Molecular dynamics simulation was applied to investigate the metabolism mechanism for quinone analogues. Favourable hydrogen bonds between ligand and NQO1, and parallel orientation between ligand and flavin adenine dinucleotide could explain the difference of metabolism rate (in micromol/min/mg) for quinone analogues. This is consistent with the experimental observation (Structure 2001;9:659-667). Then Support Vector Machines was used to construct quantitative structure-metabolism rate model. The model was evaluated by 14 test set compounds. Some descriptors selected by Support Vector Machine, were introduced into standard fields of three-dimensional quantitative structure-metabolism relationship to improve the statistical parameters of three-dimensional quantitative structure-metabolism relationship models. The results show that the inclusion of highest occupied molecular orbital and lowest unoccupied molecular orbital is meaningful for three-dimensional quantitative structure-metabolism relationship models. These in silico absorption, distribution, metabolism and excretion models are helpful in making quantitative prediction of their metabolic rates for new lead compounds before resorting in vitro and in vivo experimentation.