Weighted Superstructures for Chemical Similarity Searching

Chemical similarity searching forms an important part of the virtual screening process. In this study, we present a graph-based matching method that assembles the target query graph from a number of active molecules. We apply both a weighted and an unweighted maximum common subgraph algorithm to measure the similarities between all molecules in three different data sets and samples of generated query supergraphs. The hit lists were compared against a simple MAX data fusion and they showed some improvement in the enrichment in some cases while overall not producing any worse results.