Glycosyltransferases are an enormous and diverse class of enzyme encompassing 1% of all sequenced genomes. They catalyze the transfer of a monosaccharide from an activated donor such as a sugar-nucleotide to an acceptor molecule. Though the primary sequences of glycosyltransferases have little homology, X-ray structural studies on glycosyltransferases have revealed that there are two main folds and that the orientation of the sugar donors with respect to the folds is highly conserved. It seems that glycosyltransferases have evolved diversified specificities toward donor sugars by changing the amino acids around the monosaccharide moiety without altering the orientation of the nucleotide moiety. In this study, we designed new glycosyltransferases with altered donor specificities by use of a novel empirical model called the Epimer Propensity Index (EPI). The EPI was constructed using 221 carbohydrate-protein complex structures in the Protein Data Bank with either galactose or glucose in the complex. The blood type B synthesizing glycosyltransferase GTB, a galactosyltransferase was our target enzyme. Two GTB mutants designed to exhibit enhanced glucosyltransferase activity were cloned, expressed and characterized experimentally. The predicted GTB mutants, Ser185Asn and Ser185Cys, exhibited 4.3- and 4.8-fold elevations in k(cat)/K(m) for UDP-Glc relative to that of wild-type enzyme.