PhD Thesis
Title: The SOLVES program system for search and optimization of lead structures
Available at: Tobias-lib Uni Tübingen (in german language)
Abstract:
The main goal of this work is the design and implementation of the SOLVES program system, which combines methods for data processing and evaluation (e.g. neural networks and evolutionary algorithms) with application-field specific methods under a common, integrated user interface. This distributed program system is based on the CORBA middleware standard. It allows easy, flexible and intuitive generation of workflows. The calculated results can be archieved permanently and comprehensibly in relational databases. Several examples of possible applications of SOLVES for the prediction of molecular properties and the selection of molecular descriptors are given. Additionally, a novel method for the optimization of substructural descriptors named STEEDS is introduced, which maps SMARTS patterns to tree structures. The optimization is performed by strongly-typed genetic programming. Two successful applications of the STEEDS method are described, and the obtained results are discussed.
