Metabolisms represent highly organized systems characterized by strong regulations obeying to mass conservation law. This makes a whole chemical resource to be competitively shared between several biosynthesis components (ways) at both intra-and inter-molecular scales. Statistically, the whole shared-resource principle can be considered under a constant sum-unit constraint which represents the basis of simplex mixture rule. In this work, a new simplex-based chemometric approach was developed to extract scaffold information on different biosynthesis regulation factors responsible for the chemical structural diversity at atomic and molecular scales within a large metabolic system. This approach consisted in combining different (p) molecular clusters into a complete set of N mixtures containing a same constant (n) number of molecules which were randomly sampled by gradually varying the relative weights (n_j/n) of combined clusters j. The complete set of N combinations was initially given by the Scheffé’s mixture matrix. At the output of Scheffé’s design, each molecular mixture was represented by a theoretical average (barycentric) molecule which was trained by the characteristics of the different weighted clusters. The mixture design was iterated several (K) times by applying bootstrap technique leading to extensively explore chemical variability between and within clusters. Finally, the K response matrices issued from the K applications of Scheffé’s mixture design were averaged to obtain a smoothed matrix containing scaffold information on different regulation factors responsible for molecular diversification at inter- and intra- (atomic) molecular scales. This matrix was used as a backbone for graphical analysis of different positive and negative trends between atomic characteristics (chemical substitutions) at both mentioned scales. This new simplex approach was illustrated by cycloartane-based saponins of Astragalus genus by considering three desmosylation clusters (mono-, bi- and tridesmoside saponins) characterized by relative glycosylation levels of different aglycones’ carbons.