The maximum common substructure as a molecular depiction in a supervised classification context: experiments in quantitative structure/biodegradability relationships.

Journal of chemical information and computer sciences

PubMedID: 12376991

Cuissart B, Touffet F, Crémilleux B, Bureau R, Rault S. The maximum common substructure as a molecular depiction in a supervised classification context: experiments in quantitative structure/biodegradability relationships. J Chem Inf Comput Sci. 2002;42(5):1043-52.
The maximum common structure between two molecules (MCS) induces a similarity that enables one to group compounds sharing the same pattern. This text relates a study based on such a structural depiction in a context of quantitative structure/biodegradability relationships (QSBR). The similarity indices are based exclusively on the MCS. First, the results of statistical tests prove that these indices significantly group compounds of similar activity together. These first conclusions enable the elaboration of classification models using those structural similarities. In a second part, a population of classifiers relying on the maximum common structure and the k-nearest-neighbor algorithm is explored. Finally, a thorough examination of the best models is conducted.