Abstract: A computer implemented method is disclosed for analyzing a large data set of chemical compounds for which there is a measured activity in the same assay for each compound to determine the structural units which are responsible for the observed activity of the compounds.

Abstract: The static and electrostatic interaction energy fields between probe atoms and the atoms of a topomerically aligned fragment placed in a three-dimensional grid may be used to derive a CoMFA model. The topomeric CoMFA model coefficients may be used to predict partial activity values for fragments not derived from molecules of the activity series. The partial activities can be summed to provide a predicted activity for all fragment positions of the activity series molecules. A Virtual Library in which topomerically aligned fragments are associated with their respective steric and electrostatic interaction energies can be searched for fragments similar in shape to the fragments derived from the molecules of the activity series. The identified fragments can be used with the topomeric CoMFA coefficients to predict their activity if used in the molecular activity series.