Transcriptomic Analysis and Multiple Machine Learning Approaches Identify ZDHHC20 and Its Highly Correlated Gene AK5 as Diagnostic Markers in Multiple System Atrophy

Transcriptomic Analysis and Multiple Machine Learning Approaches Identify ZDHHC20 and Its Highly Correlated Gene AK5 as Diagnostic Markers in Multiple System Atrophy | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Transcriptomic Analysis and Multiple Machine Learning Approaches Identify ZDHHC20 and Its Highly Correlated Gene AK5 as Diagnostic Markers in Multiple System Atrophy Zhipeng Lu, Zhongqi Li, Zhibiao Yin, Jialong Liu, Pu Fang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8711800/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background Multiple system atrophy (MSA) is a fatal neurodegenerative disorder lacking effective diagnostic tools. While protein palmitoylation is crucial for neuronal function, its specific role in MSA pathogenesis remains unexplored. Methods We integrated bulk and single-nucleus RNA sequencing (snRNA-seq) data from postmortem MSA brain tissues. Eight machine learning algorithms were utilized to screen palmitoylation-related genes. Downstream analyses, including functional enrichment, cellular deconvolution, and pseudotime trajectory inference, were then conducted. Results ZDHHC20 and its highly correlated gene, AK5, were identified as hub genes. Both demonstrated significant downregulation in MSA, particularly within the cerebellar white matter. Functional enrichment analysis linked this expression pattern to mitochondrial dysfunction and impaired energy metabolism. Furthermore, snRNA-seq revealed that ZDHHC20 and AK5 are predominantly expressed in oligodendrocytes and are progressively suppressed during the abnormal terminal differentiation trajectory observed in MSA. Conclusions ZDHHC20 and AK5 represent promising diagnostic biomarkers for MSA. These findings highlight the potential role of palmitoylation in MSA pathogenesis, providing new insights into the diagnosis and treatment of MSA. Multiple System Atrophy (MSA) Palmitoylation Machine learning Energy metabolism Oligodendrocytes Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 1. Introduction Multiple System Atrophy (MSA) is a rare neurodegenerative disease that has garnered significant medical attention due to its sporadic nature, rapid progression, and fatal outcome (Krismer, et al., 2024; Poewe, et al., 2022). The clinical presentation primarily manifests as any combination of cerebellar syndrome, parkinsonism, and autonomic dysfunction (Wenning, et al., 2022). As a synucleinopathy, its pathogenesis is not fully understood. The prevailing view posits that the abnormal accumulation of misfolded α-synuclein (αSyn) as glial cytoplasmic inclusions (GCIs) within oligodendrocytes in affected brain regions—including the striatonigral and olivopontocerebellar systems—plays a critical role in the pathogenic process. This is the specific pathological hallmark of MSA (Wan, et al., 2023). Due to the high heterogeneity of its clinical manifestations and the symptomatic overlap with other disorders—including Parkinson's disease (PD), progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and sporadic adult-onset ataxia of unknown etiology (SAOA)—MSA is frequently misdiagnosed (Ndayisaba, et al., 2025 ; Poewe, et al., 2022; Wan, et al., 2023). Currently, the diagnosis of MSA is predominantly based on clinical history and neurological examination. examinations, including autonomic function tests, electromyography (EMG), and neuroimaging, can be employed to support the diagnosis or rule out other conditions (Wenning, et al., 2004). Potential MSA biomarkers from various sample sources (e.g., body fluids, tissues, gut microbiota, and imaging) have also been developed, but their clinical application remains limited (Wan, et al., 2023). For patients diagnosed with MSA, there are currently no effective treatments to slow disease progression; thus, MSA patients often have a poor prognosis (Piras, et al., 2020; Schweighauser, et al., 2020). Against this backdrop, exploring the pathogenesis of MSA and identifying effective diagnostic and therapeutic strategies are urgently needed. Palmitoylation is a common post-translational modification. Its mechanism involves the covalent attachment of a saturated fatty acid, such as palmitic acid, to amino acid residues of a substrate protein (Huang, et al., 2025; Zhang, et al., 2024). This modification subsequently influences various aspects of the substrate protein, including structure, function, and membrane localization (Ko and Dixon, 2018 ). Palmitoylation is widely present in various cellular metabolic activities, and i is also heavily involved in important processes such as neuronal development and maturation, the establishment of synaptic plasticity, and neural signal transduction (F, et al., 2024). Nearly half of all synaptic proteins and almost the entire myelin proteome are substrates for S-palmitoylation, which is the most common form of palmitoylation (F, et al., 2024; Jeong, et al., 2025). Some palmitoyltransferases, such as ZDHHC5 and ZDHHC9, are essential for oligodendrocyte maturation and myelin formation (F, et al., 2024). Owing to its pivotal role in synaptic plasticity, palmitoylation also assumes a crucial role in advanced cognitive functions (Buszka, et al., 2023). Recent studies have shown that the pathogenesis of multiple neurodegenerative diseases, including Alzheimer's disease (AD), Parkinson's disease (PD), and Amyotrophic Lateral Sclerosis (ALS), is closely associated with palmitoylation (Cho and Park, 2016 ). However, no studies have yet linked palmitoylation to multiple system atrophy. In recent years, bioinformatics and machine learning algorithms have developed rapidly. Machine learning algorithms can analyze complex and diverse data to identify optimal biomarkers and have therefore been widely applied in the medical field (Qi, et al., 2025). Bioinformatic analysis is a common analytical method for studying phenotypes and diseases; however, bioinformatics analyses regarding MSA are relatively scarce. From a bioinformatics perspective, we investigated the role of palmitoylation in MSA, identified palmitoylation-related biomarkers through various machine learning and logistic regression analyses, and discussed their functions and upstream regulatory mechanisms. This facilitates an understanding of MSA pathogenesis from a new perspective and provides a direction for the diagnosis and even personalized treatment of MSA. 2. Materials and Methods 2.1. Data acquisition The datasets GSE199715 (n = 94) and GSE199258 (n = 38) were derived from bulk RNA sequencing data of cerebellar white matter from postmortem MSA patients in different brain banks. GSE199724 (n = 12) consists of oligodendrocyte samples extracted from the cerebellar white matter tissue of postmortem MSA patients for bulk RNA sequencing (Piras, et al., 2020). We selected dataset GSE199715 as the discovery cohort to identify hub genes, used GSE199258 as the validation set, and employed GSE199724 to explore the expression of these genes in oligodendrocytes. All datasets above were obtained from the Gene Expression Omnibus (GEO) database ( http://www.ncbi.nlm.nih.gov/geo ) . The snRNA-seq dataset (Syn52662231) utilized in this study was obtained from the Synapse database ( https://www.synapse.org/ ). It consists of snRNA-seq data derived from postmortem prefrontal cortex tissues of individuals with Parkinson's disease (n = 20), Parkinsonian-type multiple system atrophy (MSA-P, n = 6), and healthy controls (HC, n = 13). Data from the MSA-P and HC groups were extracted for analysis(Nido, et al., 2025). A list of palmitoylation-regulated genes (PRGs), including 27 palmitoyltransferase genes and 9 depalmitoylation-related enzyme genes, was obtained from previous literature (Chen, et al., 2024 ; Fan, et al., 2024; Lucas, et al., 2016). The 36 PRGs are shown in Table S1 . 2.2. Screening and analysis of DEGs In this study, the "DESeq2" package in R was used to perform differential expression analysis on genes from MSA patients and healthy controls (HC) in GSE199715. Genes with an average expression level (counts) ≤ 10 were filtered out. The criteria of |log2FoldChange| > 1 and adjusted P-value < 0.05 were applied to screen for differentially expressed genes (DEGs). A heatmap of the DEGs was generated using the "pheatmap" R package, and a volcano plot was drawn using the "ggpubr" R package. We carried out Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of the DEGs using the "clusterProfiler" R package (Yu, et al., 2012). Enrichment terms meeting the criteria of P-value < 0.05 and q-value < 0.05 were considered significant. The resulting biological functions and pathways were visualized using the "ggplot2" R package. 