In silico prediction of the solubility advantage for amorphous drugs – Are there property-based rules for drug discovery and early pharmaceutical development?

Abstract

Oral delivery of poorly water-soluble compounds is often a substantial challenge. Once a drug candidate is selected, it is desirable to predict, based on chemical structure, which formulation technology has the highest potential to enhance drug solubility and absorption. Due to the importance of amorphous drug formulations, this work aimed at calculating the solubility ratio of amorphous and crystalline drug using in silico methods only. Molecular modeling together with multivariate methods was employed and a particular aim was to find simple structure-based rules for the technology selection of amorphous drug formulations. As a result, calculated estimates for reference compounds were generally higher than experimentally obtained amorphous solubility ratios; however, the rank order of the values revealed a significant correlation (p=0.036). Subsequently, a set of 56 neutral poorly water-soluble compounds resulted in a good partial least square model with R2 of 0.803. Most important for the amorphous solubility ratio was molecular weight, number of hydrogen bond acceptors, melting point, number of torsional bonds and polar surface area. By considering the Lipinsky rules, we proposed suitable ranges of these molecular predictors with respect to selecting promising amorphous drug formulations. Such structure-based guidance can help in early formulation development of challenging drug candidates, thereby leading to substantial cost savings. However, there is certainly more experimental research needed to finally assess how broadly the presented concepts can be applied.