In admetSAR (version2), six physicochemical or topological properties are used to defined the applicability domain, which are molecular weight, alogP, number of atoms, number of rings, H-bond acceptors, and H-bond donors. We analyzed the distribution of these properties in all the training sets of the predictive models.

Compounds with the molecular weight higher than 99% or lower than 99% of the training set will be regarded as warning, and if it is higher than the maximum of the training set, it will be tagged as out domain. Similarly, compounds with the alogP higher than 99% or lower than 99% of the training set will be regarded as warning, and if it is higher than the maximum or lower than the minimum of the training set, it will be tagged as out domain. For the other four stistics properties, since their lower bound is 0 and is reasonable, we only tag the overhigh compounds.