Small-world phenomena in chemical library networks: application to fragment-based drug discovery

A wide variety of networks in various fields have been characterized as small-world networks. In scale-free networks, a representative class of small-world networks, numbers of contacts (degree distributions) of nodes follow power laws. Although several examples of power-law distributions have been found in the field of chemoinformatics, the network structures of chemical libraries have not been analyzed. Here, we show that small-world phenomena are observed not only in existing chemical libraries but also in virtual libraries generated from structurally diverse fragments when represented as networks. On the basis of this observation, we propose that an efficient compound-prioritization method of fragment-based drug discovery (FBDD) would be to select those fragments as a starting point such that the linked compounds become hubs in the library and therefore allow identification of many similar compounds when all-to-all fragment linkings are performed. Moreover, our analyses indicated that the variety of linkers had a marked influence on the network structure and thus on the diversity of the compounds synthesized by linking fragment hits.