The application of three-dimensional H-suppressed BCUT metrics (BCUTs) in binary QSAR analysis wasinvestigated using carbonic anhydrase II inhibitors and estrogen receptor ligands as test cases. Variableselection was accomplished with a genetic algorithm (GA). Highly predictive binary QSAR models wereobtained for both sets of compounds within 200 GA generations. The derived binary QSAR models werevalidated with two sets of compounds not included in the training sets. The results indicate that BCUTs arevery useful molecular descriptors, and the genetic algorithm is a very efficient variable selection tool inbinary QSAR analysis.