Two extensive studies quantifying the ability of topomer shape similarity to forecast a variety of biological similarities are described. In a prospective trial of "lead hopping", using topomer similarity for virtual screening and queries from the patent literature, biological assays of 308 selected compounds (representing 0.03% of those available, per assay type) yielded 11 successful "lead hops" in the 13 assays attempted. The hit rate averaged over all assays was 39% ("activity"defined as inhibition > or =20% at 10 microM), significantly greater than an unexpectedly high negative control hit rate of 15%. The average "Tanimoto 2D fingerprint similarity" between query and "lead hop" structures (0.36) was little more than the Tanimoto similarity between random drug-like structures. Topomer shape and Tanimoto 2D fingerprint similarities were also compared retrospectively, in their tendencies to concentrate together potential and actual drugs reported to belong to the same "activity class", for twenty classes. Among the most similar 3% of structures (corresponding to "> or =0.85 Tanimoto" for these structures), an average of 62% of the topomer similar selection possessed a near neighbor belonging to the same activity class, roughly a one-third superiority over the "Tanimoto > or = 0.85" selection containing 48% actives in avoiding false positives. Conversely, the least similar 75% of structures contained 0.3% actives for topomer similarity vs 1.0% actives for Tanimoto 2D fingerprint similarity, a 3-fold superiority for topomers in avoiding false negatives.

FlexX and DrugScore are the two computer programs which were applied finally to search for inhibitors of the enzyme carbonic anhydrase II in compound libraries and to predict their binding properties. Subsequently the actual binding affinities of 13 of the best‐ranked compounds discovered in this way were measured experimentally. Three of these proved to be subnanomolar, one nanomolar, and seven micromolar inhibitors. The predicted binding mode was confirmed crystallographically for two of the inhibitors.

No abstract

Based primarily on further studies of a collection of eleven publications reporting fifteen successful 3D-QSAR relations, several phenomena are preliminarily described. The RMS error of 133 ligand binding energy predictions based on these successful 3D-QSARs is 0.75 kcal/mole, which compares favorably to the prediction accuracies of approaches that include the receptor. A similar result is obtained when topomer alignments are substituted for those published, with seemingly profound implications for the future of 3D-QSAR. The "alignment-averaged" molecular properties, log P and molar refractivity, have very little correlative power for these data sets, either alone or in combination with the 3D-QSAR field descriptors. The q (2 )metric for the number of PLS components necessarily tends to discard any unique or unconfirmed SAR information. Large drops in q (2) are thus to be expected whenever such unique information is first encountered. Predictive r (2) values from an exploratory new "series trajectory" analysis of these 3D-QSAR though highly variable do not differ much from their q (2) values, a phenomenon that seems to encourage prediction even when there are so few structures underlying a 3D-QSAR so that almost all information is unique.

Template CoMFA, a novel alignment methodology for training or test set structures in 3D-QSAR, is introduced. Its two most significant advantages are its complete automation and its ability to derive a single combined model from multiple structural series affecting a biological target. Its only two inputs are one or more "template" structures having 3D coordinates that share some Cartesian space, as may result from X-ray crystallography or pharmacophoric hypothesis, and one or more connectivity-only SAR tables associated with a common target. Template CoMFA also overcomes the major disadvantages of both existing 3D-QSAR alignment methodologies, specifically the tedium and subjectivity of familiar ad hoc approaches, and the awkwardness, occasional physicochemical heresies, and structural scope limitations of the purely topomer approach. The template CoMFA algorithms are described, and two of its application classes are presented. The first class, general models of binding to factor Xa and P38 map kinase, uses crystallographic structures as templates, with the encouraging result that the statistical qualities of each of these two combined models are equivalent to those of their constituent individual series models. The second, 15 data sets originally collected for validation of topomer CoMFA, with arbitrary structures as templates, confirms that the modeling power of template CoMFA resembles that of its predecessors.