Fig. 6: Aberrant expression and rare variant enrichment implicate GPC6 in Parkinson disease.
From: Neural networks reveal novel gene signatures in Parkinson disease from single-nuclei transcriptomes
A Bar plot showing the neural network (NN) disease classification balanced accuracy for dopaminergic neurons (DaNeurons) from an independent dataset prepared by Martirosyan et al. LIME-identified genes obtained from Kamath et al. or an equal number of randomly selected genes were input to the classifiers without (left) or with (right) balanced cell counts. Error bars represent the standard deviation of the balanced accuracy across three independent test-train permutations. For each permutation, we used distinct sets of randomly selected genes. Wilcoxon rank-sum tests were used to compare model performance when using the LIME-identified genes versus randomly selected genes. B Volcano plot showing the LIME feature importance Z-score and the gene expression log2 fold-change (Log2FC) between Parkinson disease (PD) and control DaNeurons from the post-mortem substantia nigra (SN). Only highly variable genes (HVG) input to the NN classifier with a LIME feature importance Z-score > 1.00 are shown. C Heatmap showing the Log2FC of the LIME-identified genes in DaNeurons from the post-mortem SN and induced pluripotent stem cell- (iPSC) derived DaNeurons. Asterisks denote significantly differentially expressed genes (adjusted P < 0.05) between individuals with PD and controls computed by MAST. D Dot plot showing GPC6 rare variant associations with various diseases across the Neurodegenerative Disease Knowledge Portal (NDKP) cohorts. E Cochran-Mantel-Haenszel (CMH) tests investigating whether individuals with PD from the PD Genome Project and International Parkinson Disease Genomics Consortium (IPDGC) Exome Sequencing Project cohorts were enriched for rare variants in GPC6. P-values corresponding to each variant category are shown on the right of the plot. The sample sizes for each cohort are shown below the plot. F Protein schematic showing the GPC6 pathogenic missense variants observed uniquely in PD or in both PD and controls across both cohorts.
