Predictive identification of co-formers in co-amorphous systems.

This work aims to understand the properties of co-formers that form co-amorphous pharmaceutical materials and to predict co-amorphous system formation. A partial least square - discriminant analysis (PLS-DA) was performed using known co-amorphous systems described by 36 variables based on the properties of the co-former and the binding energy of the system. The PLS-DA investigated the propensity to form co-amorphous material of the active pharmaceutical ingredients: mebendazole, carvedilol, indomethacin, simvastatin, carbamazepine and furosemide in combination with 20 amino acid co-formers. The variables that were found to favour the propensity to form co-amorphous systems appear to be a relatively large value for average molecular weight and the sum of the difference between hydrogen bond donors and hydrogen bond acceptors for both components, and a relatively small or negative value for excess enthalpy of mixing, excess enthalpy of hydrogen bonding and the difference in the Hansen parameter for hydrogen bonding of the coformer and the active pharmaceutical ingredient (API). To test the predictive power of this model, 29 potential co-formers were used to form either co-amorphous or crystalline two-component materials with mebendazole. Of these 29 two-component systems, the co-amorphous nature of a total of 26 materials was correctly predicted by the model, giving a predictive hit rate of 90 %.