2.3. Identification of hub genes using machine learning models Eight machine learning models were used to screen for palmitoylation-related core genes: Gradient Boosting Machine (GBM), Generalized Linear Model (GLM), Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), Neural Network (NNET), and Decision Tree (DT). We evaluated model performance using multiple metrics, including residuals, precision, recall, and the area under the curve (AUC). These models were used to assess the importance of PRGs, and the top 10 genes were ranked based on the feature importance scores of each model. Appropriate models were selected by integrating these performance metrics. The genes identified from the selected models were intersected with the DEGs using a Venn diagram to identify the hub genes. 2.4. Screening for ZDHHC20-correlated genes MSA samples were divided into high- and low-expression groups based on the mean expression level of ZDHHC20 in the GSE199715 dataset. The "DESeq2" R package was used to perform differential analysis between these two groups. Genes meeting the criteria of |log2FoldChange| > 2 and adjusted P-value 0.8 and P-value < 0.05 were considered highly correlated. The intersection of these highly correlated genes and the ZDHHC20-related DEGs was determined . Next, we further screened these intersected genes using LASSO and SVM algorithms. We employed the "glmnet" package to conduct the LASSO algorithm and identify the penalty coefficient (λ) associated with the minimum partial likelihood deviance. The "e1071" R package was used to implement the SVM with 10-fold cross-validation to prevent overfitting (Leisch, 2024 ). Genes identified by the intersection of LASSO and SVM results were advanced to univariate and multivariate logistic regression analyses. In the univariate analysis (using the "survival" R package), genes with a P-value < 0.05 were included in the multivariate analysis to determine the final key genes. 2.5. Gene set enrichment analysis (GSEA) To explore the potential biological mechanisms and pathways of the hub palmitoylation-related genes (HPRGs), we conducted GSEA using the "clusterProfiler" R package. The "c5.go.v2025.1.Hs.symbols.gmt" (GO) and "c2.cp.kegg_medicus.v2025.1.Hs.symbols.gmt" (KEGG) gene sets were downloaded from the MsigDB database ( http://www.gsea-msigdb.org/gsea/msigdb ). A threshold of False Discovery Rate (FDR) < 0.05 was considered statistically significant. 2.6. Cellular deconvolution To investigate the cellular composition of cerebellar tissue, we employed cell type deconvolution to quantify the proportions of CNS (central nervous system)-resident cells and infiltrating immune cells between the MSA and HC groups by utilizing the "CIBERSORT" package(Chen, et al., 2018). A signature matrix file encompassing six cell types ( oligodendrocytes, excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, and microglia) in the CNS was used in our study(Wang, et al., 2020 ). The results were visualized through the application of the "ggplot2" package. Correlations between genes and immune cells were computed using Spearman's rank correlation test and presented as heatmaps via the "pheatmap" package. 2.7. HPRG expression visualization, ROC analysis and correlation scatter plots The "ggplot2" R package was used to visualize the differential expression of HPRGs. The "pROC" package was employed to plot receiver operating characteristic (ROC) curves, and the area under the curve (AUC) was calculated to evaluate diagnostic value. The "ggplot2" and "ggpubr" packages were used to generate correlation scatter plots. 2.8. ceRNA and gene-TF regulatory network We utilized TargetScan ( http://www.targetscan.org/ ), miRTarBase ( https://mirtarbase.cuhk.edu.cn/ ), and miRDB ( http://mirdb.org/ ) databases to predict the target microRNAs (miRNAs) of HPRGs. The ENCORI database ( http://rnasysu.com/encori/ ) was employed to predict target long non-coding RNAs (lncRNAs) (Li, et al., 2014; Zhou KR).The NetworkAnalyst online tool ( http://www.networkanalyst.ca/ ) was used to analyze interactions between hub genes and transcription factors (TFs), with TF targets derived from the JASPAR database (Rauluseviciute, et al., 2024). Cytoscape software was used to build the competing endogenous RNA ceRNA and TF-gene interaction network. 2.9. snRNA-seq analysis Single-nucleus data from 6 MSA-P and 13 HC samples (Syn52662231) underwent processing using the “Seurat” package(Hao, et al., 2024). Quality control (QC) filtered cells based on following criteria: 500 < nFeature_RNA < 6000 and percent.mt < 5. Data underwent normalization (“NormalizeData” function in Seurat) and subsequent identification of Highly Variable Genes (HVGs) via “FindVariableFeatures”. Scaling for downstream dimensionality reduction was performed using “ScaleData”. The optimal number of principal components (PCs) was determined using an elbow plot. Dimensions were selected based on two primary criteria: the cumulative explained variance exceeded 90% while the single PC contribution remained below 5%, and the variance difference between adjacent PCs fell below 0.1%. Clustering resolution was established using a clustree plot, selecting the resolution that exhibited the most stable downward clustering change. Cell clusters were then identified using Seurat's “FindClusters” function. Cell type annotation was performed using previously established cell markers(Mathys, et al., 2019). Nonlinear dimensionality reduction was then conducted using “RunUMAP”, with results visualized via the “DimPlot” function. Gene expression across cell types was displayed using “VlnPlot” and “DotPlot”, and the distribution of target genes across cells was visualized with “FeaturePlot” The selected core cell types subsequently underwent dimensionality reduction and clustering utilizing this identical procedure. 2.10. Pseudotime analysis To characterize the temporal trajectory and biomarker expression dynamics of core cell types, we conducted pseudotime trajectory analysis. Using the Monocle2 package(Qiu, et al., 2017; Qiu, et al., 2017; Trapnell, et al., 2014), a CellDataSet object was generated, with size factors and dispersion calculated via the “estimateSizeFactors” and “estimateDispersions” functions. Transcripts were filtered to retain those with average expression above 0.1 and dispersion exceeding model-fitted values. Subsequently, the DDRTree algorithm was conducted to reduce dimensionality, and cells were ordered along the pseudotime trajectory using these filtered genes. We visualized the single nucleus trajectory using the “plot_cell_trajectory” function. The dynamic changes in biomarker expression along pseudotime were also visualized. Ridge plots further illustrated cell density distributions across the pseudotime axis for both disease and control groups, as well as specific cell subpopulations. 2.11. Statistical analysis All statistical analyses were performed using R software (version 4.4.3). All parameter analyses were conducted using two-tailed tests. The Wilcoxon rank-sum test was employed to compare differences between the two groups. Spearman's rank correlation test was used for all correlation analyses. The Wald test was used to evaluate the significance of the logistic regression analysis. A P-value or FDR less than 0.05 was considered statistically significant. 3. Results 3.1. Identification and functional enrichment of DEGs between MSA and HC Figure 1 illustrates the workflow of the study. In the GSE199715 dataset, a principal component analysis (PCA) plot demonstrated a clear separation between the MSA and HC groups (Fig. 2 A). Differential expression analysis identified 93 DEGs, including 81 upregulated and 12 downregulated genes (Fig. 2 B, 2 C). The GO enrichment analysis of these DEGs revealed enrichment in Biological Processes (BP) related to protein conformation and cell junctions; Cellular Components (CC) associated with cellular structure; and Molecular Functions (MF) involved in protease activity regulation (Fig. 2 D). KEGG enrichment analysis yielded only two significant pathways, suggesting low enrichment of these DEGs in gene sets related to KEGG pathways (Fig. 2 E). 3.2. Discovery of ZDHHC20 via machine learning Eight machine learning algorithms were applied to rank the importance of 36 PRGs. Based on a comprehensive evaluation of model performance (AUC, residuals, precision, F1-score, and accuracy), GBM, LASSO, and SVM were selected as the optimal models (Fig. 3 A, B, Table S2 ). GBM demonstrated the highest AUC, precision, and accuracy (Fig. 3 A, B, E, Table S2 ). LASSO showed the second-highest AUC along with the smallest residual value (Fig. 3 A, B, G, Table S2 ), and SVM yielded the best recall value (Fig. 3 A, B, I, Table S2 ). Each model ranked genes by feature importance (Fig. 3 C). The top 10 genes and the ROC from each of the three selected models are presented in Fig. 3 D–I. The intersection of the 93 DEGs and the top 10 genes from these three models yielded a hub gene: ZDHHC20 (Fig. 3 J). 