The use of molecular descriptors in the development of co-amorphous formulations

Co-amorphous systems consisting of a drug and an amino acid have been investigated extensively for the enhancement of drug solubility and amorphous stability. The purpose of this study is to investigate which molecular descriptors are important for predicting the likelihood of a successful co-amorphisation between amino acid and drug. The predictions are thought to be used in an early screening phase to identify potential drugamino acid combinations for further studies. A large variety of molecular descriptors was calculated for six drugs (carvedilol, mebendazole, carbamazepine, furosemide, indomethacin and simvastatin) and the twenty naturally occurring amino acids. The descriptor differences for all drug-amino acid combinations were calculated and used as input in the X-matrix of a Partial Least Square Discriminant Analysis (PLS-DA). The Y-matrix of the PLS-DA consisted of the X-ray powder diffraction response (co-amorphous or not co-amorphous) obtained by ball milling all combinations for 60 min. The PLS-DA model showed a clear separation of the not co-amorphous and the co-amorphous samples and was successfully predicting the class membership of 19 out of the 20 completely left out drug-amino acid combinations of mebendazole. The approach seems to be promising for predicting the ability of new drug-amino acids combinations to become co-amorphous.