Simulation of carbene chemistry and other problems in computer-assisted organic synthesis
Abstract
Three projects were undertaken as part of the CAMEO program, which predicts the products of organic reactions given starting materials and conditions: (1) A module was written to simulate carbene reactions. Mechanistically and thermodynamically based, it can predict the behavior of a wide range of compounds it has never encountered, and suggests likely side products. It provides insight into the underlying processes through messages issued when mechanistic decisions are made. The principal stages are: (a) Optional carbene formation, incorporating all major preparative routes. (b) Perception of carbene philicity, selectivity and spin multiplicity. (c) Analysis, wherein a few fundamental mechanisms are essayed on all applicable atoms and bonds. Each resulting potential product is rated using a model from transition state theory. To the mechanism's base rank are added entropic and enthalpic adjustments for substituent effects. This allows neat incorporation of the effects of catalyst, carbene selectivity, temperature and intramolecularity. (d) Product evaluation and generation. (2) A package was developed to compose aesthetic structure diagrams from connection tables. There are three modes, ranging from neatening to complete redesign. Difficult mono-and bicyclic rings are drawn from stored templates. Suggestions are made for a template dictionary and for drawing complex systems not in it. An irregular polygon technique improves the appearance of bridged and polyfused rings. Techniques for combating atom overlap are included. When more than one molecule is present they are distributed evenly about the screen in two steps: (a) the molecules are located where they have the best opportunity for dispersion; (b) they are dynamically scattered using a force field-like approach. A function to assess aesthetic appearance was invented; it can decide if a redrawn structure is improved. (3) Symmetry is perceived by iterative partitioning. Stereochemistry, including that of "U" double bonds, is treated. To identify duplicate products, two methods of deriving a canonical connection table are available. One uses fast symmetry-driven partitioning; the other is a rigorous stereochemically-extended implementation of the Morgan algorithm. Additionally, the last chapter discusses innovations in graphics and a new non-graphics interface.
Degree
Ph.D.
Advisors
Jorgensen, Purdue University.
Subject Area
Organic chemistry|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.