the system has been developed by Java and Oracle 8i with JDBC, it has little dependency for the platform. The database contains three kinds of data for various organic compounds; they are, some 4,000 full-digitized spectral data measured by 90 MHz or 400 MHz, 13,000 Structural data in CI/441 format, and 9,500 assignment data in CI format. The structural data converter is also equipped for the format conversion from CI/441 data to CS connection table and/or MDL Molfile data.

View full abstract

Structural feature analysis or similarity analysis of proteins provides us a lot of useful information for better understanding of molecular biological science and the related areas. In PROSITE database, a structural feature called motif is usually defined by a pattern of amino acid sequence using regular expressions. In our preceding works, we have constructed a 3D protein motif database that corresponded to the PROSITE. In this database, various 3D substructures that have common sequence pattern are found by visual checking. In this paper, we have investigated a classification method for such substructures. That is, a dissimilarity criterion for a pair of substructures is defined and a classification is done using a dissimilarity matrix for a group of 3D substructures. Then a substructure that has the smallest sum of dissimilarity value is defined as a representative 3D pattern. The result of 3D motif search using these representative patterns of also discussed.

View full abstract

In order to reduce frequency of structure revision of organic compounds, we have proposed a logically reasonable and systematic procedures on the basis of a concept of informational homologue. Many cases of structure revision appear to be due to not misinterpretation of structural data but simple overlooking of target structures to be determined. In a starting point of structure elucidation, all structures must be theoretically possible, and at a final stage, only the target structures should singly remain. One of unique and useful features of our procedure would be that users can arbitrarily select chemical criteria to be used to deny any structures which are not informational homologues as the needs of the case demand on the principle of a kind of axiomatism. Several cases of structure revision were discussed.

View full abstract

The intermolecular interactions between two molecular hydrogens are examined by using ab initio molecular orbital method. All calculations were performed at CCSD(T)/cc-pVTZ level of theory with Gaussian 94 program package. The calculated results indicate that these intermolecular potentials have large anisotropy. We propose their model potentials.

View full abstract

Synthetic routes from AIPHOS system are evaluated based on possibility of side-reactions using a newly developed system that recognizes reactive sites by a module for reactive site perception of a reaction prediction system SOPHIA and generates possible products including byproducts according to information on transformations in a reaction database. Some possibilities of side-reactions are predicted and evaluated for several synthetic routes proposed by AIPHOS. The results from the evaluations are visualized by a newly developed graphic user-interface. Applications to synthetic routes for ligand molecules of catalysts in Mitsui Chemicals are also demonstrated.

View full abstract

There is no appropriate information service system regarding the drug teratogenicity in Japan. We have aimed at creating an information-community for teratogenic agents, consisting of pregnant women, pharmacists, physicians, research scientists of new drugs. In this paper, we reported a trial database (DB) system for teratogenic agents. The basic structure of DB was improved in order to solve the problem in the retrieval of chemical structures from the web client. The DB performs not only retrieval of the teratogenic information and chemical structures of drugs/medicinal products, but also registration of clinical case reports and teratogenic information of various chemicals.

View full abstract

Numerous chemicals have been produced by mankind. But recently their effects on global environment are serious matter. Numerous accidents due to chemicals happened everywhere. Among chemicals available in market, quite few are ascertained on safety. Reliable data on chemical safety are few and sparse. So, it is necessary to construct a database which collect reliable data on chemical safety. Animal test to get data on chemical safety takes much cost and period. So it is impossible to get safety data on every unascertained chemicals by animal test. Here a computer prediction by QSAR is meaningful for screening. Several prediction systems by QSAR have been developed. But the performances of existing systems are all low. Considering these situations, we are developing a computer-aided evaluation system on chemical safety by structure-activity relationship named CSD. It consists of two subsystems, one is a database subsystem which contains reliable hazard data on numerous chemical substances, and another is a prediction subsystem which evaluates hazard or safety on known or unknown chemical substance by quantitative structure-activity relationship.

View full abstract

The multi-contents (compounds, proteins and pharmacological data) database has been developed. This database has been designed and constructed under the new concept of "druginformatics". The druginformatics which combine bioinformatics, proteoinfomatics, cheminformatics and some other drug related informatics is proposed in this paper. This database has been linked to a gene expression data management and analysis system for the purpose of getting drug design related information by bio-researchers easily and quickly.

