T The antioxidant and gap junctional communication(GJC)activities of carotenoids are known to be the two main anticancer mechanisms.Quantitative structure-activity relationship(QSAR)models of the two activities were developed using stepwise regression and multilayer perceptron neural network based on the calculated descriptors of quantum chemistry.The results showed that the significant molecular descriptor related to the antioxidant activity of carotenoids was the HOMO-LUMO energy gap(EHL)and the molecular descriptor related to the GJC was the lowest unoccupied molecular orbital energy(ELUMO).The two models of antioxidant activity both showed good predictive power,but the predictive power of the neural network QSAR model of antioxidant activity was better.In addition,the two GJC models have similar,moderate predictive power.The possible mechanisms of antioxidant activity and GJC of carotenoids were discussed.