Computational Modeling of Parkinson's Disease Across Scales: From Mechanisms to Biomarkers, Drug Discovery, and Personalized Therapies.

Parkinson's disease (PD) is a multifactorial neurodegenerative disorder characterized by complex interactions across molecular, cellular, circuit, and behavioral scales. While experimental and clinical studies have provided critical insights into PD pathology, integrating these heterogeneous data into coherent mechanistic frameworks and translational strategies remains a major challenge. Computational modeling offers a powerful approach to bridge these scales, enabling the systematic investigation of disease mechanisms, candidate biomarkers, and therapeutic strategies. In this review, we survey state-of-the-art computational approaches applied to PD, spanning molecular dynamics and biophysical models, cellular- and circuit-level network models, systems and abstract-level simulations of basal ganglia function, and whole-brain and data-driven models linked to clinical phenotypes. We highlight how multiscale and hybrid modeling strategies connect α-synuclein pathology, mitochondrial dysfunction, oxidative stress, and dopaminergic degeneration to alterations in neural dynamics and motor and non-motor symptoms. We further discuss the role of computational models in biomarker discovery, including imaging, electrophysiological, and digital biomarkers. In particular, eye-movement-based measures are highlighted as quantitative, reproducible behavioral signals that provide principled constraints for individualized computational modeling. We also review the emerging impact of computational approaches on drug discovery, target prioritization, and in silico clinical trials. Finally, we examine future directions toward personalized and precision medicine in PD, emphasizing digital twin frameworks, longitudinal validation, and the integration of patient-specific data with mechanistic and data-driven models. Together, these advances underscore the growing role of computational modeling as an integrative and hypothesis-generating framework, with the long-term goal of supporting data-constrained predictive approaches for biomarker development and translational applications.