View full abstract

It is useful for the chemists to know the change of molecular orbital or electron density according to the geometric transformation of molecule structures, in order to explain the mechanism of molecular reactions. The shape of molecular orbital give the important information of the bonding or anti-bonding characters. However, it is difficult to understand behaviors of electrons in molecules from molecular orbital that are represented in the forms of functions or numeric values. This paper presents visualization of orbital functions or density functions of electrons as a cloud rendered by real-time computer graphics. Electron densities are sampled and represented by voxel data in rendering. Manipulation of molecular structure, such as rotating a certain bond, is realized and the response is observed in real time. Placing transparent texture images side by side with hardware acceleration support renders transparent volume. Especially, an arbitral state of the manipulated bond angle is generated by interpolation between finite numbers of voxel data. It is also realized in real-time by a combination of rendering multiple transparent texture images. It gives chemists an advantage in understanding the features of electronic states in molecules.

View full abstract

We investigated the structures and stability of two weakly bound complexes Ar/NH3 dimer and Ar2/NH3 trimer using ab initio wavefunctions at correlated levels of theory with a large basis set. For the calculations, correlation consistent basis sets of Dunning (aug-cc-pVTZ) were used. The basis set superposition errors (BSSE) were estimated by counterpoise method. All calculations were performed with the Gaussian94 and Gaussian98 program system. For the Ar/NH3 diemer, potential energy surface exibits two minima, which have the argon atom in an equatorial location. One of the two minima is the global minimum ,which has the argon atom in the half plane that contains both the C3 axis of NH3 molecule and is between two hydrogen atoms. The other is the local minimum, which has the argon atom in the half plane that contains both the C3 axis of NH3 molecule and the hydrogen atom closest to the argon atom. The two maxima in the potential energy surface have the argon atom in the axial location. For the Ar2/NH3 trimer, our calculation suggests that the structure resembles T-shaped and Ar2 is closest to the three hydrogen atoms of NH3.

View full abstract

We present an approach to build protein backbones from alpha-carbon coordinates. The approach is analytical and based on the information of allowed regions in the Ramachandran plot. The backbone construction consists of three parts: prediction of (phi, psi) angle pairs from the alpha-carbon trace, generation of atomic coordinates with these dihedral angles, and subsequent energy minimization with molecular mechanics. Tests on several known protein structures show that the root mean square deviations in reconstructed backbones are 0.25-0.48 angstrom for coordinates and 14-34 degrees for the dihedral angles. The results indicate that our method is one of the best methods proposed in terms of accuracy. It has also been revealed that the approach is not only robust against errors in initial alpha-carbon coordinates but is also capable of providing equivalent or more reasonable models compared to other known methods. Furthermore, backbone structures has been found to be built accurately by using the (phi, psi) angles of the reference structures. This suggests that the approach would be useful in homology modeling studies.

View full abstract

Linear Free Energy Relationship (LFER) is an extremely useful tool in the exploration of reaction mechanisms. Especially, the Yukawa-Tsuno (Y-T) Eq. (1) has been applied in success to a lot of systems including benzylic cation systems. It is expected that obtained reaction constants (r and rvalues) predict characters of transition states or cationic spieces. In this point of view, the Y-T relationship was applied to the stabilities of theoretically optimized benzylic cation systems using theoretically derived substituent parameters. This examination is of critical importance in order to establish the physical meanings of reaction constants. All species have been calculated at MP2/6-31G*//RHF/6-31G* level using GAUSSIAN-98. For a theoretical s0-standard system, ring substituted a,a-dimethylbenzyl cations whose cationic p-orbitals are fixed orthogonally against benzene p-system are selected. Relative stabilities to unsubstituted derivative are determined in hydride transfer equilibria and are calculated to determine theoretical s0. The set of s+ are determined in the same manner with s0 using a,a-dimethylbenzyl cations. With these s values, Eq. (1) was applied to a lot of benzylic cation systems that have electrically and sterically various substituents at a-positions. Obtained reaction constants were examined with theoretical indices to discuss the physical meaning.