3.3. Identification of AK5 as the key ZDHHC20-correlated gene To find genes related to ZDHHC20, we stratified MSA samples into ZDHHC20-high and ZDHHC20-low groups based on the mean expression levels. Differential analysis between these groups identified 142 upregulated and 1,222 downregulated genes (Fig. S1 A, S1B). Concurrently, correlation analysis identified 197 genes highly correlated with ZDHHC20, with the threshold of correlation coefficient (Cor) > 0.8 and P < 0.05. The intersection of these two gene sets (ZDHHC20-related DEGs and highly correlated genes) yielded 29 key genes (Table S3), excluding ZDHHC20 itself (Fig. S1 C). Functional enrichment analysis showed that these 29 genes were enriched in terms related to vesicle transport, such as Golgi apparatus subcompartment, focal adhesion, and exocytic vesicle (Fig. S1 D). To further refine these 29 genes, we applied LASSO and SVM algorithms. The LASSO analysis results showed the penalty parameter associated with the minimum partial likelihood deviance (Fig. 4 A, 4 B). Ultimately, 14 genes were successfully selected. To achieve more stable and reliable results and prevent overfitting, the SVM algorithm employed 10-fold cross-validation. As depicted in Fig. 4 C and 4 D, when the top 11 genes are selected, the accuracy reaches its optimal value and the error is minimized. The intersection of the LASSO and SVM results yielded 10 genes for logistic regression (Fig. 4 E). Univariate analysis pinpointed five genes (SLCO1A2, PDGFRA, TF, CDKN1C, and AK5) that were associated with MSA (P < 0.05). Intriguingly, all of these genes exhibited an odds ratio (OR) < 1 (Fig. 4 E). In the subsequent multivariate analysis, only AK5 remained significantly associated with MSA status (P < 0.05), suggesting it may be a protective factor (Fig. 4 F). A scatter plot confirmed a strong positive correlation between ZDHHC20 and AK5 (R = 0.88, P < 2.2e-16) (Fig. 4 G). Violin plots showed that both ZDHHC20 and AK5 were significantly downregulated in the MSA group (Fig. 4 H), consistent with the heatmap in Fig. 2 C. ROC curve analysis demonstrated good diagnostic performance for both ZDHHC20 (AUC = 0.729) and AK5 (AUC = 0.718) (Fig. 4 I). We finally defined ZDHHC20 and AK5 as hub palmitoylation-related genes (HPRGs) for subsequent analysis. 3.4. GSEA reveals similar enrichment profiles for ZDHHC20 and AK5 Single-gene GSEA was performed to explore the functions of HPRGs. In the GO analysis, pathways involving establishment of mitochondrion localization microtubule mediated, NADH dehydrogenase activity, protein targeting to lysosome, oxidoreductase activity acting on NADPH quinone or similar compound as acceptor, and oxidoreduction driven active transmembrane transporter activity were enriched in the ZDHHC20 high-expression group, while keratin filament, olfactory receptor activity, sensory perception of smell, sensory perception of chemical stimulus, and detection of chemical stimulus were enriched in the low-expression group (Fig. 5 A). AK5 showed a similar enrichment profile: establishment of mitochondrion localization microtubule mediated, oxidoreductase activity acting on NADPH quinone or similar compound as acceptor, regulation of mitochondrial fission, mitochondrial fission, and NADPH dehydrogenase activity were activated in the high-expression group, whereas keratin filament, olfactory receptor activity, sensory perception of smell, sensory perception of chemical stimulus and CUL2 ring ubiquitin ligase complex were suppressed (Fig. 5 C). The KEGG pathways significantly enriched for ZDHHC20 and AK5 were almost identical, including dynein recruitment to the kinetochore, mitochondrial complex UCP1 in thermogenesis, electron transfer in complex I, mutation caused aberrant ABETA to electron transfer in complex I, mutation caused aberrant SNCA to electron transfer in complex I, mutation caused aberrant TDP43 to electron transfer in complex I, and mutation inactivated PINK1 to electron transfer in complex I (Fig. 5 B, 5 D). The “translation initiation” pathway was enriched only for ZDHHC20 (Fig. 5 B). These terms are predominantly associated with energy metabolism, suggesting that the downregulation of HPRG might be linked to dysregulated intracellular energy metabolism. 3.5. Decreased oligodendrocyte abundance in MSA and its positive correlation with HPRG expression In this study, the "CIBERSORT" package was used to assess the cellular composition between MSA and control groups. Figure 6 A depicts a box plot showing differences in the proportions of six cell types between MSA samples and controls. It reveals that the proportions of endothelial cells and inhibitory neurons were significantly higher in MSA, while oligodendrocytes were less prevalent in MSA (P-value < 0.05). Subsequently, we calculated the correlations between HPRGs and cell types (Fig. 6 B). Oligodendrocytes showed positive correlations with both ZDHHC20 and AK5, whereas excitatory neurons displayed negative correlations. Additionally, scatter plots depicting the correlations between HPRGs and cells were generated (Fig.s 6C–6D). 3.6. Validation and subtype-specific expression patterns of ZDHHC20 and AK5 in MSA We used the independent validation set GSE199258 to confirm our findings. In this dataset, ZDHHC20 and AK5 were significantly downregulated in the MSA group (Fig. 7 A), consistent with our initial findings (Fig. 4 H). ROC analysis validated their good diagnostic performance (Fig. 7 B). The strong coefficient was also confirmed (R = 0.77, P = 2.1e-08, Fig. 7 C). Together, these findings suggest that the differential expression of HPRGs is not coincidental. We next explored HPRG expression in MSA clinical subtypes. Compared to the HC group, the expression of ZDHHC20 and AK5 was significantly downregulated in both the MSA cerebellar type (MSA-C) and the parkinsonian type (MSA-P). No significant difference was observed between the two clinical subtypes (Fig. 7 D, 7 E). Subsequently, we analyzed the expression in pathological subtypes. ZDHHC20 was significantly decreased in the olivopontocerebellar atrophy (OPCA) and striatonigral degeneration-olivopontocerebellar atrophy (SND-OPCA) subtypes compared to HC, but not in the striatonigral degeneration (SND) subtype (Fig. 7 F). Furthermore, ZDHHC20 expression in the OPCA group was significantly lower than in the SND group, while there was no significant difference between SND-OPCA and either OPCA or SND (Fig. 7 F). AK5 expression patterns were highly similar: expression was significantly downregulated in the OPCA and SND-OPCA subtypes compared to HC, but not in the SND subtype (Fig. 7 G). AK5 expression in the SND subtype was significantly higher than that in the OPCA and SND-OPCA subtypes, and there was no significant difference between OPCA and SND-OPCA (Fig. 7 G). Finally, we conducted a further exploration of the expression status of HPRGs in oligodendrocyte samples. Both genes showed a non-significant downward trend in MSA samples (Fig.. 7H). 3.7. Construction and analysis of upstream regulatory networks for ZDHHC20 and AK5 Target miRNAs for HPRGs were predicted using the miRTarBase, miRDB, and TargetScan databases. Eleven of these miRNAs had predictable target lncRNAs in the ENCORI database. Eight miRNAs (hsa-miR-653-5p, hsa-miR-548o-3p, hsa-miR-6509-3p, hsa-miR-2114-5p, hsa-miR-30d-5p, hsa-miR-642b-5p, hsa-miR-3200-5p, and hsa-miR-20b-5p) were associated with ZDHHC20, and three (hsa-miR-134-5p, hsa-miR-642b-3p, and hsa-miR-342-3p) were related to AK5. We predicted 35 common lncRNAs from these two groups of miRNAs (Table S4) and constructed a ceRNA network (Fig. 8 A). Based on this network, we were able to gain a rough understanding of the common miRNAs shared between two genes and lncRNAs. The size and color of the lncRNA nodes, along with the size of their respective labels, represent the number of miRNAs with which the lncRNAs are associated (denoted as “degree”). We also investigated upstream transcription factors (TFs) using the NetworkAnalysis platform. The resulting TF-gene interaction network identified GATA2 as a common TF for both ZDHHC20 and AK5 (Fig. 8 B). 