A major study published in Advanced Materials reveals that machine learning can be used to design self-assembling nanoparticles that enhance drug delivery in fibrotic tissues, potentially revolutionizing treatment for conditions like heart disease and pancreatic cancer. The research, led by a team of scientists, demonstrates how a single molecule can act as both a drug delivery enhancer and a therapeutic agent, offering a new approach to overcoming the challenges of low solubility and poor tissue penetration.

Overcoming the Challenges of Drug Delivery

Many small-molecule drugs used in modern medicine face challenges related to low solubility and rapid clearance from the body. Traditional nanocarriers, while effective, often require complex manufacturing processes that limit their scalability. In contrast, small-molecule nano-self-assembly offers a simpler and potentially more scalable approach to drug delivery.

The study focuses on fibrotic diseases, where tissue stiffening and increased stromal density reduce drug penetration. Fibrosis is a common complication in conditions such as liver cirrhosis, pulmonary fibrosis, and heart disease. Researchers turned to FAP, a membrane-bound serine protease enriched in activated fibroblasts within fibrotic lesions, as a potential target for both drug delivery and therapeutic action.

“FAP is not only a target for antifibrotic therapy but also a potential ‘handle’ for drug delivery,” said the lead researcher. “By repurposing an existing FAP inhibitor as a co-assembly excipient, we were able to create stable nanoparticles with a range of hydrophobic drugs.”

Machine Learning and Molecular Design

The researchers used a combination of molecular dynamics simulations and machine learning to identify the key features that determine whether a drug can co-assemble with SP-13786 (SP), a small-molecule FAP inhibitor. Starting with 4,810 computed molecular descriptors, they narrowed down to 356 interpretable physicochemical features and ultimately identified 228 descriptors that predict co-assembly outcomes.

Key predictors included aromaticity, molecular rigidity, and nitrogen-related interaction features. The machine learning models translated these insights into design cues, allowing researchers to predict which hydrophobic drugs are likely to co-assemble with SP. This approach could significantly reduce the trial-and-error process in drug formulation.

“The machine learning models helped us identify patterns that would have been difficult to discern through traditional methods,” said a co-author of the study. “This data-driven approach provides a roadmap for developing new drug delivery systems with high efficiency.”

Testing in Fibrotic Tissues

The researchers tested their nanoparticle system, called SCAN, in both in vitro and in vivo models. In cell studies, they observed that SCAN interacted with FAP-expressing fibroblasts, leading to morphological changes and uptake dynamics. In vivo experiments focused on a murine model of myocardial ischemia/reperfusion (IR) injury, a condition that leads to progressive fibrosis.

Using PET/CT with a radiolabeled FAP tracer (68Ga-FAPI-04), the team validated in vivo FAP targeting, while fluorescence imaging helped track the biodistribution of SCAN. The results showed enhanced accumulation of SCAN in the injured, fibrotic myocardium compared to free-drug controls.

“SCAN showed better retention in fibrotic tissue, even as the tissue stiffened and became less permeable,” said a researcher involved in the study. “This suggests that SP’s inhibition of FAP may help overcome the physical barriers posed by fibrosis.”

The study also evaluated SCAN in a stromal-rich pancreatic cancer (PDAC) setting, indicating that the platform could be useful in a variety of conditions where dense stroma and fibroblast-driven biology limit drug access.

As fibrosis progresses, the effectiveness of free-drug delivery declines sharply, while SCAN showed a more gradual decline. This effect is attributed to the ongoing inhibition of FAP by SP, which may ease the physical and biochemical barriers associated with fibrosis.

The findings suggest that the combination of a molecule chosen for its biological activity (FAP inhibition) and its material properties (co-assembly) can produce a high-loading, relatively simple nanoparticle system. This approach could be especially valuable for fibrotic and stromal-rich diseases, where both biochemical and physical barriers hinder drug delivery.

The study’s authors are now working to translate these findings into clinical applications. They plan to conduct further preclinical trials to evaluate the safety and efficacy of SCAN in larger animal models before moving toward human trials.