Over recent decades, numerous models have been developed to study PD. Although no single model fully recapitulates the complexity of human pathology, these systems provide valuable insights into disease mechanisms and therapeutic limitations.

Transgenic mouse models are typically based on pathogenic gene mutations identified in approximately 10% of hereditary PD cases. These models target genes such as SNCA, LRRK2, Parkin, PINK1, and DJ-1, which are involved in mitochondrial function, protein degradation, and oxidative stress pathways. They are particularly useful for mechanistic studies and target validation. However, their major limitation lies in their inability to fully reproduce key pathological features of human PD, especially the robust loss of dopaminergic neurons.

Neurotoxin-induced models are among the most widely used and well-established PD models. These models mimic sporadic PD by selectively damaging dopaminergic neurons through various neurotoxins. The core model relies on three types of toxins that induce lesions in the nigrostriatal pathway by impairing mitochondrial function and elevating oxidative stress. Notably, some of them can also trigger α-synuclein aggregation^[7].

MPTP penetrates the blood-brain barrier and is converted into MPP+ by MAO-B in glial cells, followed by further metabolism into active toxic compounds. MPP+ is then taken up by dopamine transporters, impairing mitochondrial respiration through the inhibition of complex I of the electron transport chain, thereby activating oxidative stress and programmed cell death pathways.

6-OHDA enters cells via dopamine transporters and induces mitochondrial dysfunction, generating ROS and quinones that lead to the degeneration of dopaminergic neurons in the substantia nigra and striatum.

Characterized by high lipophilicity, rotenone easily permeates the cell membrane, triggering both α-synuclein aggregation and mitochondrial dysfunction, ultimately resulting in the production of ROS and quinones.

Stereotactic injection of pre-formed α-synuclein fibrils (PFFs) into specific brain regions seeds endogenous α-synuclein aggregation and facilitates trans-synaptic propagation. The aggregates primarily consist of cytoplasmic Lewy body/neuronal structures. This pathology spreads along connected neural circuits, mimicking the propagation of α-synuclein from the striatum to other regions in PD. Compared to traditional toxin models, the α-synuclein PFF model better recapitulates the formation and dissemination of Lewy body pathology. Furthermore, as it induces progressive neuroinflammation and dopaminergic neuron loss, this model is widely used to study the "prion-like transmission" mechanism of α-synuclein^[8].

By using recombinant adeno-associated virus (rAAV) to overexpress human α-syn in target areas (such as the substantia nigra), the intracellular α-syn content accumulates significantly. The resulting aggregates exhibit a predominantly nuclear or fibrillary pattern and remain confined to the virus-transduced neurons. While this model effectively simulates neurodegeneration driven by excessive α-syn levels, it shows limited aggregation and weak propagation^[8].

The future urgently demands next-generation disease models that capture longitudinal, multi-system pathological features and adopt biology-centric trial designs, rather than simply optimizing for operational convenience. Such advancements will lay the groundwork for the rigorous evaluation of neuroprotective therapeutics. Currently, while machine learning models trained on small, single-center data achieve high accuracy, their poor performance in external validation highlights a susceptibility to sampling bias—a limitation shared by traditional in vivo models. Ultimately, no single platform can fully simulate the intricate interplay of aging, polygenetics, and environmental factors in PD. Therefore, multi-platform cross-validation (integrating in vitro cellular models, diverse in vivo animal strains, and in silico computational tools) is crucial.