3.8. Single-nucleus RNA-seq analysis reveals oligodendrocyte downregulation of HPRGs in MSA Quality control yielded 165,577 cells and 24,214 genes (Fig. S2 A). Subsequent dimensionality reduction utilized the top 2,000 HVGs (Fig. S2 B). Twenty-four PCs were selected (Fig. S2 C), and a resolution of 1 was determined for cell clustering based on results shown in Fig. S3. This process classified 50 cell subpopulations (Fig. S2 D), which were annotated into eight major cell types: excitatory neurons (65487), inhibitory neurons (23765), astrocytes (18636), oligodendrocytes (42310), microglia (5096), oligodendrocyte progenitor cells (OPCs) (7153), endothelial cells (1601), and pericytes (1529) (Fig. 9 A, Fig. S4). Fig.s 9B and Fig. S5A illustrate the composition and distribution of these cell types in the MSA and HC groups, respectively. Analysis of HPRG expression patterns revealed that ZDHHC20 was significantly expressed across all cell types except pericytes, whereas AK5 expression was restricted to oligodendrocytes, excitatory neurons, and inhibitory neurons (Fig. 9 C). Crucially, both genes exhibited peak expression in oligodendrocytes (Fig. 9 C, D), suggesting pivotal roles within this lineage. Within the dot plot (Fig. 9 C), both HPRG expression levels and the proportion of expressing cells were reduced in the MSA group relative to controls in oligodendrocytes, corroborating our findings in Fig. 7 H. Distribution plots (Fig. S5B, S5C) showed that ZDHHC20 and AK5 expression was generally high and densely concentrated in oligodendrocytes, consistent with Fig. 9 C and D. Consequently, oligodendrocytes were selected for subsequent analysis. 3.9. MSA exhibits a distinct oligodendroglial composition compared to HC Analysis commenced with 42,310 oligodendrocytes. Dimensionality reduction and clustering leveraged the top 2,000 HVGs and 19 PCs, ultimately identifying 19 clusters at a resolution of 0.5 (Fig. 10 A, Fig. S6). This clustering revealed significant compositional shifts between the multiple system atrophy (MSA) and healthy control (HC) groups (Fig.s 10B and 10C). For instance, six clusters (1, 7, 10, 13, 16, 18) were nearly exclusive to MSA, whereas nine clusters (0, 2, 4, 5, 8, 9, 11, 12, 14) predominated in HC (Fig. 10 B). HPRG expression patterns across these 19 subpopulations are detailed in Fig. 10 D. 3.10. Abnormal terminal differentiation and dynamic downregulation of HPRGs along the pseudo-temporal trajectory in MSA Pseudotime analysis was conducted on a random subset of 20,000 microglial cells (N = 42,310) using 1,418 HVGs for dimensionality reduction (Fig. S7A). Cells were subsequently ordered along a computed developmental trajectory, with color intensity indicating differentiation time (Fig. 11 A). Seven distinct differentiation states were identified along the trajectory (Fig. S7B). Concurrently, the relative expression levels of ZDHHC20 and AK5 consistently decreased with increasing pseudotime (Fig. 11 B). Group distribution densities showed divergent, opposing trends (Fig. 11 C): HC cells were concentrated in the early and middle pseudotime stages, whereas MSA cells demonstrated significant late-stage accumulation. This pattern suggests that oligodendrocytes in MSA patients may progress toward an abnormal terminal differentiation state. Visualization of subpopulation distribution further revealed that MSA-specific clusters were primarily restricted to the mid-to-late stages (excluding cluster 18, which occupied the early-to-mid phases). Conversely, HC-specific clusters (e.g., clusters 1, 2, 4, 9, 11, 14) were concentrated in the early and middle stages, with cluster 8 representing an exception, located predominantly in the late stage (Fig. 11 D, Fig. S7C). Collectively, these findings indicate that the relative suppression of ZDHHC20 and AK5 expression in MSA oligodendrocytes is associated with pathological alterations during disease onset and progression. 4. Discussion Multiple System Atrophy (MSA), a sporadic, adult-onset, rapidly progressive, and ultimately fatal neurodegenerative disease, presents significant challenges in both diagnosis and treatment (Ndayisaba, et al., 2025 ; Poewe, et al., 2022; Wan, et al., 2023). Therefore, it is necessary to identify effective diagnostic markers and therapeutic targets. Palmitoylation, one of the most common protein post-translational modifications, plays an indispensable role in neuronal development, synaptic plasticity, neural signal transduction, and neuroinflammation (B and Talwar, 2025 ; F, et al., 2024). Abnormal palmitoylation is often associated with neurodegenerative disorders (Cho and Park, 2016 ). Currently, the relationship between palmitoylation and MSA remains unclear, and systematic analyses are lacking. Consequently, we performed a bioinformatics analysis to examine the expression patterns of palmitoylation-related genes in MSA and their potential functions. In this study, by integrating transcriptomic data, we identified two palmitoylation-associated genes, ZDHHC20 and AK5, and discovered a strong correlation between them; these two genes may serve as promising biomarkers. Our findings may contribute to elucidating and refining the pathogenesis of MSA. palmitoylation, the most prevalent form of palmitoylation, is catalyzed by a family of enzymes known as palmitoyltransferases (PATs). This enzyme family typically harbors a conserved cysteine-rich zinc finger-like domain that encompasses the aspartate-histidine-histidine-cysteine (Asp-His-His-Cys, DHHC) sequence. Consequently, this family of enzymes is designated as the ZDHHC protein family (Buszka, et al., 2023; Fan, et al., 2024). These enzymes function by covalently attaching a 16-carbon saturated fatty acid, palmitic acid, to the cysteine residues of substrate proteins via thioester bonds, playing a pivotal role in diverse biological processes (Ko and Dixon, 2018 ). ZDHHC20, a member of ZDHHCs, has been predominantly studied in the context of tumors and the immune system. For example, one study showed that ZDHHC20 regulates cellular sensitivity to ferroptosis by palmitoylating GPX4 that is a core inhibitor of ferroptosis. Inducing ferroptosis is associated with the suppression of tumor occurrence and metastasis (Huang, et al., 2025). Inhibition of ZDHHC20 can lead to hyperactivation of the EGFR signaling pathway, thereby increasing tumor sensitivity to specific targeted therapies. Interferon-induced transmembrane protein 3 (IFITM3) is regulated by palmitoylation, and ZDHHC20 can significantly enhance the antiviral activity of IFITM3 (McMichael, et al., 2017). ZDHHC20 also plays a critical role in regulating the function of the T-cell co-stimulatory molecule CD80 (Lu, et al., 2024). Despite the fact that ZDHHC20 assumes a pivotal role in tumorigenesis and immune regulation, data from The Human Protein Atlas ( https://www.proteinatlas.org ) suggest that it is also prominently expressed in brain tissue. Furthermore, research has found that ZDHHC20 is considerably expressed in oligodendrocytes (Jeong, et al., 2025). Recently, numerous Mendelian Randomization (MR) studies have established associations between it and a variety of psychiatric disorders (Guo, et al., 2025). These findings suggest that ZDHHC20 may play an important role in the brain. This study is the first to find that ZDHHC20 expression is downregulated in MSA patients, suggesting that it may be a potential protective factor in MSA. Adenylate kinases (AKs) are phosphotransferase enzymes widely present in organisms. They are responsible for catalyzing the reversible interconversion of adenine nucleotides and play a crucial role in maintaining adenine nucleotide balance and energy homeostasis (Fujisawa, 2023 ; Siddiqui, et al., 2024 ). Adenylate kinase 5 (AK5) was first discovered in brain tissue in 1999 and was considered to be specifically expressed in the brain (Van Rompay, et al., 1999 ). Since then, elevated expression of AK5 has been detected in the neurons of the cerebral cortex and hippocampus. These brain regions are closely related to higher cognitive functions, learning, and memory (Siddiqui, et al., 2024 ). As a member of the AK family, AK5 participates not only in nucleotide conversion and energy metabolism but is also associated with the pathogenesis of various diseases, including cancer, asthma, diabetes, and neurological disorders (Fujisawa, 2023 ; Siddiqui, et al., 2024 ). Notably, AK5 has been found to be significantly downregulated in the substantia nigra of patients with mid- to late-stage Parkinson's disease (PD) (Garcia-Esparcia, et al., 2015). This phenomenon has also been observed in the entorhinal cortex and frontal cortex of patients with late-stage Alzheimer's disease (AD) (Ansoleaga, et al., 2015). A recent study indicated that AK5 downregulation may be associated with the onset of temporal lobe epilepsy (Lai, et al., 2016). By analyzing AK5 expression levels in the cerebellar white matter of MSA patients and healthy controls, we also observed this downregulation phenomenon. ZDHHC20 and AK5 exhibit overlapping functional pathways. AK5 is predominantly associated with energy metabolism, with significant enrichment of these pathways observed in the AK5 high-expression group—consistent with its established biological role (Siddiqui, et al., 2024 ). Notablely, the KEGG biological pathways enriched for ZDHHC20 and AK5 were nearly identical. Most of these pathways are related to mitochondria and respiratory chain complex