Probabilistic neural network-based on QSAR for the prediction of the diuretic activity of the active constituents of traditional Chinese medicinal herbs

A quantitative structure-activity relationship(QSAR)method is used for the first time to de- velop the correlation models between the diuretic activity of the active constituents of traditional Chinese medicinal herbs and a set of three molecular descriptors.Molecular descriptors derived solely from struc- ture were used to represent molecular structures.A subset of the calculated deseriptors selected using correlation coefficient matrix and forward regression was used in the QSAR model development.Linear discriminant analysis and probabilistic neural network(PNN)were utilized to construct the linear and nonlinear QSAR model.respectively.The optimal QSAR model developed was based on a PNN with the smoothing parameter σ=0.75.Fractions correct representing the fraction of cases classified correctly of training,cross validation and test data were all 100%,respectively.It proves that this PNN is a perfect classifier network.