The modern approach to drug discovery has resulted in drug candidates that tend to be poorly absorbed when taken by mouth. These compounds may benefit from clever medicine design and can be paired with other ingredients called excipients to enhance their absorption. Identifying the best excipients, however, is difficult and requires extensive trial-and-error. Many products on the market are not well-designed, leading to inefficient and variable absorption. The aim of this research was to develop new computational methods to guide excipient selection and accelerate the clinical translation of breakthrough drugs.
A database of small molecule oral drug products marketed in the European Union was constructed using code. Machine learning-inspired approaches were developed to determine whether excipient selection can be explained by the drug’s chemical structure. The final database contained 1154 oral drug products. Immediate release oral tablets (n=464) were explored as a case study.
Drugs paired with an excipient called copovidone differed from all other drugs in terms of their size, shape, and structural features. It was also found that drugs formulated with copovidone and hypromellose acetate succinate were more structurally similar to each other than to all other drugs. These results provide evidence that, in some cases, excipient selection can be guided by drug structure without the need for laboratory testing.
This research will impact our shared future by providing tools to lower the cost of development, reduce the need for wasteful experiments, and ensure faster market access for breakthrough therapies.
Recent Comments