An interpretable whole-blood transcriptomic model for Parkinson’s disease identification and molecular profiling
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Sporadic Parkinson’s disease (PD) lacks a molecular taxonomy that stratifies patients by dominant molecular drivers, limiting enrollment for disease-modifying trials, especially anti-inflammatory agents targeting NLRP3/NF-κB.
Methods
We developed ORIGIN-PD, an interpretable whole-blood transcriptomic classifier comprising 34 gene-level features and a composite low-expression score, with gene-level contributions organized into eight prespecified functional modules: midbrain dopaminergic (mDA) neuron identity/development, pathway-related, oxidative/mitochondrial, proteostasis, inflammation/ immunity, microglia, ncRNA, and other. Using LASSO regression on stringently quality-controlled whole-blood RNA-seq data from the Parkinson’s Progression Markers Initiative (PPMI) (Training Set: 204 PD and 92 controls; Locked Test Set: 51 PD and 23 controls), we assigned each patient a module-contribution profile. Module dominance was defined as >30% of total score.
Results
ORIGIN-PD discriminated PD from controls with high accuracy (Locked Test Set AUC 0.944, sensitivity 0.843, specificity 0.913). Among 255 sporadic PD patients, 49.4% exhibited a dominant module-contribution pattern—most frequently mDA neuron identity/development (22.0%), followed by pathway-related (11.8%), microglia/glial (7.5%), and inflammation/immunity (3.5%). Among the 50.6% with multi-module contributions, mDA neuron identity/development ranked in the top three in 62.8% of patients versus 35.7% for inflammation/immunity, indicating broader involvement of mDA neuron identity genes than inflammatory genes. The model showed high specificity against Alzheimer’s disease, ALS, and inflammatory controls.
Conclusions
ORIGIN-PD enables rational enrichment for NLRP3/NF-κB inhibitor trials by identifying patients with inflammatory involvement. In addition, the predominant contribution of the mDA neuron identity/development module, centred on WNT1 , MSX1 , NPM1 , and LMX1B , identifies a larger patient subgroup with altered expression of mDA neuron identity module genes, suggesting that ORIGIN-PD may also be useful for patient enrichment in future trials of Wnt/β-catenin pathway agonists. This dual-purpose framework provides a transparent, interpretable basis for pathogenic subtyping and precision enrichment in sporadic PD, with the aim of guiding disease-modifying therapy.
