The biological activities of two sets of a total of 30 different polysulfides were investigated using QSAR. The semiempirical AM1 in Gaussian 2003 for windows was used to optimize the structures whereas a subsequent calculation of hundreds of various types of descriptors at the density functional (B3LYP/6-31G*) using CODESSA package was employed. The known 15-lipoxygenase inhibitory activity data (IC50) of 19 polysulfides out of the whole data set were correlated in a multiple linear regression procedures with the computed descriptors. Statistically, the most significant overall correlations were five- and four- parameter equations with good statistical parameters; R2= 0.9981, R2CV = 0.9970 and R2=9967, R2CV = 0.9933 respectively. The models concluded that biological activity of polysulfides is mainly attributed to quantum-chemical, geometrical and topological descriptors with neither electrostatic contribution nor chief role of sulfur atoms. Also sulfur related descriptors were not the most significant contributors in the concluded models. The obtained models were efficiently employed to estimate the biological activities of the other 11 polysulfides available in natural products such as garlic and mushroom.