Composition–activity Relationship(CAR) Model Of Antitumor Effect Of The Combination Of Curcuma Longa L. And Glycyrrhiza Extracts

Curcuma longa L.and Glycyrrhiza are both traditional Chinese medicines,which has been used in a long history.According to the modern researches,Curcuma has obvious antitumor effect.And on the basis of theory of Chinese medicines,Glycyrrhiza can moderate the property of herbs in order to display the better effect.In this study,by optimizing the process of modeling,fifty batches of combination of Curcuma longa L.and Glycyrrhiza extracts were studied to find out the relationship between their antitumor activity on Hela cells and their constituents to predict the anti-tumor activity according to the components.The main work is as follows:Curcuma longa L.and Glycyrrhiza from fifty batches were extracted by using ultrasonic extraction,respectively.An MTT(3-(4,5-dimethylthiazol-a-yl)-2,5-diphenyltetrazolium bromide)assay was used to acquire the antitumor activity data,which is the inhibition ratio.The results showed that the content of differdent constituents and ingibition ratio can vary a lot from batch to batch.BP neural network(BPNN)were used to establish the CAR models in which the input variable is relative area of 46 characteristic peaks and the output variable is the inhibition ratio of 50 batches of extracts of Curcuma longa L.and Glycyrrhiza against Hela cells.Root mean square error(RMSE)was used to evaluate the result.Finally,the optimum conditions of BP modeling are 5 neures of hidden layer and trainlm or trainrp as training functions.Support vector regression machine(SVR)were used to establish the CAR models in which the input variable is relative area of 46 characteristic peaks and the output variable is the inhibition ratio of 50 batches of extracts of Curcuma longa L.and Glycyrrhiza against Hela cells.Root mean square error(RMSE)and Correlation coefficient(R)were used to evaluate the result.Finally,the optimum conditions of SVR modeling are ?-SVR and LKF and RBKF as kenral functions.Genetic algorithm(GA)and particle swarm optimization(PSO)were used to optimizing the parameters of obtained models,and finally the optimal model was found—PSO-SVR-RBKF(C=2.9441;g=0.1000).