Abstract: AIM　To predict skin permeability of drugs with theoretical parameters. METHODS　The semiempirical self-consistent field molecular calculation AM1 method is utilized to obtain the structural parameters of drug molecules. Stepwise multiple regression analysis or BP neural network is then utilized to establish the correlation between skin permeability of drugs and their structural parameters. RESULTS　The calculated human skin permeability coefficients (k_p) of 22 model drugs in vitro or the R values (R=absorbed/unabsorbed) of 17 drugs in vivo are in good agreement with their observed values. CONCLUSION　Theoretical parameters can be used to predict skin permeability of drugs.