METHOD FOR DETERMINING A SHAPE SPACE FOR A SET OF MOLECULES USING MINIMAL METRIC DISTANCES

摘要

I describe several techniques for characterizing molecules based on theshapes of their fields. The minimal distance between two molecular fields isused as a shape-based metric, independent of the underlying chemicalstructure,and a high-dimensional shape space description of the molecules is generated.I then show how these attributes can be used in creating, characterizing, andsearching databases of molecules based on field similarity. In particular,theyallow searches of a database in sublinear time. Next, I extend the utility ofthis approach by describing a way to automatically break molecules into aseries of fragments by using an ellipsoidal Gaussian decomposition. Not onlycan these fragments then be analyzed by the shape metric technique describedabove, but the parameters of the decomposition themselves can also be used tofurther organize and search databases. The ellipsoidal method can also be usedto describe binding or active sites on macromolecules, providing a templatefor searching for complementary molecules in a database such as I describe.The most immediate application of these techniques is to pharmaceutical drugdiscovery and design.