Internal talk by Marco Bozza
Virtual populations are a central component of Quantitative Systems Pharmacology (QSP) and related mechanistic modeling frameworks. By reproducing the biological variability observed across individuals, they enable the simulation of heterogeneous treatment responses and support the investigation of complex physiological systems.
This topic was recently explored during a COSBI seminar presented by Marco Bozza, PhD student at COSBI, working under the supervision of Stefano Giampiccolo and Prof. Luca Marchetti.
Beyond their application in treatment-response simulations, virtual populations are also valuable tools for investigating model non-identifiability and systematically exploring parameter spaces in mechanistic models. However, generating virtual populations for large and complex systems often requires extensive computational resources, with simulations potentially lasting days or even weeks.
To address these limitations, Marco presented an approach based on the use of surrogate models to accelerate the most computationally demanding phases of virtual population generation. In this framework, the surrogate model is used to approximate the system behavior efficiently, while the original mechanistic model is primarily employed for validation purposes.
Preliminary results are promising, suggesting that this strategy can significantly reduce computational time while maintaining the reliability of the generated virtual populations.
This work highlights the growing role of hybrid computational approaches in improving the scalability and applicability of QSP workflows for increasingly complex biological systems.