I. Furthermore, GO analysis of ZDHHC20 identified many energy metabolism-related pathways, some of which also appeared in the AK5 enrichment results. These findings suggest a potential functional interaction between ZDHHC20 and AK5. While the precise mechanism remains to be elucidated, palmitoylation is known to play a critical role in the establishment and maintenance of normal neurological function. (B and Talwar, 2025 ; F, et al., 2024). Therefore, we speculate that the palmitoylation activity of ZDHHC20 may influence the expression of AK5 through certain mechanisms, such as by regulating transcription factors. Furthermore, mitochondria are the "powerhouses" of the cell, and their dysfunction is a common feature of many neurodegenerative diseases (Mantle and Hargreaves, 2022 ). This pathological feature may also involve in the pathogenesis of MSA(Krismer, et al., 2024; Wan, et al., 2023). Our results suggest that ZDHHC20 and AK5 may be involved in pathological processes in MSA similar to those in the aforementioned neurodegenerative diseases (Fig. 5 B, 5 D). Consequently, we propose that downregulation of ZDHHC20 and AK5 may exacerbate MSA progression by impairing mitochondrial function. In our study, the overall expression levels of ZDHHC20 and AK5 were significantly downregulated in the cerebellar white matter tissues of MSA patients, and a positive correlation was observed between the expression of these two genes. In human white matter tissue, oligodendrocytes are the predominant cell type, and these two markers exhibit considerable expression levels in oligodendrocytes (Fig. 9 D). We speculate that their low expression may result from extensive loss and death of oligodendrocytes in the affected tissues of late-stage MSA samples, while their high correlation may be related to the degree of oligodendrocyte loss—though the possibility of co-expression between them cannot be ruled out. Analyses of various MSA subtypes in the study also indirectly support this notion. As shown in Fig. 7 D—G, the differential expression results of the two core genes across clinical and pathological subtypes of MSA are not entirely consistent. Specifically, the expression levels of HPRGs are significantly decreased in both clinical phenotypes of MSA compared to the control group. However, this trend is not observed in MSA-SNP and subtypes, whereas decreased expression is observed in MSA-OPCA and SND-OPCA. Generally, MSA-P and MSA-C correspond pathologically to SND and OPCA, respectively. (Poewe, et al., 2022). However, a growing body of pathological research indicates that, in addition to MSA subtypes dominated by either SND or OPCA pathology, a considerable number of cases present as mixed-type MSA with comparable pathological severity of SND and OPCA (Jellinger, et al., 2005 ; Ozawa, et al., 2004). This suggests that pathological changes in MSA are not confined to a single brain region but become more widespread as the disease progresses(Ndayisaba, et al., 2025 ; Wan, et al., 2023). Moreover, results from the single-nucleus analysis (including dot plots shown in Fig. 9 C and expression changes along pseudotime presented in Fig. 11 B) consistently show widespread downregulation of HPRGs in oligodendrocytes of MSA patients. Consistent findings were also observed in our analysis of the GSE199724 dataset (Fig. 7 H). This indicates that the overall low expression of HPRGs in tissues is not solely due to the extensive loss of oligodendrocytes but may also result from suppressed expression within dysfunctional oligodendrocytes. These findings further support the potential of HPRGs as diagnostic markers. Oligodendrocytes are the earliest and most severely affected cell type in MSA, playing a central role in the onset and progression of the disease(Hsiao, et al., 2023; Poewe, et al., 2022). The accumulation of misfolded α-synuclein within oligodendrocytes underlies the formation of GCIs(Schweighauser, et al., 2020; Wan, et al., 2023). Oligodendrocytes subjected to prolonged high GCI burden exhibit structural and functional impairments, often accompanied by demyelination and subsequent neuronal loss(Ndayisaba, et al., 2025 ; Poewe, et al., 2022). In the early to middle stages of the disease, GCI density shows a strong positive correlation with the extent of neuronal loss and disease duration(Ozawa, et al., 2004). However, in advanced-stage MSA cases, this correlation reverses. Multiple studies have reported a decline in GCI density in severely atrophic brain regions, primarily due to the widespread death of host oligodendrocytes(Jellinger, et al., 2005 ; Tanaka, et al., 2025). Various mechanisms contribute to this cell death: when intracellular protein load exceeds a certain threshold, or when mitochondrial dysfunction compromises the cell’s capacity to sustain basic metabolic activities, oligodendrocytes may undergo apoptosis or necrosis, leading to the release of GCIs. These insoluble fibrils may remain in the extracellular matrix for a short period but are ultimately phagocytosed and removed by microglia, accompanied by severe neuroinflammation(Tanaka, et al., 2025). In MSA patients, oligodendrocyte precursor cells (OPCs) fail to differentiate into mature myelin-forming cells to repair the damage, a process inhibited by mechanisms such as the toxic effects of α-synuclein on OPCs(Hsiao, et al., 2023; May, et al., 2014). Consequently, in end-stage tissue, severe demyelination and neuronal loss may be observed, yet the visible GCI burden is lower than in less affected regions. In our cell deconvolution analysis of MSA cerebellar white matter transcriptomic data, we observed a reduction in oligodendrocytes (Fig. 6 A). Both ZDHHC20 and AK5 showed positive correlations with oligodendrocyte proportion (Fig. 6 B), and their downregulation may indirectly reflects the extensive loss of oligodendrocytes in the terminal stage of MSA. The findings of this study have been discussed in detail above and are of significant importance. However, this study also has several limitations. First, although the ROC curves of HPRGs in the training and validation sets demonstrated good diagnostic performance (AUC > 0.7), we did not construct a diagnostic model. If these markers can be combined with other influencing factors to build a diagnostic model in the future, it may lead to even better diagnostic outcomes. Second, the bulk RNA-seq data and single-nucleus data used in this study were derived from different tissue sources. The prefrontal cortex of advanced MSA-P patients may exhibit a certain degree of involvement, although the extent of atrophy is generally mild(Poewe, et al., 2022; Tanaka, et al., 2025). While we observed low expression of these biomarkers in oligodendrocytes in such tissues, the expression of HPRGs in severely affected tissues (such as cerebellar white matter) remains unclear. Additionally, all analyzed samples were obtained from end-stage MSA patients, and the expression patterns of these genes in early-stage patients require further investigation. Third, the current use of HPRGs as diagnostic markers is limited, as obtaining corresponding brain tissue from MSA patients for diagnosis is highly invasive. Therefore, it is necessary to identify other tissues—such as blood, cerebrospinal fluid, or even skin—that may reflect such expression patterns. Fourth, due to the scarcity of publicly available samples for MSA, all bulk RNA-seq data used in our study were derived from a single study. Fortunately, this study comprised three independent cohorts, with samples sourced from different brain banks worldwide, exhibiting significant batch effects. This provided a basis for selecting training and validation sets. Moreover, MSA patients are extremely rare, and obtaining postmortem brain tissue from them is particularly challenging. As a result, experimental validation was not conducted in this study. If conditions permit in the future, further experimental validation should be carried out to verify the reliability of these findings. Fifth, the differential expression results of HPRGs in oligodendrocytes showed a non-significant downward trend, which may be related to the small sample size of oligodendrocytes (n = 12). A small sample size may lead to discrepancies between the results and expectations. In addition, heterogeneity in factors such as study populations, sample processing time and methods, and various clinical variables (e.g., disease duration, age at death, comorbid conditions) may introduce bias. 5. Conclusions By applying multiple machine learning algorithms and logistic regression analysis, we identified ZDHHC20 and its highly co-expressed gene, AK5. We verified their significant downregulation in MSA and demonstrated their strong performance as potential diagnostic markers. Furthermore, we explored their associated biological functions and potential