New Methods for Designing Chemical Structures Using Graph Theory and Kernel Methods

点击翻译按钮获取中文摘要

In this project, we studied a new approach for design of novel chemical compounds by combining graph theory and kernel methods. We developed branch-and-bound algorithms for enumerating tree-like chemical structures from a given feature vector consisting of frequency of labeled paths, dynamic programming-based algorithms for enumerating stereoisomers from a given tree-like or outer-planar chemical structure, and kernel functions that can discriminate stereoisomers. Some of these algorithms were implemented in a newly developed web server.