Selection of reagents for combinatorial synthesis using clique detection

The identification of a database subset in which all pairs of molecules have similarities less than some user-defined threshold is shown to be equivalent to the graph-theoretic problem of clique-detection. A comparison of several different clique-detection algorithms demonstrates that the algorithm due to Babel is by far the most efficient for this particular problem. The database subsets identified by the Babel algorithm are slightly more diverse than those identified by alternative, less time-consuming algorithms for dissimilarity-based compound selection.