View full abstract

Theoretical MO study is presented on the electronic spectra of new polymethine dyes with a heterocyclic ring incorporated into the methine chain. 1-Alkyl-2 or 4-methylpyridinium salts and related compounds reacted at the exocyclic double bond with cyano- or methoxycarbonyl heterocycles bearing methylthio group in the presence of potassium carbonate in dimethyl sulfoxide at room temperature to give new polymethine dyes in good yields. This reaction occurred by an addition-elimination mechanism, involving nucleophilic attack of cyclic enamines at carbon deficient atoms on heterocycles followed by elimination of methylmercaptan. These polymethine dyes are red, violet, and blue in color and appeared at 450-615 nm in absorption spectra in ethanol. Some MO calculations including INDO/S, RPA, TDDFT etc. were performed to investigate their electronic spectra and fairly good agreements were obtained between experimental and calculated their absorption maxima.

View full abstract

Grand canonical Monte Carlo method was applied to simulate liquid-liquid solubility. To overcome the difficulty of insertion procedure, configurational bias Monte Carlo method was used.

View full abstract

Simulated Annealing improved by Tsallis statistics was applied to search a globally stable conformation of collagen-like oligopeptides. Recently the structures of two collagen model peptides were determined by X-ray crystallography. One of them is made of a repeating unit of L-Pro-L-Hyp-Gly. They form triple helical coils as were described in many textbooks. On the other hand, we analyzed a model peptide Boc-L-Pro-L-Pro-Gly by X-ray crystallography. Its conformation is same as the conformation of repeating unit of collagen. It suggests the helical conformation of collagen molecule is determined in the stage of single strand. The new statistical dynamics by Tsallis generalizes the ordinary thermodynamic and is going to be applied in various fields including simulated annealing. We tried the search of globally stable conformations of six model peptides, Boc-(L-Pro-L-Pro-Gly)n-methylamide, acetyl-(L-Pro-L-Hyp-Gly)n-methylamide, n = 1,2 and 3, using the improved transition and acceptance probabilities based upon Tsallis entropy. Semiempirical molecular orbital calculation package(WinMopac) was used as an energy minimizer. X-ray structures could be searched for the model peptides successfully. We conclude that the simulated annealing algorithm with semiempirical MO calculations gives globally stable conformations for collagen-like oligopeptides.

View full abstract

We have coded the programs for the fast elecrton repulsion integrals (ERIs) calculation using the hybrid algorithm and embedded them into the ABINIT-MP program developed to perform the fragment molecular orbital calculation. Furthermore, parallelization and performance evaluation of our routines were also carried out. As the results, our routines shows better perforance than original routines coded using only VRR algorithm and high parallelization efficiency. Our performance evaluation indicated that the decreases of not only floating point operations but memory accesses are needed for the fast ERI calculation.

View full abstract

A rate-constant expression is proposed for an excitation-energy transfer between two molecules that is regarded as a nonadiabatic transition between two adiabatic electronic states. It is shown that under weak intermolecular interaction the approximately derived rate constant increases with the increase in pseudo-Coulombic interaction and depends on the differences in the potential energies and the potential derivatives of the initial and final electronic states. In addition, the rate constant under strong interaction is shown to decrease with increasing pseudo-Coulombic interaction. We show that our expression has the advantage that it can be applied to the excitation-energy transfer under the (not very) weak and the strong coupling limits which cannot be addressed by Forster's theory.

