First in-patient tests with the Virtual Parkinsonian Patient

Parkinson’s disease​ is ​characterized by the loss of dopamine-producing neurons primarily in one region of the brain, the substantia nigra, which is involved in movement control. But how this local loss of neurons reverberates over the whole-brain dynamics remains poorly understood. The new model, which combines various types of brain data, explicitly includes the dynamics of Dopamine concentrations and can predict whole-brain dynamics as a function of Dopamine levels.  

In a first demonstration, the researchers showed that the new model can be inverted, yielding a prediction about the levels of dopamine at the individual subject level. To test this, ten patients were recorded utilizing EEG and deep electrodes, before and after the administration of a commonly prescribed medication, called Levodopa or L-DOPA. 

L-DOPA provokes the rise of dopamine levels in the brain and is used to alleviate movement-related symptoms of the disease. For each patient, the large-scale dynamics have been simulated (in terms of aperiodic activities, so-called neuronal avalanches) as a function of dopamine. The inversion procedure, carried out utilizing AI, correctly inferred higher and lower dopamine states in all patients tested.  

“We are now working to modify our model so that complex therapies such as deep brain stimulation might also be simulated.” explains last author Pierpaolo Sorrentino. About future uses, he adds: “We hope that our models can be deployed in clinical settings, becoming a new aid to the clinicians in order to efficiently identify the most effective therapy for each patient and, possibly, predict the onset of side-effects.”

Original Publication: Angiolelli, M., Depannemaecker, D., Agouram, H. et al. The Virtual Parkinsonian patient. npj Syst Biol Appl11, 40 (2025).https://doi.org/10.1038/s41540-025-00516-y