Computational Prediction of Drug Solubility in Lipid Based Formulation Excipients

To investigate if drug solubility in pharmaceutical excipients used in lipid based formulations (LBFs) can be predicted from physicochemical properties. Solubility was measured for 30 structurally diverse drug molecules in soybean oil (SBO, long-chain triglyceride; TG(LC)), Captex355 (medium-chain triglyceride; TG(MC)), polysorbate 80 (PS80; surfactant) and PEG400 co-solvent and used as responses during PLS model development. Melting point and calculated molecular descriptors were used as variables and the PLS models were validated with test sets and permutation tests. Solvation capacity of SBO and Captex355 was equal on a mol per mol scale (R (2) = 0.98). A strong correlation was also found between PS80 and PEG400 (R (2) = 0.85), identifying the significant contribution of the ethoxylation for the solvation capacity of PS80. In silico models based on calculated descriptors were successfully developed for drug solubility in SBO (R (2) = 0.81, Q (2) = 0.76) and Captex355 (R (2) = 0.84, Q (2) = 0.80). However, solubility in PS80 and PEG400 were not possible to quantitatively predict from molecular structure. Solubility measured in one excipient can be used to predict solubility in another, herein exemplified with TG(MC) versus TG(LC), and PS80 versus PEG400. We also show, for the first time, that solubility in TG(MC) and TG(LC) can be predicted from rapidly calculated molecular descriptors.