An artificial intelligence algorithm has selected modified-release tablet formulations and dose levels during an ongoing clinical study, according to interim findings reported by Quotient Sciences. The contract research, development and manufacturing organization said its proprietary algorithm achieved a predefined pharmacokinetic target within three dosing periods, meeting the study’s interim objectives.

The approach is intended to address a longstanding challenge in modified-release drug development: establishing how changes in formulation composition translate into drug exposure in humans. Conventional development can require several cycles of laboratory formulation work followed by clinical testing, potentially extending the development timeline.

“Predicting how a modified-release tablet will behave in humans is difficult,” says Andrew Lewis, PhD, Chief Scientific Officer at Quotient Sciences. “The interim data show that the algorithm learned that relationship quickly and accurately, reaching our preset target within three dosing periods.”

The clinical study builds on earlier laboratory work conducted by Quotient Sciences using the same algorithm. During that work, the model was used to learn the relationship between tablet composition and in vitro drug release. According to the company, it was able to map the formulation design space after screening approximately one-third fewer formulations than would have been required using conventional approaches. That result provided the basis for testing whether a similar model could learn how formulation changes affect pharmacokinetic performance in humans.

For the clinical study, the algorithm began with training based solely on in vitro release data. Following each dosing period, new dissolution and pharmacokinetic data from healthy participants were incorporated into the model. The updated algorithm then selected the formulation composition and dose for the subsequent period. The process incorporated human oversight. Quotient Sciences established operating constraints for the algorithm, including a maximum dose for the initial prototype, while a safety committee reviewed and approved each formulation before it was manufactured and administered.

The trial is being conducted using a generic drug with an established safety profile and substantial published information. Quotient Sciences said the compound was selected specifically as a test case for the algorithm and is not intended to become a commercial development program. The company is using the study to investigate whether an AI-driven approach can learn the relationship between formulation composition and human pharmacokinetics quickly enough to reduce the number of clinical iterations normally required during modified-release formulation development.

“We set out to answer three questions,” Lewis says. “Can the model learn the relationship between formulation composition and performance in humans, and if so, how quickly and how accurately?”

Based on the interim findings, Quotient Sciences said the answer to those questions is encouraging, although dosing in the study has not yet finished. The work forms part of Quotient Sciences’ Translational Pharmaceutics platform, which brings together drug product development, manufacturing and clinical testing.

The company said its AI-enabled formulation approach is designed to support model-informed drug development by creating a digital representation linking formulation composition with both in vitro drug-release behavior and human pharmacokinetics.

The remaining dosing periods are expected to continue through the end of 2026, with full study data planned for release after the trial concludes. If confirmed in the final dataset, the findings could provide further evidence for using adaptive computational approaches to reduce formulation-development iterations. For now, however, the clinical results remain interim and are based on a single study using a generic compound.

For more information, visit www.quotientsciences.com.