[A method for predicting activity of traditional Chinese medicine based on quantitative composition-activity relationship of neural network model].

OBJECTIVE

To study a method for evaluating the quality of traditional Chinese medicine (TCM) according as their activity.

METHOD

Combined with partial least squares (PLS), BP and RBF neural networks were selected to establish the model of quantitative composition-activity relationship (QCAR) due to their strong approximation capabilities for nonlinear function respectively. The activity of TCM was predicted with the QCAR model, and the quality of TCM was evaluated according to the predicted activity.

RESULT & CONCLUSION: The proposed method was applied to evaluate the quality of Chuanxiong. The results indicated that, in the indexes including training error, prediction error and correlation coefficient, the established model is better than the model established by principal component regression or PIS regression. The new model can accurately represent the complicated nonlinear relationship between the components and the bioactivity of Chuanxiong. Consequently, this method has potential to evaluate the quality of TCM according to bioactivity.