The Pulse
Quotient Sciences’ AI Picks Drug Formulations in Three Dosing Periods
Quotient Sciences says a proprietary AI algorithm selected modified-release tablet compositions and doses during an ongoing clinical study. The system reached a preset pharmacokinetic target within three dosing periods, with final results e
AI.info Team ·
Three dosing periods were enough to hit the preset target
Quotient Sciences says its proprietary artificial intelligence algorithm selected modified-release tablet compositions and dose levels during an ongoing clinical study, reaching the trial’s preset pharmacokinetic target within three dosing periods. The Nottingham, England-based contract research, development and manufacturing organization reported the interim findings on September 17, 2026, in a company release distributed through PR Newswire.
The result matters because modified-release formulations are designed to control how a drug enters the bloodstream, but their performance in people can be difficult to predict from laboratory testing alone. Drug developers typically cycle through formulation design, manufacturing and clinical testing before finding a composition that produces the desired exposure profile.
“Predicting how a modified-release tablet will behave in humans is difficult,” Andrew Lewis, chief scientific officer at Quotient Sciences, said in the release. “The interim data show that the algorithm learned that relationship quickly and accurately, reaching our preset target within three dosing periods.”
The model changed its next recommendation after each dosing period
Quotient Sciences says the algorithm entered the trial trained only on in vitro drug-release data. After each dosing period, the company retrained it using tablet dissolution results and pharmacokinetic data from healthy participants, then used the updated model to select the next formulation composition and dose.
The process did not give the system unrestricted control over the study. Quotient Sciences set operating limits, including a dose cap for the first prototype, while a safety committee reviewed and approved every proposed composition before manufacture and dosing. The arrangement kept human oversight in the clinical decision loop while allowing the model to use results from one dosing period to guide the next.
The study uses a generic drug with an established safety record and extensive published data. Quotient Sciences says the compound was selected to test the formulation method rather than to serve as a development candidate, and the company does not plan to advance it as a product.
Laboratory screening provided the starting point
The clinical work follows earlier laboratory screening in which the same algorithm learned the relationship between tablet composition and in vitro drug release. According to Quotient Sciences, the system mapped the formulation design space after screening one-third fewer formulations than conventional methods.
That earlier result led to the clinical hypothesis now being tested: a model that can connect composition with release behavior in the laboratory may also learn how composition affects pharmacokinetics in humans. The current study is intended to test that connection directly rather than treat laboratory performance as a sufficient substitute for clinical data.
Quotient Sciences began working with Intrepid Labs on AI-guided formulation development before the clinical study. In a December 2025 announcement, the companies said Intrepid’s machine-learning model would be incorporated into Quotient’s Translational Pharmaceutics platform, which combines formulation development, drug-product manufacturing and early clinical testing. The companies describe the system as a machine-learning model for pharmaceutical formulation science, not a general-purpose language model. Quotient Sciences’ partnership announcement provides the earlier background.
Interim data do not establish the final performance
The reported finding is preliminary. Dosing continues, and Quotient Sciences says it expects to publish complete results after the study finishes toward the end of 2026. The interim release does not disclose the drug’s identity, the number of participants, the exact compositions selected, the size of the pharmacokinetic error, or the target profile used by the study.
Those details will determine how broadly the result can be applied. Reaching a predefined target in one controlled study would show that the model can support an adaptive formulation program, but it would not by itself establish that the approach works across different molecules, release mechanisms, doses or patient populations.
“We set out to answer three questions,” Lewis said. “Can the model learn the relationship between formulation composition and performance in humans, and if so, how quickly and how accurately? On the interim evidence, it can, and quickly enough that we expect to need less clinical testing to develop modified-release formulations for other molecules.”
A digital twin links formulation choices to human data
Quotient Sciences says the AI-enhanced solution will form part of its Translational Pharmaceutics platform and support model-informed drug development. The company describes the output as a digital twin linking formulation composition with in vitro performance and human pharmacokinetics.
The practical claim is narrower than replacing clinical testing: the model is intended to help decide which formulation to test next, reducing unnecessary rounds of manufacturing and dosing. The study’s design also makes the boundary clear. The algorithm proposes a composition, but laboratory manufacturing, safety review and clinical administration remain human-controlled steps.
Until the final data are available, the strongest verified result is the one Quotient Sciences has reported: its algorithm reached the study’s preset pharmacokinetic target within three dosing periods while selecting the formulation and dose during the trial.