Cluster analysis of Structure-based Virtual Screening by Using Protein-ligand Interactions

Autor: Cheng-Neng Ko, 葛振寧
Rok vydání: 2006
Druh dokumentu: 學位論文 ; thesis
Popis: 94
We developed a cluster analysis method for post analysis of structure-based virtual screening. The analysis was composed of two stages based on protein-ligand interactions and compound structures, respectively. The first stage was to generate a protein-ligand interaction cluster by translating 3D structural binding information from a protein-ligand complex into a 1D real number representation, and using hierarchical clustering method to preliminarily cluster our screening results. In the second stage, we extracted molecular topology by atom-pair representation of each compound to re-grouping the clusters derived from the first stage. Each interaction cluster could be further divided into sub-clusters according to their topological similarities. The two-staged cluster analysis could be used to organize, analyze, and visualize the data of virtual screening and mining the representative candidates for future biological test. We validated this method on data sets having five classes: thymidine kinase inhibitors, dihydrofolate reductase inhibitors, estrogen receptor agonist, estrogen receptor antagonists and neuraminidase inhibitors. Our method on these pharmaceutical interest targets provided a suggestion of cluster threshold and helped to mining diversely representative structures from large number of virtual screening data. Our method also has been applied on the practical inhibitor screening analysis for Helicobacter pylori shikimate kinase (HpSK). After virtual screening in CMC database, we selected compounds from top 300 and selected 23 representative candidates. Five of 23 representative candidates were tested in vivo, and one of the five candidates, furosemide, was identified being able to inhibit HpSK by cooperated laboratory of Dr. Wen-Ching Wang.
Databáze: Networked Digital Library of Theses & Dissertations
Externí odkaz: