A 3D-QSAR CoMFA study of piperidine-based analogues of cocaine with flexible 3 alpha-substituents is described. A series of pharmacophore models were generated based on three representative compounds 1p, 2i, and 3c using the Genetic Algorithm Similarity Program (GASP) method. The flexible superposition of all studied compounds was performed for each pharmacophore model using the FlexS algorithm and the three-dimensional structure of 2i as a template. All sets of the overlaid structures with the top-ranked conformers were used for CoMFA modeling. Two best initial CoMFA models were selected and further optimized by identifying the best-fitting conformer of each compound. Compared with the initial models, the conventional correlation coefficients r(2) for the optimized models 1 and 2 were improved from 0.90 and 0.837 to 0.997 and 0.993, respectively. The leave-one-out cross-validated coefficients q(2) for the optimized models 1 and 2 were improved from 0.515 and 0.296 to 0.828 and 0.849, respectively. The results of the two CoMFA models suggest that both steric and electrostatic interactions play important roles in the binding of the 3 alpha-substituents of the piperidine-based analogues of cocaine. The contributions from steric and electrostatic fields for model 1 were 0.621 and 0.379, respectively. The contributions from steric and electrostatic fields for model 2 were 0.493 and 0.507, respectively. The two highly predictive CoMFA models indicate that the 3 alpha-substituent has two possible binding modes at the DAT. The CoMFA contour maps provide a visual representation of prospective binding modes of the 3 alpha-substituent of the piperidine-based analogues of cocaine and can be used to design novel DAT inhibitors that may be useful for the treatment of cocaine abuse and certain neurological disorders.