Chemical diversity and biological activity

The application of diversity selection for identification of biologically active molecules is based on the similarity property principle according to which no compound similar to a compound already in the screening needs to be screened since a similar outcome can be expected. However, experimental errors of the screening experiment and the only limited validity of the similarity property principle can lead to the situation that a diversely selected subset does not cover more hit series of a screening summary data set than a random selection. Due to the size bias in chemical similarity functions diversely selected subset can have a reduced complexity, which causes unwanted side effects on the distribution of IC50 values.