Tetrahedron Computer Methodology最新文献

Structures of 103 ligands previously tested as inhibitors of chymotrypsin catalysis were docked into the active site of the enzyme by use of the dock computer program. The goodness of fit was evaluated according to an approximate Lennard-Jones potential scoring routine. A statistical analysis indicated that dock correctly ranked the database when viewed in terms of a contingency table of four categories: true positives, false positives, true negatives, and false negatives. Eight of the top ten scoring molecules in the computerized docking procedure had been previously reported to be effective competitive inhibitors of chymotrypsin. This agreement between the computer predictions and experimental observations was encouraging and suggests that the dock computer program may be useful in evaluating other receptors for potential binding ligands.

Without 3D information on the structure of the target biomolecule, the rational design of bioactive compounds from 3D properties is based on structure-activity relationships. The design of the series of compounds tested will determine the amount of information gained. We compared a series of chemist-designed D2 agonists with a series designed by the computer by 3D searching and automated structure transformation. We described shape by the steric energies calculated on a 2Å lattice surrounding the molecules as is done in the Comparative Molecular Field Analysis Method. To select the series of computer-designed compounds from the total of 554, we used a cluster analysis based on the 25 highest principal components of the steric fields. Compared to the chemist-designed series, this series shows a larger variation in steric properties and has less correlation between steric properties. It explores all space explored by the former series. The mean and range of the forecast D1 and D2 dopaminergic receptor-binding affinities for the two series are not different. We conclude that multivariate statistical methods based on steric potential energy fields can be used to select the best possible series of molecules for a CoMFA analysis of shape properties. These methods also reduce a large set of suggested molecules to one that is reasonable to synthesize.

We present a language for acyclic chemical groups, which has a context-free grammar, and which is close to current chemical denotations. An interpreter which accepts the language, produces a connection table, and performs valency checks has been written in PASCAL. The source code and executable program are included on disk.

This paper illustrates applicability of the COMPASS (Common Geometrical Pattern Search System) developed in our previous work for the analysis of the three-dimensional common structural features of drug molecules referred to as a pharmacophoric pattern. The three-dimensional common structural features of nootropics agents (rolziracetam, aniracetam and a tripeptide of Pro-Ala-Gly which is a vasopressin C-terminal analogue) have been examined using the system. Conformational analysis of the peptide was carried out in an exhaustive manner using the ECEPP/2 program, resulting in 232 stable conformations within 5 kcal/mol above the global energy minimum. Common geometrical pattern search between each of the conformers and rolziracetam, a rigid molecule, was performed with COMPASS. Seven different types of common geometrical patterns involving five constituent atoms were obtained with the consideration of their electronic environments. The resulting geometrical features were compared with those of an alternative COMPASS analysis between rolziracetam and aniracetam. The active conformation of the peptide and the pharmacophoric patterns of the present molecules are discussed, and the usefulness of our system for such work is also described.

The structural matching algorithm HBA (Heuristic-Backtracking Algorithm) has been applied to prediction of multi-dimensional NMR spectra. It is first shown that coherence transfer pathways in NMR spectroscopy can be represented by combinations of n-branched trees. With this representation, efficient algorithms that use HBA to predict spectra are developed. NMR data and assignments for melittin bound to fully deuterated dodecylphosphocholine micelles are used to test the efficiency of the spectral prediction programs. At the present level of development, the prediction programs can be used to aid manual interpretation of complex NMR spectra. The program is implemented in the C++ language on a SUN $\text{4}\text{280}$ computer.

This paper discusses the requirements a computer representation of organic reactions should satisfy in order to be adapted to the machine learning of generic reactions. Existing formalisms are classified and analyzed in relation with their characteristics and their use. A new representation is introduced and described: RECOUR (REpresentation COmpacte et Unifiée des Réactions = Compact and Unified Representation of Reactions).

The SEAL method has been developed which will optimize the alignment of two three-dimensional structures using their atomic partial charges and steric volumes as factors. In addition, this method will perform the superimposition many times using randomly generated starting configurations and keep only the best unique results based on the value of the alignment function. The computer generated alignments of methotrexate and dihydrofolic acid are compared with the alignments found in the active site of Dihydrofolate Reductase. The marine neurotoxins, saxitoxin and tetrodotoxin, are aligned and the results compared to other alignment techniques. The relationship of the SEAL alignments to the potential field error is explored. Program on disk.

We have designed a feed-forward neural network to classify low-resolution mass spectra of unknown compounds according to the presence or absence of 100 organic substructures. The neural network, MSnet, was trained to compute a maximum-likelihood estimate of the probability that each substructure is present. We discuss some design considerations and statistical properties of neural network classifiers, and the effect of various training regimes on generalization behavior. The MSnet classifies mass spectra more reliably than other methods reported in the literature, and has other desirable properties.

This paper describes a program for modeling complex chemical reaction mechanisms. The described program constructs the required differential equations from a symbolic description of the reactions and solves the equations numerically. The program can also fit the model to observed data by adjusting user selected rate constants. The rate constants are adjusted using simplex optimization to give the least-squares deviation between calculated and observed data.

An original semi-empirical COMPASS program for Computer-Assisted Organic Synthesis is developed. This program is based on the combination of pure combinatorial methods with empirical rules of retrosynthetic analysis. A small knowledge base containing generalized description of synthetic methods is discussed. The principles of the DIssection-Approximation (DIA) approach to formalization of organic reactions are given. A general flow-chart of the program is presented. The rule's interpreter as well as the basic principles of the rule's construction and selection criteria are discussed. The demo version of the COMPASS program for PC-compatible computers is included on disk.