QUANTITATIVE STRUCTURE-ACTIVITY RELATIONSHIP STUDIES OF SULFAMATES RNHSO3NA - DISTINCTION BETWEEN SWEET, SWEET-BITTER, AND BITTER MOLECULES

Several different QSAR techniques have been applied to sweetness data for 50 sulfamates, RNHSO3-Na (21 sweet, 20 sweet-bitter, and 9 bitter) . Stepwise discriminant analysis has been used to separate the 50 mole cules into 3 classes, sweet, sweet-bitter, and bitter. Cluster analysi s using two principal components can clearly distinguish between the s weet and sweet-bitter molecules but not between all three classes. Reg ression analysis has been used to develop equations for parameters fit ting to log(RS) (RS, relative sweetness). The genetic algorithm method has been used to select parameters, and high correlations between log (RS) and a range of parameters have been achieved. Molecular field ana lysis followed by selection of relevant grid points by genetic algorit hm yielded a result in which six grid points gave a high correlation c oefficient (r(2) = 0.958, XVr(2) = 0.902).