upstream regulators, and elucidated their distribution and expression across various cell types at the single-nucleus level. We also simulated alterations in their expression levels during the progression of MSA. The findings provide novel insights into the investigation of the pathogenesis of MSA and concurrently offer innovative alternatives for the diagnosis of MSA. Authorship Contributions Zhipeng Lu : Writing - original draft,Writing - review & editing, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization. Pu Fang : Writing - review & editing, Project administration, Supervision, Funding acquisition. Zhongqi Li : Writing - review & editing, Methodology, Formal analysis. Zhibiao Yin : Writing - review, Supervision, Methodology. Jialong Liu : Writing - review & editing. Abbreviations AD Alzheimer's disease AK Adenylate kinase ALS Amyotrophic lateral sclerosis AUC Area under the curve CBD Corticobasal degeneration ceRNA Competing endogenous RNA Cor Correlation coefficient DEG Differentially expressed gene DT Decision tree EMG Electromyography FDR False discovery rate GBM Gradient boosting machine GCI Glial cytoplasmic inclusion GEO Gene expression omnibus GLM Generalized linear model HPRG Hub palmitoylation-related gene HVG Highly Variable Gene KNN K-nearest neighbors LASSO Least absolute shrinkage and selection operator MR Mendelian randomization MSA Multiple system atrophy MSA-C MSA cerebellar type MSA-P MSA parkinsonian type NNET Neural network OPC Oligodendrocyte progenitor cell OPCA Olivopontocerebellar atrophy OR Odds ratio PAT Palmitoyltransferase PCA Principal component analysis PD Parkinson's disease PRG Palmitoylation-related gene PSP Progressive supranuclear palsy QC Quality control RF Random forest ROC Receiver operating characteristic SAOA Sporadic adult-onset ataxia of unknown etiology SND Striatonigral degeneration SNCA Synuclein, alpha SVM Support vector machine TFs Transcription factors αSyn α-synuclein Declarations Authorship Contributions Zhipeng Lu: Writing - original draft,Writing - review & editing, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization. Pu Fang: Writing - review & editing, Project administration, Supervision, Funding acquisition. Zhongqi Li: Writing - review & editing, Methodology, Formal analysis. Zhibiao Yin: Writing - review, Supervision, Methodology. Jialong Liu: Writing - review & editing. Acknowledgments We express our sincere gratitude to all the authors who actively participated in this research. In particular, we are indebted to the corresponding author for his invaluable assistance in facilitating the progress of this study. Simultaneously, we are profoundly thankful to the GEO and Synapse database and to all the contributors who uploaded the relevant transcriptomic data. Funding : The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. . Competing interests : The authors have no relevant financial or non-financial interests to disclose. Ethics approval: Not applicable. Consent to Participate: Not applicable. Consent for Publication: Not applicable. Competing interests: The authors declare no competing interests. References Ansoleaga B et al (2015) Deregulation of purine metabolism in Alzheimer's disease. Neurobiol Aging 36(1):68–80 B SP, Talwar P (2025) Influence of palmitoylation in axonal transport mechanisms in neurodegenerative diseases. Front Cell Neurosci 19:1613379 Buszka A et al (2023) S-Palmitoylation of Synaptic Proteins in Neuronal Plasticity in Normal and Pathological Brains. Cells ;12(3) Chen B et al (2018) Profiling Tumor Infiltrating Immune Cells with CIBERSORT. Methods Mol Biol 1711:243–259 Chen Y, Li Y, Wu L (2024) Protein S-palmitoylation modification: implications in tumor and tumor immune microenvironment. Front Immunol 15:1337478 Cho E, Park M (2016) Palmitoylation in Alzheimer's disease and other neurodegenerative diseases. Pharmacol Res 111:133–151 F SM et al (2024) Mechanisms and functions of protein S-acylation. Nat Rev Mol Cell Biol 25(6):488–509 Fan Z et al (2024) Modulators for palmitoylation of proteins and small molecules. Eur J Med Chem 271:116408 Fujisawa K (2023) Regulation of Adenine Nucleotide Metabolism by Adenylate Kinase Isozymes: Physiological Roles and Diseases. Int J Mol Sci ;24(6) Garcia-Esparcia P et al (2015) Purine metabolism gene deregulation in Parkinson's disease. Neuropathol Appl Neurobiol 41(7):926–940 Guo Q et al (2025) Multi-Omics Mendelian Randomization Identifies a DNA Methylation-ZDHHC20-Immune Axis Associated With Schizophrenia Risk. Brain Behav 15(8):e70722 Hao Y et al (2024) Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 42(2):293–304 Hsiao JT et al (2023) Role of Oligodendrocyte Lineage Cells in Multiple System Atrophy. Cells ;12(5) Huang B et al (2025) Palmitoylation-dependent regulation of GPX4 suppresses ferroptosis. Nat Commun 16(1):867 Jellinger KA, Seppi K, Wenning GK (2005) Grading of neuropathology in multiple system atrophy: proposal for a novel scale. Mov Disord 20(Suppl 12):S29–36 Jeong HK et al (2025) , . Micro-scale control of oligodendrocyte morphology and myelination by the intellectual disability-linked protein acyltransferase ZDHHC9. Elife ;13. Ko PJ, Dixon SJ (2018) Protein palmitoylation and cancer. EMBO Rep ;19(10) Krismer F et al (2024) Multiple system atrophy: advances in pathophysiology, diagnosis, and treatment. Lancet Neurol 23(12):1252–1266 Lai Y et al (2016) Down-regulation of adenylate kinase 5 in temporal lobe epilepsy patients and rat model. J Neurol Sci 366:20–26 Leisch DM (2024) a.E.D.a.K.H.a.A.W.a.F. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. Arxiv preprint arXiv 1(1):548 Li JH et al (2014) starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res 42(Database issue):D92–97 Lu B et al (2024) zDHHC20-driven S-palmitoylation of CD80 is required for its costimulatory function. Acta Pharmacol Sin 45(6):1214–1223 Lucas C et al (2016) Yeast Gup1(2) Proteins Are Homologues of the Hedgehog Morphogens Acyltransferases HHAT(L): Facts and Implications. J Dev Biol ;4(4) Mantle D, Hargreaves IP (2022) Mitochondrial Dysfunction and Neurodegenerative Disorders: Role of Nutritional Supplementation. Int J Mol Sci ;23(20) Mathys H et al (2019) Single-cell transcriptomic analysis of Alzheimer's disease. Nature 570(7761):332–337 May VE et al (2014) α-Synuclein impairs oligodendrocyte progenitor maturation in multiple system atrophy. Neurobiol Aging 35(10):2357–2368 McMichael TM et al (2017) The palmitoyltransferase ZDHHC20 enhances interferon-induced transmembrane protein 3 (IFITM3) palmitoylation and antiviral activity. J Biol Chem 292(52):21517–21526 Ndayisaba A, Halliday GM, Khurana V (2025) Multiple System Atrophy: Pathology, Pathogenesis, and Path Forward. Annu Rev Pathol 20(1):245–273 Nido GS et al (2025) Single-nucleus transcriptomics reveals disease- and pathology-specific signatures in α-synucleinopathies. Brain 148(5):1588–1603 Ozawa T et al (2004) The spectrum of pathological involvement of the striatonigral and olivopontocerebellar systems in multiple system atrophy: clinicopathological correlations. Brain 127(Pt 12):2657–2671 Piras IS et al (2020) Transcriptional profiling of multiple system atrophy cerebellar tissue highlights differences between the parkinsonian and cerebellar sub-types of the disease. Acta Neuropathol Commun 8(1):76 Poewe W et al (2022) Multiple system atrophy. Nat Rev Dis Primers 8(1):56 Qi X et al (2025) Integrated Analysis of Ferroptosis- and Cellular Senescence-Related Biomarkers in Atherosclerosis Based on Machine Learning and Single-Cell Sequencing Data. J Inflamm Res 18:9283–9305 Qiu X et al (2017) Single-cell mRNA quantification and differential analysis with Census. Nat Methods 14(3):309–315 Qiu X et al (2017) Reversed graph embedding resolves complex single-cell trajectories. Nat Methods 14(10):979–982 Rauluseviciute I et al (2024), JASPAR. : 20th anniversary of the open-access database of transcription factor binding profiles. Nucleic Acids Res 2024;52(D1):D174–d182. Schweighauser M et al (2020) Structures of α-synuclein filaments from multiple system atrophy. Nature 585(7825):464–469 Siddiqui MS, Shahi MH, Castresana JS (2024) The role of the adenylate kinase 5 gene in various diseases and cancer. J Clin Transl Sci 8(1):e96 Tanaka MT et al (2025) Pathological and Molecular Insights into the Early Stage of Multiple System Atrophy. Cells ;14(24) Trapnell C et al (2014) The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 32(4):381–386 Van Rompay AR, Johansson M, Karlsson A (1999) Identification of a novel human adenylate kinase. cDNA cloning, expression analysis, chromosome localization and characterization of the recombinant protein. Eur J Biochem 261(2):509–517 Wan L et al (2023) Multidimensional biomarkers for multiple system atrophy: an update and future directions. Transl Neurodegener 12(1):38 Wang J, Devlin B, Roeder K (2020) Using multiple measurements of tissue to estimate subject- and cell-type-specific gene expression. Bioinformatics 36(3):782–788 Wenning GK et al (2004) Multiple system atrophy. Lancet Neurol 3(2):93–103 Wenning GK et al (2022) The Movement Disorder Society Criteria for the Diagnosis of Multiple System Atrophy. Mov Disord 37(6):1131–1148 Yu G et al (2012) clusterProfiler: an R package for comparing biological themes among gene clusters. Omics 16(5):284–287 Zhang H et al (2024) ZDHHC20-mediated S-palmitoylation of YTHDF3 stabilizes MYC mRNA to promote pancreatic cancer progression. Nat Commun 15(1):4642 Zhou KR, Liu HJ, Zheng S, Liu WJ SR, An Encyclopedic Regulatory and Functional Atlas of RNA Interactomes Additional Declarations No competing interests reported. Supplementary Files Supplementarymaterials1.xlsx Supplementarymaterials2.