Identifying Drugs Associated With Parkinson's Disease Risk Using Machine Learning.

Area of Science:

• Pharmacology
• Computational Biology
• Epidemiology

Background:

• Machine learning (ML) offers a promising avenue for identifying repurposed drugs for Parkinson's disease (PD).
• Previous studies have explored ML for drug discovery, but validation in large populations is crucial.

Purpose of the Study:

• To apply an ML-based signal detection method to identify drugs associated with reduced Parkinson's disease risk.
• To evaluate the generalizability of the ML method using a large case-control study.

Main Methods:

• A case-control study utilized Finnish register data, including 12,257 PD cases and 81,103 controls.
• An algorithm combining subsampling and lasso logistic regression analyzed drug exposure (ATC classification) with an 8-year lag.
• Drug dispensation frequency over two years prior to the index date was considered.

Main Results:

• Three drug subgroups and two individual drugs were associated with a decreased risk of Parkinson's disease.
• Inhalant anticholinergics, specifically tiotropium bromide, demonstrated the most significant association with reduced PD risk.
• Other potential protective agents included antimalarial drugs (aminoquinolines) and lincosamide antibiotics.

Conclusions:

• The ML-based approach is feasible for identifying potential drug repurposing candidates for Parkinson's disease.
• Tiotropium bromide and other identified drug classes warrant further investigation for their disease-modifying potential in PD.
• Observed associations may stem from direct pharmacological effects or the treatment of PD prodromal symptoms.