We derived new parameters for the semi-empirical PM3 method, PM3(MS), designed for studies of saccharides. The parameters were optimized using genetic algorithm procedure. We used the MP2/cc-pVTZ energies and optimized geometries of monosaccharides as the reference set. Results show that compared to the standard PM3, the new PM3(MS) parameters predict better ordering of the conformers energy both for the members of the training set, as well as those outside the training set. (C) 2010 Elsevier B.V. All rights reserved.