Deriving Structures for Lead Drug Discovery from Cell-Line Screens.

摘要

The primary aim of this research was to develop a suite of computational tools for the examination of tumor screening data from the NCI's tumor screening databases. These tools were designed to easily process data from over one hundred immortalized tumor cells screened for growth inhibition by over 30,000 synthetic compounds. This analysis consisted of a self-organizing-map (SOM) clustering of compounds based on their screening responses. Our results find that clearly defined classes of compounds are clustered based on their mechanism of action. Six general groupings were identified according to the broadly defined putative classes of cellular action for these agents: nucleic acid biosynthesis, mitosis, kinase and phosphatase signaling pathways, membrane function (integrity and transport), protein metabolism, and a class of agents that exclude the previous 5 classes, and have not yet been associated with a particular cellular function. These results provide a facile means of relating previously screened compounds to the large libraries of untested compounds. This effort will increase opportunities for the discovery of novel anti-tumor agents.