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 21 Feb, 2026 Reviews received at journal 15 Feb, 2026 Reviewers agreed at journal 10 Feb, 2026 Reviewers agreed at journal 10 Feb, 2026 Reviewers invited by journal 10 Feb, 2026 Editor assigned by journal 29 Jan, 2026 Submission checks completed at journal 29 Jan, 2026 First submitted to journal 27 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8711800","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":590439104,"identity":"19bd468e-0d4b-4ece-84b5-080ecc62806e","order_by":0,"name":"Zhipeng Lu","email":"","orcid":"","institution":"First Affiliated Hospital of Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Zhipeng","middleName":"","lastName":"Lu","suffix":""},{"id":590439107,"identity":"00a0323c-8724-41cd-98af-aaae5deb46ab","order_by":1,"name":"Zhongqi Li","email":"","orcid":"","institution":"First Affiliated Hospital of Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Zhongqi","middleName":"","lastName":"Li","suffix":""},{"id":590439111,"identity":"b7327c13-8f07-4e1c-bbcc-c8388d4365d8","order_by":2,"name":"Zhibiao Yin","email":"","orcid":"","institution":"First Affiliated Hospital of Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Zhibiao","middleName":"","lastName":"Yin","suffix":""},{"id":590439113,"identity":"e7adf80d-19b5-49be-8628-1e841e44a5e5","order_by":3,"name":"Jialong Liu","email":"","orcid":"","institution":"Second Affiliated Hospital of Nanchang University","correspondingAuthor":false,"prefix":"","firstName":"Jialong","middleName":"","lastName":"Liu","suffix":""},{"id":590439115,"identity":"068deb16-c58a-448e-8dbf-b7c20c110e90","order_by":4,"name":"Pu Fang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3ElEQVRIiWNgGAWjYBAC9gYwJSFnf7z5wIEPP4jQwnMATFkYM5w5lnhwZg/xWioSG27kGB/mYCNGC/vZw68L2yQSGxtyPhxm4GGQ5xc7QEALT16a9cw2CeNmhrMbDhdYMBjOnJ2AX4s9Q46ZMW+bhGwbY++GwzN4GBIMbhPQwsP/BqyFsYeZ58FhHjZitEjkGD8GalGcwcbDQKyWN2bMM85JGBvwsBkAA1mCsF94+HOMPxeU1ckZyD9+/OHDDxt5fmkCWoCATRqJI0FQOQgwfyZK2SgYBaNgFIxcAABWxEM/Ex7e+gAAAABJRU5ErkJggg==","orcid":"","institution":"First Affiliated Hospital of Nanchang University","correspondingAuthor":true,"prefix":"","firstName":"Pu","middleName":"","lastName":"Fang","suffix":""}],"badges":[],"createdAt":"2026-01-27 14:40:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8711800/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8711800/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102747815,"identity":"52e65bb0-1381-410c-bec6-5a358f1ccf41","added_by":"auto","created_at":"2026-02-16 09:05:25","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":180455,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the study design.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/2df589adabbb25b5d5416958.png"},{"id":102602028,"identity":"1ec7941c-16f7-4f15-a0b8-45e25f1cba45","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":626611,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification and functional annotation of DEGs in GSE199715. \u003cstrong\u003e(A)\u003c/strong\u003e PCA plot between MSA (red) and HC (blue) samples. Each point represents a sample, and the ellipse indicates the 95% confidence interval. \u003cstrong\u003e(B)\u003c/strong\u003e Volcano plot of DEGs between the MSA and HC groups. Red and blue dots indicate up- and downregulated genes, respectively. Grey dots represent relatively stable genes. The top 10 most significantly up- and downregulated genes (by |log2FoldChange|) are labeled. \u003cstrong\u003e(C)\u003c/strong\u003eHeatmap illustrating the expression patterns of DEGs in MSA and HC samples. Red indicates upregulation; blue indicates downregulation. \u003cstrong\u003e(D)\u003c/strong\u003e GO enrichment analysis of DEGs, displaying the top 10 terms for BP, CC, and MF categories. \u003cstrong\u003e(E)\u003c/strong\u003eKEGG pathway enrichment analysis of DEGs. PCA: Principal Component Analysis; MSA: Multiple system atrophy; HC: Healthy control; DEGs: Differentially expressed genes.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/14a6145c47a744025e463573.png"},{"id":102602030,"identity":"0a65de94-a232-4bee-b482-f469acb605a1","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":377367,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of ZDHHC20 using multiple machine learning algorithms. \u003cstrong\u003e(A)\u003c/strong\u003eBox plots comparing the absolute residuals of eight machine learning models. \u003cstrong\u003e(B)\u003c/strong\u003eROC curves evaluating the performance of the eight algorithms. AUC values for each model are displayed in the legend. \u003cstrong\u003e(C)\u003c/strong\u003e Top 10 genes ranked by feature importance from all eight machine learning algorithms. \u003cstrong\u003e(D–I)\u003c/strong\u003e The top 10 genes based on feature importance and corresponding ROC curves for the GBM, LASSO, and SVM models are shown separately. \u003cstrong\u003e(J)\u003c/strong\u003e Venn diagram illustrating the intersection of the top 10 genes identified by the GBM, LASSO, and SVM models with the previously identified DEGs. ROC: Receiver operating characteristic; AUC: Area under the curve; GBM: Gradient boosting machine; LASSO: Least absolute shrinkage and selection operator; SVM: Support vector machine; DEGs: Differentially expressed genes.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/44ae0e59edcae4be63ff9ea3.png"},{"id":102602036,"identity":"514941af-b3a9-4520-9833-61b2136f2dc1","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1151911,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification and validation of AK5 as a key ZDHHC20-correlated gene.\u003cstrong\u003e (A–B)\u003c/strong\u003e LASSO regression analysis of key genes: 14 genes were screened by selecting the -log(λ) value corresponding to the minimum partial likelihood deviance.\u003cstrong\u003e (C–D)\u003c/strong\u003eSVM feature selection, showing model accuracy\u003cstrong\u003e (C)\u003c/strong\u003e and error rates\u003cstrong\u003e (D)\u003c/strong\u003eused to select the optimal 11 genes. \u003cstrong\u003e(E) \u003c/strong\u003eForest plot of univariate logistic regression for the 10 common genes from LASSO and SVM. \u003cstrong\u003e(F)\u003c/strong\u003eForest plot of multivariate logistic regression, identifying AK5 as the sole significant hub gene. \u003cstrong\u003e(G)\u003c/strong\u003e Scatter plot demonstrating the strong positive correlation between ZDHHC20 and AK5. The pink line and gray shadow around it represent the fitted linear regression with its 95% confidence interval. \u003cstrong\u003e(H)\u003c/strong\u003eViolin plot comparing HPRG expression in MSA and HC groups. \u003cstrong\u003e(I)\u003c/strong\u003e ROC curves assessing the diagnostic efficacy of ZDHHC20 and AK5. *** indicates P \u0026lt; 0.001. λ: penalty coefficient; HPRGs: Hub palmitoylation-related genes, referring to ZDHHC20 and AK5. R: Correlation coefficient.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/29e98f8bbb3c5e45936b2cd5.png"},{"id":102602032,"identity":"02508888-e218-4914-adad-7631745937c6","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1510122,"visible":true,"origin":"","legend":"\u003cp\u003eGene Set Enrichment Analysis (GSEA) of HPRGs. \u003cstrong\u003e(A–B)\u003c/strong\u003e GO and KEGG enrichment analysis for ZDHHC20. \u003cstrong\u003e(C–D)\u003c/strong\u003e GO and KEGG enrichment analysis for AK5.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/4e38d44fa364b5fb319550e5.png"},{"id":102747321,"identity":"b618cadc-f42c-42a7-96e5-6b2030f7f0f7","added_by":"auto","created_at":"2026-02-16 09:04:30","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":241346,"visible":true,"origin":"","legend":"\u003cp\u003eCell deconvolution analysis. \u003cstrong\u003e(A)\u003c/strong\u003e Differences in the proportions of neuronal and glial cell populations between MSA and HC samples. \u003cstrong\u003e(B)\u003c/strong\u003e Lollipop plot depicting correlations between HPRGs and both oligodendrocytes and inhibitory neurons, with correlation strength indicated by the size and color intensity of the lollipop symbols. \u003cstrong\u003e(C–D)\u003c/strong\u003e Scatter plots showing the associations between HPRGs and oligodendrocytes as well as inhibitory neurons. *: P \u0026lt; 0.05; **: P \u0026lt; 0.01; ***: P \u0026lt; 0.001; ns: P \u0026gt; 0.05.