Fig. 2
From: Profiling and analysis of chemical compounds using pointwise mutual information
List of compound sets. Synthetic accessibility interrelation patterns are analyzed for one set of HS compounds (nonpher compound set) and three sets of ES compounds (zinc_random, savi and scubidoo compound sets). Extremely complex compounds in these data sets (_complex compound subsets) are also considered to be HS. zinc_random compound set does not contain any excessively complex compound. nonpher and zinc_random compound sets are augmented into the training set S used to train the RF classifier