DOSE-DEPENDENT LEVODOPA EFFECTS IN PHENOTYPE-AWARE DIGITAL TWIN MODELS OF PARKINSON’S DISEASE

Digital twin models are increasingly used to investigate disease mechanisms and treatment responses across disorders, including epilepsy and neurodegenerative diseases (Jirsa et al., 2017; Wang et al., 2024). In Parkinson's disease (PD), a virtual patient framework introduced a whole brain multi-scale digital twin based on the Dopamine Dynamics (Dody) model, showing higher inferred dopaminergic tone in the ON-medication (levodopa administration) state compared to the OFF-medication state (Angiolelli et al., 2025). However, this framework does not model how levodopa dosage shapes brain dopamine concentration across PD stages at key treatment milestones. To address this gap, we integrated a levodopa pharmacokinetic and pharmacodynamic model into the Dody framework (Baston et al., 2016), enabling simulation of dose-dependent dopamine concentration dynamics across different PD phenotypes. Subcortical regions together with the cortical Desikan parcellation were used to construct structural connectomes (Desikan et al., 2006). Cortical regional membrane potentials were mapped to scalp electroencephalography (EEG) using a forward lead-field model, while activity of the subthalamic nucleus (STN) was obtained directly from its membrane potential, consistent with recordings from sensing-capable deep brain stimulation electrodes placed near the left and right STN. Large-scale brain dynamics were characterized using avalanche transition matrices (ATM) (Angiolelli et al., 2025) and EEG microstate dynamics (Trujillo-Barreto et al., 2024). Differences in EEG microstate dynamics and ATM patterns across levodopa doses and PD phenotypes were analyzed to identify optimal dosage profiles for each treatment milestone. This work provides a mechanistic forward model for studying dose-dependent levodopa effects with potential impact on personalized medicine.