Determining similarity of conformational polymorphs

Conformational polymorphs are identical molecules that crystallize in different spatial formations. Understanding the amount of difference between the polymorphs might aid drug design as there is a widespread assumption that there exists a direct connection between the conformations in the crystallized form of the molecule and the conformations in the solvent. We define a measure of similarity between conformational polymorphs and present an algorithm to compute it. For this end we weave together in a novel way our graph isomorphism method and substructure matching. We tested our algorithm on conformational polymorphs from the Cambridge Structural Database. Our experiments show that our method is very efficient in practice and has already yielded an important insight on the polymorphs stored in the data base.