Biomedical engineers at Duke University have pioneered a platform that leverages automated wet lab techniques paired with artificial intelligence (AI) to design nanoparticles specifically for drug delivery. This innovative approach promises enhanced efficiency in delivering therapeutics that are often difficult to encapsulate.

Background

This breakthrough was documented in a proof of concept study where the research team successfully formulated nanoparticles that effectively delivered a challenging therapeutic for leukemia while optimizing the design of a second anti-cancer nanoparticle. The study is published in the journal ACS Nano.

AI-based tools have reshaped the drug development landscape by improving the prediction of the biological, chemical, and physical properties of potential therapeutic compounds. While significant advances have been made in early-stage drug discovery, later stages, including drug delivery optimization, remain comparatively underexplored. According to Zilu Zhang, a Ph.D. student in the lab of Daniel Reker, an assistant professor of Biomedical Engineering (BME), “Finding the right molecule is only half the battle; a new drug must be delivered accurately.”

Challenges in Nanoparticle Design

The formulation of effective nanoparticles relies heavily on the precise combination of both active therapeutic agents and excipients (inactive substances). Current AI models have limitations; they tend to focus either on the selection of promising delivery molecules or on optimizing material ratios but not both. This restricts the effectiveness of existing systems.

  • Existing AI platforms: They often can only process one type of ingredient, either the active drug or the excipients.
  • Machine learning models: These systems are too rigid, trained with fixed material ratios that do not adapt to varying combinations that could enhance delivery effectiveness.

As Daniel Reker states, “If you don’t mix them with the drug at a specific ratio, they won’t create a stable nanoparticle.” Therefore, identifying optimal mixture ratios is crucial for successful nanoparticle formation.

Introducing the TuNa-AI Platform

The research team introduced a novel platform named Tunable Nanoparticle AI (TuNa-AI). This solution uses automated liquid handling to create a comprehensive dataset comprising 1,275 distinct formulations, combining various therapeutic molecules and excipients.

“By utilizing robotics to systematically combine different ingredients, we were able to yield data that our AI model used to determine the optimal nanoparticle design under specific conditions,” said Zilu Zhang.

Results Achieved

The results of using TuNa-AI were significant:

Parameter TuNa-AI Standard Approaches
Increase in Successful Nanoparticle Formation 42.9% Baseline
Reduction in Carcinogenic Excipients 75% N/A
Encapsulation Efficiency Improved Lower

The platform successfully formulated nanoparticles that encapsulated venetoclax, a chemotherapy agent used for leukemia treatment, leading to enhanced solubility and improved effectiveness in inhibiting leukemia cell growth compared to non-encapsulated formulations.

In another successful case study using TuNa-AI, the team demonstrated reduced levels of a potentially carcinogenic excipient, while still preserving the drug's efficacy and optimizing its biodistribution in mouse models.

Future Directions

The potential of TuNa-AI extends beyond improving existing drugs; the research team aims to adapt their platform to create and optimize other types of biomaterials for a wide array of therapeutic applications. They are actively collaborating with medical researchers and physicians within and beyond Duke University to explore advancements in treating difficult diseases.

Reker expressed optimism regarding the implications of their work: “This platform serves as a foundational step towards designing and optimizing targeted delivery systems for therapeutic applications.” With TuNa-AI, researchers are excited to enhance the efficacy and safety of existing and emerging therapies.

References

  • Zilu Zhang et al, “TuNa-AI: A Hybrid Kernel Machine to Design Tunable Nanoparticles for Drug Delivery,” ACS Nano (2025). DOI: 10.1021/acsnano.5c09066

For further information on this promising research and related advancements in the field, you may refer to the original article published on Phys.org.