Pharmacometric expertise in Marseille bridges research, routine care, and emerging plans for broader dissemination.
During the SophIA Summit, we spoke with Sébastien Benzekry, head of the Inria–Inserm COMPO (Computational Pharmacology & Clinical Oncology) team, about the role of mathematical modeling and AI in oncology. The discussion covered several complementary—but distinct—applications of pharmacometric approaches developed and used in Marseille, ranging from pioneering research studies to routine-care clinical tools, as well as current plans to further expand their reach.
COMPO is a cross-disciplinary team combining mathematicians, pharmacologists, and oncologists. Jointly affiliated with Inria and Inserm and based at the Centre de Recherche sur le Cancer de Marseille (Inserm U1068), the team develops computational models that integrate mechanistic representations of biological and pharmacological processes with statistical and machine-learning approaches. This methodology, referred to as mechanistic learning, is used to support clinical decision-making, inform the design of clinical trials, and explore biological hypotheses. Close ties with early-phase clinical trials, notably through the Marseille CLIP2 center, and with routine-care practice enable COMPO to work at the interface between research and the clinic.
It is important to distinguish between the two separate strands of work discussed during the interview. On the one hand, Marseille has been the site of a first-in-class math-based clinical trial that demonstrated how mathematical models could be used to guide chemotherapy dosing within a research protocol. On the other hand, Marseille pharmacists also use a dosing-adaptation software in routine care, based on population pharmacokinetics. While both rely on shared scientific principles, the clinical trial and the software are independent in their development and use.
READ: Immuno-Oncology: Mechanistic Learning, Digital Twins & AI
Marseille’s Singular Access to a Proven Clinical Tool
Today, Marseille remains the only city in France where hospital pharmacists routinely use this specific dosing-adaptation software grounded in population pharmacokinetics. The reason is not preferential policy or exclusivity by design, but history: the Marseille hospital purchased four licenses before the software’s original commercial trajectory slowed, allowing continued local use.
The software in Marseille supports pharmacists in adapting chemotherapy doses for individual patients, particularly in drugs known for high inter-patient pharmacokinetic variability. In routine care, the software provides a structured, user-friendly interface that translates patient data into dose recommendations, helping clinicians balance efficacy with tolerability.
“This is something that pharmacists use in routine care,” Sébastien Benzekry explained. “They adapt the dose based on the patient’s pharmacokinetics, and it works very well for certain drugs.”
Elsewhere in France, pharmacists and clinicians are aware of the tool but cannot currently access it under the same conditions. As Benzekry notes, this situation is not the result of medical disagreement or lack of interest, but of how the software’s licensing and ownership evolved over time.
Marseille’s continued use, however, has provided a rare opportunity: long-term, real-world experience with a pharmacometric tool embedded directly into daily practice. That experience continues to inform discussions about how such tools might be shared more broadly.
A Path Forward: Recent Improvements
Discussions are currently underway regarding the future development and broader distribution of the dosing-adaptation software used in Marseille. According to Benzekry, recent improvements have been made, and several avenues are being explored to ensure the tool can continue to evolve and, potentially, reach additional clinical sites under appropriate conditions. These efforts reflect a shared recognition of the software’s clinical utility and the need to adapt its deployment to current technical and regulatory environments.
“[Other model-informed dosing tools] exist, but only for a specific drug for specific therapy areas; the software in Marseille allows users to enter their own population and data for a drug, which is important in oncology because we consistently have new drugs.”
The interest in model-informed dosing tools is growing more broadly. A Swiss company has developed a modern dosing tool, for example, but Marseille cannot install it because the hospital’s IT department blocks third-party installations. An open-source, browser-based system could, in principle, bypass some of these institutional barriers by avoiding local installation requirements.
For now, Marseille remains the sole center routinely using the specified dosing-adaptation software, benefiting from long-standing experience with a pharmacometric tool embedded in daily oncology practice. With ongoing development efforts and discussions around future distribution models, there is cautious optimism that the system could become more accessible over time. As Benzekry notes:
“Patients can benefit from the use of the software.”
And if access were to expand beyond Marseille, many more could.
