MEANS AND METHOD FOR RECURSIVE PARTITIONING ANALYSIS OF LARGE STRUCTURE-ACTIVITY DATA SETS

The identification of three-dimensional pharmacophores from large, heterogeneous data sets is achieved by combining a conformational and correspondence search procedure. More specifically, each flexible ring in a compound is isolated by cutting off one or more side chains while keeping the side chain neighbors nearest the flexible ring. Every flexible corner of the isolated ring is perturbed relative to an average ring plane. The chains that were cut off are reconnected and torsional ranges for each rotatable bond is calculated. Responsive to the identification of points within the calculated torsial ranges, the identified conformations are uniformly sampled and stored. The conformation generation process is integrated into a correspondence search procedure to maximize both the efficiency and speed.