Figure 6. Vulnerability and resilience modules in territories, neighborhoods, and the ML clusters.
(A, B) Violin plots of the vulnerability module across territories and neighborhoods, with percentage of normalized cell loss shown at the bottom per territory and neighborhood, respectively. (C, D) Violin plots of the resilience module across territories and neighborhoods, with percentage of normalized cell loss shown at the bottom per territory and neighborhood, respectively. Pairwise comparisons of vulnerability module scores across territories and neighborhoods, respectively. (E) Pairwise comparisons of resilience module scores across neighborhoods. (F) Vulnerability module gene list, ranked by average Log2FC. (G) Resilience module gene list, ranked by average Log2FC. Summary statistics center lines are mean values. na = not applicable. ns = not significant. Sample size per condition (n) = 6. See Materials and methods for the statistical tests used.
Figure 6—figure supplement 1. Normalized cell loss per cluster and its correlation with the vulnerability module.
(A) Calculated normalized cell loss for dopaminergic-only clusters from the 6 territories shown, visualized as percentages, and color-coded by territory. Section of the hierarchical dendrogram of the same clusters. The vulnerable (black triangle) and resilient (black circle) clusters are marked. See Materials and methods for details. Error bars show standard deviation, SD. (B) Heatmap of adjusted p values for the pairwise comparisons of cell loss among the same clusters in A. For the Conover-Iman test; p-value = P(T ≥ |t|), and null hypothesis (H0) was rejected at p ≤ α/2, which is 0.025. (C, D) Linear regression models, fitted for the normalized cell loss in the specified clusters versus the square root of the vulnerability module scores calculated with and without Slc6a3, respectively. The grey bands represent the 95% confidence interval limits. p-value = 3.89e-10 for C, and p-value = 1.73e-09 for D.
Figure 6—figure supplement 2. Vulnerability module scores per cluster for integrated Th+/Slc6a3+dataset.
(A) UMAP projection of integrated dataset with territory color-coding, and labels of the clusters. (B) Heatmaps of adjusted p values for pairwise comparisons of vulnerability and resilience module scores among the dopaminergic-only clusters, excluding the non-mDA clusters. See Materials and methods for the statistical tests used. (C, D) Violin plots of vulnerability and resilience module scores for all clusters in the integrated dataset.
Figure 6—figure supplement 3. Co-expression network analysis of lesioned nuclei from mDA neuron territories.
(A) Transcriptomically similar nuclei were aggregated into over 4800 Metacells, and their aggregated expression profiles are visualized in two dimensions with UMAP, grouped by their respective territories. Non-mDA clusters are colored gray. (B) Dendrogram visualization of the hierarchical clustering of genes into co-expression modules based on the topological overlap matrix (TOM). (C) Dotplot of scaled module eigengene (ME) for the 9 calculated modules across mDA territories, with their corresponding abundance per territory. (D) The co-expression network of the 9 modules, based on the gene expression of mDA territories from the lesioned side. Genes are shown as nodes, co-expression relationships between genes are shown as edges. The point size is scaled based on the kME values. Intramodular edges are the same color as their corresponding modules, while intermodular edges are colored gray. The strength of the co-expression links between genes is used to scale the opacity of the edges in the network. Network edges are downsampled to 65% for better visualization. The top 20 hub genes and other 5 randomly selected genes per module were used. (E) Correlation between each module pair based on their MEs. Modules 2 and 3 are negatively correlated while modules 2 and 8 are positively correlated as highlighted by dashed boxes. (F) Boxplot showing average kME values per module. The black line in the box indicates the median while the lower and upper hinges correspond to the first and third quartiles (the 25th and 75th percentiles). The upper / lower whiskers extend from the hinge to the largest / smallest values no further than 1.5 * IQR from the hinge. IQR = the inter-quartile range. (G) GO terms plots for the co-expression modules 2 and 3, showing the top 20 terms ranked by p-values. The enrichment analysis was performed on the top 100 genes per module, ranked by kME, using the GO_Biological_Process_2021 database.