The Maggiora paper1 is a very short paper, but it points out one fundamental reason of QSAR models disappoints: activity cliffs. This paper’s is so important in the activity landscape research, that it is cited by all the other four papers summarized below. The paper pointed out that because similar molecules can show significantly different biological activities, activity are often mispredicted for these molecules, even when the overall predictivity of the models are high. The author explained this by introducing the concept of activity landscape, which is related to the chemical space representation of the compounds. Then she introduced another concept activity cliffs, which is the area in the activity landscape that very similar molecules possessing very different activities. She also pointed out that outliers may reflect the presence of activity cliffs and additional compounds need to be analyzed in the neighborhood of the cliffs to ensure the landscape are adequately represented. She also discussed the lack of invariance of chemical space could be a crucial issue in QSAR models, as the distance relationships of molecules will not be preserved in different spaces. 1. Maggiora G. On Outliers and Activity Cliffs-Why QSAR Often Disappoints. J. Chem. Inf. Model. 2006 Jul 24;46(4):1535.