Panel-Gene Transcriptomic Associations and Published-Variant Architecture in Parkinson Disease

Parkinson disease (PD) was examined in four primary adult-brain GEO cohorts (GSE49036, GSE20141, GSE8397, GSE7621) using a frozen 468-gene redox, metal, xenobiotic, and barrier panel. Inverse-variance random-effects meta-analysis (DerSimonian–Laird) produced 419 analyzable genes, 91 at Benjamini–Hochberg FDR < 0.05, and 56 high-confidence genes (FDR < 0.05, measured in ≥ 3 cohorts, direction-concordant in ≥ 70% of cohorts, and nominally significant in ≥ 2 cohorts). Median I² among high-confidence genes was 16.4%. High-confidence calls included higher abundance of selected iron-handling and stress-response transcripts and lower abundance of selected glutathione- and mitochondrial-related transcripts. These are transcript-abundance findings; they do not themselves prove iron accumulation, glutathione depletion, or complex-I failure. A compiled variant inventory includes PD-relevant published alleles (including CYP2D6 poor-metabolizer and GST/PON1/ABCB1 pesticide-interaction literature). Expression and variant layers are not combined into a clinical score. GVI, DDI, and IDRS remain uncalibrated research formulae and are not used for inference in this paper. No statistical pooling with Alzheimer, ALS, multiple sclerosis, or autism cohorts was performed. This paper is a disease-restricted companion to the methods-and-panel manuscript of this series (1).