A new approach to finding natural chemical structure classes.

In modern drug discovery, large compound libraries need to be compared and the diversity of compound libraries needs to be analyzed. Classification algorithms are important tools for accomplishing these tasks. In this paper, a chemical structural scaffold based classification approach is reported. The goals of the approach are to find natural structure families from a large (millions of entries) compound library within a feasible time period and to view the library in two-dimensional data space using chemically meaningful methods.