Hit Rate Parameter and Chemical Profiling

Models Bayesian score Sensitivity TH Specificity % of unclassified 

To check the chemical diversity of the dataset, compounds tvere clustered in 843 singletons and 1836 chemical classes containing at least tv ro molecules. The biggest cluster had 72 members. The singleton set contains only 10% of promiscuous compounds. Classes tvith more than 20 compounds contain a higher number of promiscuous compounds. In fact, classes tvith 20-30 members have 25% promiscuous members and classes with 30 members have 40% promiscuous members. 

The explanation of this trend is that more compounds are submitted to the safety panel during the lead optimization when a series of compounds turns out to have off-target activities at safety targets. 

---