View full abstract

Opiate molecules, which shows the action of analgesia, sedation and euphoria, are major interest in classical opiate pharmacology. Opioid agonists and antagonists often have similar molecular structures, though they act differently. In this study, we have theoretically investigated the structures of opiate molecules by molecular orbital calculations. 25 opiate molecules are geometry optimized by semi-empirical AM1 calculations. Calculated molecules are morphine, oxymorphone, levorphanol, nalorphine, naloxone, levallorphan, heroin, codeine, etorphine, dihydromorphine, normorphine, pentazocine, apomorphine, buprenorphine, butorpahnol, dihydrocodeine, oxycodone, thebaine, bremazocine, cyclazocine, ketocyclazocine, naloxazone, naltrexazone, naltrexone, oxymorphazone. 25 molecules have common molecular skeleton that includes ten atoms(one nitrogen atom, one oxygen atom and 8 carbon atoms). We have obtained highest occupied molecular orbital(HOMO) for 25 molecules. Electron densities of HOMO on common ten atoms are used to perform multivariate statistical analysis. Though only 64% of HOMO densities of ketocyclazocine are localized on common ten atoms, all of the other 24 molecules have more than 75% localization of HOMO densities on common ten atoms. Principal component analysis using 10x25 data group is performed. Plot of scores of the first and the second principal component classified 25 molecules into three major groups. The first one includes only ketocyclazocine. The second one is constituted from 7 molecules. The third one is constituted from 17 molecules.

View full abstract

We reported in this paper the role of inorganic acid, H2SO4,as a catalyst in Beckmann rearrangement of cyclohexanone oxime into ipsiron-caprolactam. Density functional theory calculation revealed that proton transfer between H2SO4 and cyclohexanone oxime occurred during the rearrangement reaction. H2SO4 act as both a donor and acceptor of proton for cyclohexanone oxime during the rearrangement reaction. This functional of H2SO4 would be one of cause of the acid catalytic activity.

View full abstract

In order to improve the precision of the prediction of the electronic spectra of organic dyes, novel two-center electron repulsion integral, new-gamma, for the PPP molecular orbital method has been investigated. In this paper we propose the estimation of the spectrochemical softness parameter (k-value) in new-gamma using the absolute hardness from PPP (N*M-gamma) or Hueckel method, and report the calculated wavelengths of the first absorption maxima of the anthraquinones, diarylmethans, and cyanine derivatives. The wavelengths calculated by the PPP method with the k values obtained from the absolute hardness represented the observed values better than the other methods.

View full abstract

To understand electronic properties and reactivities of proteins, we have developed a gaussian-based density functional program, ProteinDF. In fact, an all-electron wavefunction of hemoprotein cytochrome c was calculated by this program. It is necessary for all-electron calculations on proteins to find any method estimating a good initial guess. We developed the quasi-cannonical localized orbital method which creates accurate initial MOs of elongated peptide chain from MOs of small peptides. However, the calculation of the quasi-cannonical localized orbitals is proportional to the 4th power of the number of basis functions. In this study, we accelerated this calculation.

View full abstract

The B-NMR chemical shifts for cationic boron species from chlorodiphenylborane were measured. The optimized structures and B chemical shifts for the cationic boron species were calculated by using ab initio at GIAO-HF/6-311+G(2d,p)//B3LYP/6-31+G(d). The agreement between the calculated and found chemical shifts strongly suggests that the cationic species formed by the reaction of chlorodiphenylborane with silver perchlorate in CD3NO2 is not diphenylboronium ion but diphenylboronium perchlorate. On the basis of structural characters of this cationic species, cationic boron species were discussed the possibility as Lewis acids.

View full abstract

It is known that up to Li5 the most stable forms are planar and Li6 cluster has one planar isomer. However stable planar isomers for larger Li clusters have never been found. In this paper, we searched stable planar isomers for Lin (n=7 -10) clusters by using B3LYP density functional method. We didn't find planar isomers for Lin (n=7 -8) clusters but found one planar isomer for each Li9 and Li10. We discussed the characters of bonds in the planar isomers for Li9 and Li10.

View full abstract

In order to elucidate the existence of the hydrogen transfer in the 12-membered cyclic conjugated hydrogen bonding system for enolic forms of 2,2'-bridged bisdimedones 1a-c and the related compound 2, we compared molecular structures determined by X-ray crystallography and C-13 NMR chemical shifts measured in the solid state with the corresponding values obtained by DFT calculations using B3LYP functionals with the 6-31G** basis sets. The calculated bond lengths for the CC and CO bonds in the 12-memberd ring of 1a are in accord with the experimental values. The calculated single(double) bonds in the 12-memberd rings for 1b, 1c, and 2 are longer(shorter) than those determined at RT. The calculated chemical-shift differences for the C1 and C3 carbons are 7 ppm larger than, 2 ppm larger than, and nearly equal to the experimental values at 206K for 1a, 1b, and 1c, respectively. These results are closely related to the differences in the calculated barriers to the double hydrogen transfer in the 12-membered conjugated hydrogen bonds, supporting the existence of the hydrogen transfer.

