Abstract: In this paper quantitative structure-activity relationships(QSAR)were developed on HBV DNA polymerase inhibitors to uncover the relationship between biological activity and the key structural features. The 2D QSAR model of fingerprint fragments was built using genetic function approximation(GFA)method with good internal(adjusted R-squared: 0. 911 9)and external prediction(r²_pred = 0. 848 9). The features of 8 fingerprint fragments obtained are consistent with 3D pharmacophore. These fingerprints are notably more effective than those based on a fragment dictionary or hashing library. The combination of these fingerprint fragments to novel molecule library provides an effective and efficient approach to virtual screening and de novo drug design.