A multivariate chemical similarity approach to search for drugs of potential environmental concern.

A structural similarity tool was developed and aimed to search for environmentally persistent drugs. The basis for the tool was a selection of so-called anchor molecules and a multidimensional chemical map of drugs. The map was constructed using principal component analysis covering 899 drugs described by 67 diverse calculated chemical descriptors. The anchor molecules (diclofenac, trimethoprim, and carbamazepine) were selected to represent drugs of known environmental concern. In addition 12 chemicals listed by the Stockholm Convention on persistent organic pollutants were used representing typical environmental pollutants. Chemical similarity was quantified by measuring relative Euclidean distances in the five-dimensional chemical map, and more than 100 nearest neighbors (kNNs) were found within a relative distance of less than 10% from each drug anchor. The developed chemical similarity approach not only identified persistent or semipersistent drugs but also a large number of potentially persistent drugs lacking environmental fate data.