The study compares GRIND-based 3D-QSAR and CoMFA [A. Cavalli, E. Poluzzi, F. De Ponti, M. Recanatini, J. Med. Chem, 45(2002), 3844-53] to investigate a biological topic dominated by hydrophobic interactions, e.g. hERG K(+) channel blocking activity. As expected, models are found by both methods and there is a fine agreement between statistical and graphical results as well. However, a closer inspection revealed that failures in the prediction of hERG blocking activity for lipophilic compounds were registered for both methods. The study explores the reasons for these failures which are strongly dependent on the chosen method, and gives some suggestions to handle with these topics.