N Sukumar - Investigation of similarity and diversity threshold networks generated from diversity-oriented and focused chemical libraries

Version 1

      Publication Details (including relevant citation   information):


      J. Math. Chem. (2016)


      Topological properties of chemical library networks, such as the   average clustering coefficient, average path length, and   existence of hubs, can serve as indicators to describe the   inherent complexities of chemical libraries. We have used   Diversity-Oriented Synthesis (DOS) and Focussed Libraries to   investigate the appearance of scale-free properties and absence   of small-world behavior in chemical libraries. DOS aims to elicit   structural complexity in small compounds with respect to   skeleton, functional groups, appendages and stereochemistry.   Complexity here indicates incorporation of           sp3 carbons, hydrogen bond acceptors   and donors in the molecule. Biological studies have shown how   structural complexity enhances the interaction of molecules with   complex biological macromolecules. In contrast, Focussed   Libraries concentrate on specific scaffolds against a specific   biological target. We have quantified the diversity in several   DOS and Focussed Libraries based on properties of similarity and   dissimilarity threshold networks formed from them. Similarity and   dissimilarity networks were generated from diverse chemical   libraries at various Tanimoto similarity coefficients (tc) using FP2 and MACCS fingerprints.   The dissimilarity networks at very low           tc threshold led to the absence   of small-world behaviors, as evidenced by low average clustering   coefficient and high average path length in comparison to   Erdös–Renyi networks. Dissimilarity networks exhibit scale free   topology as evidenced by a power law degree distribution. The   similarity networks at high           tc threshold have shown high   clustering coefficients and low average path lengths, without the   appearance of hubs. Combining dissimilarity and similarity   threshold graphs revealed assortative and dissortative behaviors   in the DOS libraries, leading to the conclusion that the vertices   of the dissimilarity communities are more likely to share   similarity edges, but it is quite unlikely for the vertices in a   similarity community to share dissimilarity edges. We propose a   simple and convenient diversity quantification tool, QuaLDI   (Quantitative Library Diversity Index) to quantify the diversity   in DOS and Focussed libraries. We anticipate that these   topological properties can be used as descriptors to quantify the   diversity in chemical libraries before proceeding for synthesis.

      Address (URL): http://dx.doi.org/10.1007/s10910-016-0657-0