View full abstract

Gangliosides are the sphingolipids containing sialic acids. Some enzymes recognize their sugar chain. It has been reported that gangliosides inhibit the activity of an enzyme NAD+ glycohydrolase(CD38), and that those with a tandem sialic acid in the sugar chain have the largest inhibitory effect. However, the mechanism at higher levels of structure has not been elucidated. The objective of our investigation is to clarify the cause of difference in the inhibitory effect in terms of three-dimensional structures and electronic structures of gangliosides. We assessed the similarity of the calculated structures of NAD+ and tandem sialic acid using CONFLEX, MM3 and PM3, on the assumption that CD38, an NAD+ reacting enzyme, cross-reacts with the tandem sialic acid in gangliosides. It was found that there were similarities in the distance between oxygen atoms having negative charges, the dipole moment, and HOMO between the POO--O-POO- in NAD+ and the two COO- groups in tandem sialic acid. A strong correlation between the orbital energy of HOMO and the magnitude of the inhibitory effect was also found. Our study suggests that CD38 recognizes the POO- groups in NAD+ and the two COO- groups in tandem sialic acid, which is consistent with the experimental results.

View full abstract

Dioxins have very strong toxicity to human health and so they are considered to be one of the most important environmental problems. Many scientists are trying to find the way of degrading dioxins, but the specific way other than the thermal decomposition at high temperature have not been found. We call polychlorodibenzo-p-dioxin (PCDD) "dioxins". Both polychlorodibenzofuran (PCDF) and Co-PCB are analogues of dioxins. PCDD have 75 congeners and they have different toxicity by the numbers and the positions of chlorines. These structures include that two benzene rings are bridged by oxygen. And these bonds are similar to the C-O-C bond of ether. The major process to break the bond of ethers is cleavage by acids. So we examine the cleavage of dioxin by hydroiodic acid compared with that of ether. And we also examine compared with the difference between the gas phase reaction and the liquid phase reaction by the use of the B3LYP hybrid density functional theory.

View full abstract

The structure of cyclic hexapeptides, hexaglycine (c-Gly6) and that with two leucine residures (c-(LeuGlyGly)2), were estimated with molecular mechanics calculation (CAChe Mechanics, Fujitsu Co.) and CONFLEX (Conflex Co.). The most stable structures of c-Gly6, a pair of mirror images with each other, show a twisted oval feature with four pairs of intramolecular hydrogen bonds. Calculation for the c-(LeuGlyGly)2 was done with six initial structures, which were formulated by adding a pair of leucine branches to the c-Gly6 structures (three cases for each Gly6 structure). As a result, five different types of stable structures were afforded, all of which basically retained the original hexapeptide skeleton. The difference in total energies among the stable structures was less than 0.9 kcal/mol. One molecule of isobutylamine was added to one of the stable structures and calculation was made to afford stable structures for the 1:1 complex between c-(LeuGlyGly)2 and the amine. The resulting structure have the ring structure similar to the original hexaglycine ring. The stabilization energy for the 1:1 complex was calculated as -10.2 kcal/mol, with the van der Waals term as the major contribution.

View full abstract

Folding and unfolding molecular dynamics simulations of a b-hairpin peptide [Ace1-Ile2-Thr3-Val4-Asn5-Gly6-Lys7-Thr8-Tyr9-Nme10] in explicit water were carried out. The obtained nuclear Overhauser effect (NOE) intensity was in good agreement with the intensity using NMR spectroscopy. The energy landscape was separated into two regions; deep energy wells and rugged surface. The deep wells consisted of b-hairpin conformations, and the rugged surface had random-coils. Both the random-coil and the b-hairpin were stable at 300 K, while only the random-coil was stable at > 450 K. Therefore, folding mechanism of b-hairpin was funnel-like. The unfolding of b-hairpin conformations at 400 - 500 K occurred along the minimum energy region of the energy landscape at the corresponding temperature.

View full abstract