A predictive 3D-QSAR model that correlates the biological activities with the chemical structures of a series of sialyltransferase inhibitors, exemplified by the sugar:nucleotide derivatives, was developed by means of comparative molecular field analysis (CoMFA). The resulting cross-validated value (q(2)=0.629), non-cross-validated value (r(2)=0.965) and standard error of estimate (SEE=0.288) indicate that the obtained pharmacophore model indeed mimics the steric and electrostatic environment where inhibitors bind to the enzyme. The developed model also possesses promising predictive ability as discerned by the testing on the external test set, and should be useful to further understand the molecular nature of inhibitor-enzyme interactions and to aid in the design of more potent sialyltransferase inhibitors.