How a virtual rodent is quietly reshaping how we test nanomedicine.
Before a new drug can reach a patient, it has to survive a lengthy and often sobering gauntlet: laboratory testing, animal studies, clinical trials, regulatory review. Somewhere in the early stages of that process — specifically, the part where researchers try to figure out where a nanoparticle goes once it’s inside a body — a lot of mice have historically been involved.
That may be changing.
A researcher at Empa, the Swiss Federal Laboratories for Materials Science and Technology, has developed what is essentially a virtual mouse. Not a simulation in the video game sense, but a computational model sophisticated enough to predict how different nanoparticles distribute through a mouse’s body — liver, kidneys, lungs, spleen — based purely on their physical properties: size, surface coating, electrical charge.
The model is called a physiologically based pharmacokinetic model, or PBPK model. The name is a mouthful, but the concept is fairly intuitive: it maps how a substance moves through biological compartments over time, the same way you might track a package through a courier network. What makes this particular version notable is how it was trained. Jimeng Wu, the doctoral student behind the work, fed it data from 18 peer-reviewed mouse studies, then layered in a machine learning method — multivariate linear regression — so the model can adjust its own parameters to match whichever nanoparticle is being tested. That last part matters. Traditional PBPK models are calibrated for one specific material at a time. This one generalises.
The practical implication: a researcher could, in theory, screen a range of nanoparticle formulations before manufacturing them, identifying the most promising candidates without running a single animal experiment. It’s not a perfect replacement — the training dataset is still relatively small, and the model’s reliability will improve as more well-documented studies are added — but as a screening tool, it offers something genuinely useful: a reasoned shortlist before the expensive work begins.
There’s also a broader concept at play here, one the researchers describe as Safe and Sustainable by Design. The idea is that safety considerations shouldn’t be an afterthought bolted on at the end of development — they should be built in from the start. A model that predicts biodistribution early makes that more feasible.
Wu’s next step is adapting the framework for human physiology — a more complex undertaking, partly because the organs involved are different and the ethical stakes of getting things wrong are higher. One application she’s specifically interested in is the blood-brain barrier: the highly selective membrane that separates the brain from the bloodstream, and that most drugs cannot cross. It’s a known bottleneck in treating brain tumours, and one where nanoparticles have long been proposed as a potential solution. Whether that potential materialises depends on finding the right particles — which is precisely what a well-validated in silico model could help identify.
It’s incremental progress, not a breakthrough announcement. But incremental progress, done rigorously, is how the field moves.
Sources
- Wu, J., et al. (2025). Data-Driven Prediction of Nanoparticle Biodistribution from Physicochemical Descriptors. ACS Nano. https://doi.org/10.1021/acsnano.5c03040
- Empa press release via News Medical, May 28 2026: Virtual AI mouse tests nanomedicine without using living animals
- Tshinanu Munyendo et al. (2022). PBPK modelling in drug development — an overview. Drug Discovery Today (for general PBPK context)
- Bhardwaj, P., et al. (2022). Nanoparticles as drug delivery systems for the blood-brain barrier. Pharmaceutics — good background on BBB + nano
- European Chemicals Agency: Safe and Sustainable by Design framework — https://echa.europa.eu/hot-topics/safe-and-sustainable-by-design
