Tiny molecule medications combat numerous health conditions, yet their therapeutic impact frequently falls short due to pharmacokinetic challenges — essentially how the human body processes pharmaceutical compounds. Once introduced into the system, biological mechanisms determine absorption rates, organ distribution, and the speed at which these compounds are broken down and eliminated from the body.
Microscopic delivery vehicles, typically constructed from lipid-based materials, polymer structures, or hybrid compositions, offer potential solutions to these pharmacokinetic limitations. However, these nanocarriers present manufacturing challenges and typically possess limited drug-carrying capacity.
Certain pairings of compact cancer-fighting compounds with molecular colorants have demonstrated the ability to spontaneously organize into nanoparticle structures containing substantial medication quantities. Yet, forecasting which molecular combinations will successfully form these nanostructures remains challenging given the millions of potential molecular partnerships.
Scientists at the Massachusetts Institute of Technology have engineered an innovative screening methodology that integrates artificial intelligence algorithms with rapid experimental processes to swiftly discover self-assembling nanocarriers. Their groundbreaking research, featured in Nature Nanotechnology, involved evaluating 2.1 million combinations of miniature drug compounds with traditionally "inactive" pharmaceutical components, uncovering 100 novel nanostructures with promising applications across oncology, respiratory conditions, infectious diseases, and microbial infections.
"Our prior investigations have highlighted both beneficial and detrimental impacts of excipient compounds on medication performance. Through coordinated efforts spanning multiple research laboratories and specialized facilities, we've now developed a methodology that specifically harnesses the advantageous properties of these components in nanoscale drug formulations," explains Giovanni Traverso, who holds the Karl Van Tassel (1925) Career Development Professorship in Mechanical Engineering and served as senior corresponding author for this investigation.
The research outcomes reveal a dual-purpose approach that simultaneously addresses manufacturing complexities associated with nanoparticle production and overcomes the challenge of incorporating substantial therapeutic payloads within these microscopic carriers.
"Numerous pharmaceutical compounds fail to achieve their complete therapeutic potential due to challenges with precise targeting, limited systemic availability, or accelerated metabolic breakdown," notes Daniel Reker, the study's primary author and former postdoctoral researcher in Robert Langer's laboratory. "By operating at the convergence of computational analytics, artificial intelligence, and therapeutic delivery systems, we aim to dramatically enhance our capabilities to ensure medications reach their intended destinations within the body and effectively improve patient outcomes."
Langer, who serves as the David H. Koch Institute Professor at MIT and contributes to the Koch Institute for Integrative Cancer Research, also participated as a senior author in this publication.
AI-Driven Nanoparticle Discovery Platform
To construct an artificial intelligence system capable of recognizing self-assembling nanostructures, the research team initially needed to compile a comprehensive training dataset. They carefully selected 16 self-aggregating miniature drug molecules featuring diverse chemical architectures and therapeutic applications, alongside 90 readily available compounds—including excipients currently incorporated into medications to enhance palatability, extend shelf life, or improve stability. Since both active pharmaceutical ingredients and inactive components have previously received FDA approval, the resulting nanocarriers are expected to demonstrate enhanced safety profiles and accelerated regulatory pathways.
Leveraging resources at the Swanson Biotechnology Center—an integrated collection of specialized facilities offering sophisticated technical support within the Koch Institute—the research team systematically evaluated every possible combination of drug molecule and inactive component. Following the mixing of compounds and robotic placement of 384 samples simultaneously onto nanowell plates in the High Throughput Sciences facility, scientists transported the plates—frequently containing rapidly degrading materials—to the adjacent Peterson (1957) Nanotechnology Materials Core Facility, where particle size analysis was conducted using high-throughput dynamic light scattering technology.
With 1,440 training data points established (including 94 previously identified nanostructures), the artificial intelligence system was then deployed to analyze a substantially expanded compound library. By evaluating 788 miniature drug molecules against over 2,600 inactive pharmaceutical ingredients, the platform successfully identified 38,464 potential self-assembling nanoparticles from among 2.1 million possible combinations.
The research team selected six nanostructures for additional verification, including a formulation combining sorafenib—a standard treatment for advanced liver and other malignancies—with glycyrrhizin, a compound widely utilized as both a food and pharmaceutical additive most recognized for its licorice flavor profile. Despite sorafenib's status as the primary treatment for advanced liver cancer, its therapeutic efficacy remains constrained. In laboratory cultures of human liver cancer cells, the sorafenib-glycyrrhizin nanostructure demonstrated twice the effectiveness of sorafenib alone, attributable to enhanced cellular uptake of the medication. Collaborating with the Preclinical Modeling, Imaging and Testing facility at the Koch Institute, researchers administered treatments to mouse models of liver cancer to compare outcomes between the sorafenib-glycyrrhizin nanoparticles and either compound administered independently. Results revealed that the nanostructure significantly reduced biomarkers associated with liver cancer progression relative to mice receiving sorafenib alone, while also extending survival compared to animals treated with either sorafenib or glycyrrhizin individually. Furthermore, the sorafenib-glycyrrhizin nanoparticle exhibited enhanced liver targeting compared to both oral sorafenib administration—the current clinical standard—and sorafenib injection following combination with cremophor, a commonly used solubilizing agent associated with adverse toxicological effects.
Personalized Medicine Through AI Optimization
This innovative platform holds promise for applications extending beyond merely enhancing active pharmaceutical compound performance—it offers potential for tailoring inactive ingredients to address individual patient requirements. Previous research by team members had revealed that certain excipients could trigger allergic responses in specific patient populations. Now, leveraging this expanded artificial intelligence toolkit, additional formulation options can be generated to provide suitable alternatives for these sensitive individuals.
"We now have the capability to consider personalizing drug delivery systems according to individual patient characteristics," elaborates Reker, currently serving as an assistant professor of biomedical engineering at Duke University. "Our approach can incorporate factors such as drug absorption patterns, genetic profiles, and even allergic sensitivities to minimize adverse effects during medication delivery. Regardless of specific genetic mutations or medical conditions, a pharmaceutical compound only qualifies as the appropriate treatment if it demonstrates genuine effectiveness for the individual patient."
While the fundamental tools for developing safe and effective drug delivery systems are available, integrating all necessary components has traditionally been a time-intensive process. The integration of artificial intelligence, high-throughput screening methodologies, and predictive capabilities for material interaction analysis will significantly expedite both medication design and the development of nanostructures responsible for systemic delivery. Current research efforts extend beyond improving drug delivery efficiency to exploring opportunities for developing therapeutic alternatives for patients who cannot tolerate standard formulations. By leveraging big data analytics to examine genetic backgrounds, allergic profiles, and food sensitivities, the team aims to address challenges in smaller patient populations that might otherwise be overlooked in conventional pharmaceutical development.
