A Multi-Modal Graph Neural Network Framework for Parkinson's Disease Therapeutic Discovery

Area of Science:

• Computational biology
• Neuroscience
• Pharmacology

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder with limited disease-modifying treatments.
• Current therapeutic strategies often fail to address the multifactorial nature of PD.
• Drug repurposing offers a faster route to identify novel therapeutic agents.

Purpose of the Study:

• To identify and prioritize multi-target drug repurposing candidates for Parkinson's disease.
• To leverage network biology and deep learning for drug discovery in neurodegenerative diseases.
• To explore polypharmacology for simultaneously targeting key PD pathological pathways.

Main Methods:

• Integrated large-scale protein-protein interaction networks with a multi-modal graph neural network (GNN).
• Applied network analysis and clustering to delineate functional modules within the PD interactome.
• Utilized a novel Functional Centrality Index to identify critical network nodes.

Main Results:

• The GNN model predicted drug candidates targeting lysosomal dysfunction, mitochondrial impairment, synaptic disruption, and neuroinflammation.
• Identified specific compounds including dithiazanine, ceftolozane, DL-α-tocopherol, bromisoval, imidurea, medronic acid, and modufolin as potential candidates.
• Demonstrated the efficacy of a polypharmacology approach in uncovering repurposing opportunities.

Conclusions:

• Network-based deep learning frameworks can accelerate the discovery of multi-target therapies for PD.
• The identified drug candidates offer potential for novel therapeutic strategies.
• This approach holds promise for treating complex neurodegenerative diseases beyond Parkinson's.