ABSTRACT
A significant hurdle for discovery of plant-derived products is the numerous trial-and-error experiments required to develop an effective purification strategy. To overcome the experimental burden, a quantum mechanics-based molecular modeling approach - known as the COnductor-like Screening Model for Real Solvents (COSMO-RS) - was used to predict a suitable two-phase solvent system to purify six silymarins from an aqueous mixture. Silymarins, a class of flavonolignans found in milk thistle (Silybum marianum L.), are well suited for assessing the use of a molecular modeling approach to predict partitioning in a countercurrent chromatography (CCC) separation because they are well characterized and previous studies report low purity fractionation in liquid-liquid solvent systems. They also present an opportunity to evaluate the use of COSMO-RS in predicting the partitioning of structurally similar isomeric compounds that are present together in an aqueous solution upon extraction from their native source. The COSMO-RS model results predicted the partition coefficients in: three traditional ARIZONA solvent systems (composed of heptane, ethyl acetate, methanol, and water), nine additional variations of this quaternary solvent system, and two chloroform, methanol, and water solvent systems. Predicted results were concise but not accurate when compared to experimental results determined by the shake flask method. The 1:4:3:5 n-heptane:ethyl acetate:methanol:water (v/v/v/v) system was identified to be an improvement on the 1:4:3:4 system previously reported. The present study verified the ability of COSMO-RS to hone in on one or two solvent systems that will yield the best fractionation using CCC.