Duke Engineers Use AI and Robotics to Design Smarter Nanoparticles for Better Drug Delivery

Biomedical engineers at Duke University have developed an AI-powered platform that combines automated laboratory techniques with artificial intelligence to design nanoparticles for improved drug delivery. The system, called TuNa-AI, aims to overcome key challenges in formulating effective and stable nanoparticles, particularly for drugs that are difficult to encapsulate. In a proof of concept published in ACS Nano, the team used TuNa-AI to create nanoparticles capable of delivering venetoclax, a chemotherapy agent used in leukemia treatment. The AI-guided platform significantly enhanced the drug’s solubility and effectiveness in lab tests, leading to better inhibition of leukemia cell growth compared to the free drug. In a second study, the platform reduced the use of a potentially cancer-causing excipient by 75% in another chemotherapy formulation, without sacrificing efficacy and while improving how the drug spread through the body in mouse models. The innovation lies in the platform’s ability to simultaneously optimize both the choice of materials and their precise ratios—something most existing AI tools cannot do. Traditional models either focus on material selection or assume fixed ratios, limiting their ability to find optimal formulations. TuNa-AI overcomes this by using a robotic liquid handling system to generate a diverse dataset of 1,275 unique nanoparticle formulations, each varying in drug and excipient composition. By training the AI model on this rich, experimentally generated data, the team enabled the system to learn how different combinations affect nanoparticle stability, formation, and performance. The result was a 42.9% improvement in successful nanoparticle formation compared to conventional methods. Zilu Zhang, a Ph.D. student in the lab of Daniel Reker, an assistant professor of biomedical engineering, emphasized that the success of a nanoparticle depends not just on what materials are used, but on how they are mixed. “If you don’t mix them at the right ratio, even the best drug won’t form a stable particle,” he said. Reker noted that while AI has transformed early-stage drug discovery, its application in later stages—like formulation and delivery—remains underdeveloped. TuNa-AI fills that gap by enabling data-driven optimization of delivery systems. The team plans to expand the platform to include a wider range of biomaterials and therapeutic applications. They are also collaborating with researchers and clinicians at Duke and beyond to apply the technology to challenging diseases where current delivery methods fall short. “This platform marks a major step forward in designing smarter, safer, and more effective nanoparticle-based therapies,” Reker said. “We’re now poised to make existing treatments better and unlock new possibilities for future medicines.”