In the present work, QSAR models for predicting the activities of ursolic acid analogs against human lung (A-549) and CNS (SF-295) cancer cell lines were developed by a forward stepwise multiple linear regression method using a leave-one-out approach. The regression coefficient (r(2)) and the cross-validation regression coefficient (rCV(2)) of the QSAR model for cytotoxic activity against the human lung cancer cell line (A-549) were 0.85 and 0.80, respectively. The QSAR study indicated that the LUMO energy, ring count, and solvent-accessible surface area were strongly correlated with anticancer activity. Similarly, the QSAR model for cytotoxic activity against the human CNS cancer cell line (SF-295) also showed a high correlation (r(2) = 0.99 and rCV(2) = 0.96), and indicated that dipole vector and solvent-accessible surface area were strongly correlated with activity. Ursolic acid analogs that were predicted to be active against these cancer cell lines by the QSAR models were semisynthesized and characterized on the basis of their (1)H and (13)C NMR spectroscopic data, and were then tested in vitro against the human lung (A-549) and CNS (SF-295) cancer cell lines. The experimental results obtained agreed well with the predicted values.