Automated carbon-13 nuclear magnetic resonance spectrum simulation based on database direct retrieval and model building techniques.

摘要

Carbon-13 nuclear magnetic resonance (;The wealth of information made available by the ;Chapter 3 describes an algorithm which is used to encode the chemical environment of each of the carbons in a large structural database in which the chemical shifts of the carbons have been experimentally determined and assigned. The prediction of chemical shifts of carbons can be performed by a direct retrieval of the carbon in the database whose chemical environment is most similar to the chemical environment of the carbon whose shift is to be predicted.;Chapter 4 describes a speed enhancement of the vector-based search system used to find the most similar chemical environment vectors in the database of vectors. This enhancement utilizes a partitioning of the database to allow a greatly reduced number of vector comparisons while still finding the most similar vectors.;Chapter 5 describes the calculation of regression models based on a subset of carbon atoms which are retrieved from the database using the environmental encoding described in Chapter 3. The independent variables used in the regression models are structural descriptors of the topological, geometrical, and electronic environment of the carbon atom.;The use of geometrical structural descriptors requires the calculation of three-dimensional atomic coordinates for all the structures in the database. Chapter 2 describes the parallelization of the molecular mechanics software used to generate these atomic coordinates.;Each of the subsections of the simulation package is studied in detail, and the integration of these parts to form an overall software package is described.