Generation of Matched Molecular Pairs

In order to identify MMPs and contexts in an unsupervised manner (i.e., without specifying the transforms beforehand), a recent approach has been foDowed that is based onfragment indexing, andhas been recently developed in GlaxoSmithKline[21]. A very similar algorithm was also described a few years earlier by Wagener and Lommerse [11]. 

2) RDKit open-source cheminformatics, version 2010v4. Available from http //www.rdkitorg. 

For the needs of the present study, the output of the fragment indexing algorithm was processed and filtered appropriately, according to the following successive steps  

2) For each unique molecular pair, only the instance with the largest common substructure (or equally the smallest transformation) was kept while the remaining intermediate common substructures were discarded. 

---