broadly applicable way to estimate key safety metrics (Cmax and AUC) reasonably well. Big thanks to Zeynep Edizcan, Stephan Schaller, and Lars Kuepfer for the excellent collaboration that made this work possible.
whether this is due to greater variability in the underlying PK data and clinical study protocols, or because dermal absorption is inherently more complex. Still, we conclude that despite the challenges in capturing absorption dynamics, the method offers a pragmatic and…
One key aspect of the study is a systematic comparison of high-throughput PK prediction accuracy for dermal versus oral exposure. We found that dermal administration is more challenging to predict, though it’s unclear…
the most reliable systemic PK predictions of 52 substances. With the best strategy, 75% of Cmax and 75% of AUC values fell within a tenfold range of observed human plasma data.
be applied across pharmaceuticals, cosmetics, and industrial chemicals. We ran 14,336 PBK simulations using the mechanistic skin permeation model of the Open Systems Pharmacology Suite to identify which combinations of LogP, solubility, and other parameters yield…
In this work, we present a fully in silico HT-PBK workflow that predicts systemic exposure after dermal application using only QSAR-derived input parameters. It’s designed for early-stage use cases where no compound-specific in vitro or formulation data are available, and can…
Our new preprint on high-throughput PBK modelling for #dermal exposure is now available 📄👇
“High-throughput PBK modelling for dermal exposure: a pragmatic approach to predict systemic pharmacokinetics” doi.org/10.1101/2025...
and provide a case study showing how mathematical models (#QSP) enable deeper understanding of in vivo biological effects.
Big thanks to Ahenk Zeynep Sayin, Lars Kuepfer, Stephan Schaller for the great teamwork and discussions that made this possible.
And this in a way that is fully mechanistic and transparent. As if this was not enough, our analysis also gives new insights into why DILI often manifests in an “idiosyncratic” fashion. And on top of that, we also gain new insights into the in vivo relevance of #BSEP inhibition...
But using high-throughput PBK modelling allows us to circumvent this problem and to generate such predictions from chemical structure alone. Thereby, we can show that already today available in vitro and in silico methods allow accurate predictions of DILI.
…combining in vitro toxicity measurements and PK predicts DILI risks with up to 96% ROC AUC, all without animal testing. Traditionally, the prospective utility of this prediction approach would have been limited by the need for in vivo PK data from human studies.
In this new study, we collected a huge dataset of 241 drugs with known clinical DILI outcomes, pharmacokinetic data, as well as in vitro hepatotoxicity data from 17 in vitro assays. This allowed us to show that…
“Integration of In Vitro and In Silico Approaches Enables Prediction of Drug-Induced Liver Injury” is now available on #bioRxiv: doi.org/10.1101/2025...
Still thinking that Drug-Induced Liver Injury (DILI) is hard to foresee? That LogP and Dose are good DILI predictors? Wondering why DILI often manifests idiosyncratically?
Then you should read our new preprint! 📄👇
🤝 Interested in applying PSA to your own chemical portfolio? Let’s connect and chat!
#OpenAccess #ExposureLed #NGRA #ChemicalSafety #NAMs #PBK #RiskAssessment #Toxicology #RegulatoryScience
Huge thanks to my co-authors Andrew Worth, Alicia Paini, Lars Kuepfer and Stephan Schaller, and to everyone who debated these ideas at the NAM Designathon workshop in Ispra 🇮🇹. Your insights were invaluable!
• However, most substances fall into a “medium PSA concern” band meaning bioactivity and external exposure still matter and it is not possible to (de)prioritise the majority of chemicals on their ADME alone.
• A handful of false-positives, mainly borderline cases.
𝗞𝗲𝘆 𝗙𝗶𝗻𝗱𝗶𝗻𝗴𝘀
• Zero false-negatives: high-concern chemicals were never missed.
3. Reality check – Manual expert annotation and public-data review to see where the model shines (and where it doesn’t).
𝗢𝘂𝗿 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵
1. Theory first – We spell out when it’s scientifically sound to rely purely on ADME.
2. High-throughput PBK modelling – Two PSA scoring systems (peak & time-average) applied to 139/150 Designathon compounds.
...classify chemicals independently along TK and TD axes. Potential Systemic Availability (PSA) is our answer.
𝗧𝗵𝗲 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲
Can we (de)prioritise chemicals in safety assessments using only their ADME/TK properties before we even look at bioactivity or external exposure? This is a core question in the move toward exposure-led risk assessment, highlighted by the EPAA NAM Designathon, which asked us to...
𝗛𝗼𝘁 𝗼𝗳𝗳 𝘁𝗵𝗲 𝗽𝗿𝗲𝘀𝘀: the second peer-reviewed paper of my PhD “Potential Systemic Availability (PSA) Classification of Chemicals for Safety Assessment” has just been published in Environment International and it’s fully open access.
🔗 doi.org/10.1016/j.en...
...Ispra (Italy) last year. 🌍
For more details about our work and the EPAA Designathon:
single-market-economy.ec.europa.eu/calls-expres...).
💡 Still, I think this was a really interesting, thought-provoking exercise and it was really fun to think about computational methods can inform regulatory decision making. And especially, it was great to engage in discussions with lots of bright people at the NAM Designathon workshop in...
The crux is that most chemicals appear to simply belong in the “medium concern” category, meaning their toxicity also depends on other factors like bioactivity and exposure. Hence, it is not possible to outright prioritise or deprioritise them based on their ADME properties alone.
Key Findings: Indeed, we can correctly predict the classification of the large majority of compounds. Only for a few we get false positive mispredictions, and we have no false-negatives.
we then evaluated 139/150 compounds from the EPAA NAM Designathon. Finally, we confirm the plausibility of our predictions by manually annotating the PSA concerns of chemicals by expert-judgement and review of publicly available data.