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/6cad903a1d2329eb75b767e5.png"},{"id":102602029,"identity":"cf610e35-bf45-4e9e-a065-a79f8d01af5c","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":259110,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of HPRGs and expression analysis in different subtypes and oligodendrocytes. \u003cstrong\u003e(A)\u003c/strong\u003eViolin plots showing HPRG expression in the validation cohort (GSE199258). \u003cstrong\u003e(B)\u003c/strong\u003eROC curves for ZDHHC20 and AK5 in the validation cohort, with AUC values displayed. \u003cstrong\u003e(C)\u003c/strong\u003e Scatter plot confirming the strong positive correlation between ZDHHC20 and AK5 in GSE199258. The pink line and the gray shadow surrounding it denote the fitted linear regression along with its 95% confidence interval.\u003cstrong\u003e (D–E)\u003c/strong\u003e HPRG expression in clinical subtypes of MSA.\u003cstrong\u003e(F–G)\u003c/strong\u003e HPRG expression in pathological subtypes of MSA. \u003cstrong\u003e(H)\u003c/strong\u003e HPRG expression in oligodendrocytes from MSA and HC samples. Significance levels are denoted as follows: *, P \u0026lt; 0.05; **, P \u0026lt; 0.01; ***, P \u0026lt; 0.001; ns, P \u0026gt; 0.05.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/450a004b4a80a45846b188e2.png"},{"id":103503974,"identity":"1f6ab553-a014-47ee-a4cf-297789b443bd","added_by":"auto","created_at":"2026-02-26 13:06:19","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":834288,"visible":true,"origin":"","legend":"\u003cp\u003eUpstream regulatory network analysis of HPRGs. \u003cstrong\u003e(A)\u003c/strong\u003e The ceRNA network for HPRGs. The size of each lncRNA label and the color of the icon correspond to the number of miRNAs with which they interact (denoted as “degree”) ; a larger label and a color closer to yellow indicate a higher degree. \u003cstrong\u003e(B)\u003c/strong\u003e The TF-gene interaction network displaying predicted upstream transcription factors for HPRGs.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/a201c1e9cd83896ca6e8e2f2.png"},{"id":102602034,"identity":"c686ec64-7634-4afc-94b6-39995822d231","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":376818,"visible":true,"origin":"","legend":"\u003cp\u003eCell annotation and expression of key genes. \u003cstrong\u003e(A)\u003c/strong\u003e Eight major cell populations were identified following annotation of 50 subpopulations. \u003cstrong\u003e(B) \u003c/strong\u003eProportions of major cell types in MSA and HC samples. \u003cstrong\u003e(C–D)\u003c/strong\u003e Expression patterns of HPRGs across different cell types in MSA and HC. In panel C, dot size reflects the percentage of cells expressing each gene within a given cell type, and color and intensity represent the average expression level.\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/25c8b96218b9a9dd20a12688.png"},{"id":102602039,"identity":"f18cfbf9-2d3c-463f-9cef-93d6be898d45","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":335995,"visible":true,"origin":"","legend":"\u003cp\u003eDimensionality reduction and clustering of oligodendrocyte subpopulations. \u003cstrong\u003e(A)\u003c/strong\u003e UMAP plot displaying the segregation of oligodendrocytes into 19 distinct clusters. \u003cstrong\u003e(B)\u003c/strong\u003eProportions of each cluster in MSA and HC samples. \u003cstrong\u003e(C) \u003c/strong\u003eDistribution of clusters between disease and control groups. \u003cstrong\u003e(D) \u003c/strong\u003eExpression of ZDHHC20 and AK5 across all clusters, where dot size reflects the percentage of cells expressing the gene within each cluster, and red color indicates higher mean expression levels.\u003c/p\u003e","description":"","filename":"image10.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/1724e38eea70c492e0d8029f.png"},{"id":102602040,"identity":"01ee185e-0eba-432e-ac61-9a458aaa940d","added_by":"auto","created_at":"2026-02-13 13:13:21","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":1435669,"visible":true,"origin":"","legend":"\u003cp\u003ePseudotime analysis. \u003cstrong\u003e(A)\u003c/strong\u003e Pseudotemporal distribution of oligodendrocytes, with lighter colors indicating later stages of the developmental trajectory. \u003cstrong\u003e(B)\u003c/strong\u003eDynamic changes in marker gene expression patterns along the inferred developmental trajectory. \u003cstrong\u003e(C–D)\u003c/strong\u003e Density distribution of MSA and HC samples and each oligodendrocyte cluster along the pseudotime axis.\u003c/p\u003e","description":"","filename":"image11.png","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/05759702dfb06ec7f9c84671.png"},{"id":103508805,"identity":"b83854eb-020d-467d-8be2-2931b0efe7e7","added_by":"auto","created_at":"2026-02-26 13:54:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8077053,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/ce1edd90-2116-4d1c-9512-d0e87017ccfd.pdf"},{"id":102747805,"identity":"234ea9b8-9a05-4997-8aff-c1aad18adb30","added_by":"auto","created_at":"2026-02-16 09:05:25","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":13566,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterials1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/ef2b0fda26bd1ccd8d460900.xlsx"},{"id":102748093,"identity":"affc7049-a214-4c7b-99d5-442580d7276f","added_by":"auto","created_at":"2026-02-16 09:05:55","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":8076326,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterials2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8711800/v1/10648cce9d6ff3e2f2d3caf6.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Transcriptomic Analysis and Multiple Machine Learning Approaches Identify ZDHHC20 and Its Highly Correlated Gene AK5 as Diagnostic Markers in Multiple System Atrophy","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eMultiple System Atrophy (MSA) is a rare neurodegenerative disease that has garnered significant medical attention due to its sporadic nature, rapid progression, and fatal outcome (Krismer, et al., 2024; Poewe, et al., 2022). The clinical presentation primarily manifests as any combination of cerebellar syndrome, parkinsonism, and autonomic dysfunction (Wenning, et al., 2022). As a synucleinopathy, its pathogenesis is not fully understood. The prevailing view posits that the abnormal accumulation of misfolded α-synuclein (αSyn) as glial cytoplasmic inclusions (GCIs) within oligodendrocytes in affected brain regions\u0026mdash;including the striatonigral and olivopontocerebellar systems\u0026mdash;plays a critical role in the pathogenic process. This is the specific pathological hallmark of MSA (Wan, et al., 2023). Due to the high heterogeneity of its clinical manifestations and the symptomatic overlap with other disorders\u0026mdash;including Parkinson's disease (PD), progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and sporadic adult-onset ataxia of unknown etiology (SAOA)\u0026mdash;MSA is frequently misdiagnosed (Ndayisaba, et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Poewe, et al., 2022; Wan, et al., 2023). Currently, the diagnosis of MSA is predominantly based on clinical history and neurological examination. examinations, including autonomic function tests, electromyography (EMG), and neuroimaging, can be employed to support the diagnosis or rule out other conditions (Wenning, et al., 2004). Potential MSA biomarkers from various sample sources (e.g., body fluids, tissues, gut microbiota, and imaging) have also been developed, but their clinical application remains limited (Wan, et al., 2023). For patients diagnosed with MSA, there are currently no effective treatments to slow disease progression; thus, MSA patients often have a poor prognosis (Piras, et al., 2020; Schweighauser, et al., 2020). Against this backdrop, exploring the pathogenesis of MSA and identifying effective diagnostic and therapeutic strategies are urgently needed.\u003c/p\u003e \u003cp\u003ePalmitoylation is a common post-translational modification. Its mechanism involves the covalent attachment of a saturated fatty acid, such as palmitic acid, to amino acid residues of a substrate protein (Huang, et al., 2025; Zhang, et al., 2024). This modification subsequently influences various aspects of the substrate protein, including structure, function, and membrane localization (Ko and Dixon, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Palmitoylation is widely present in various cellular metabolic activities, and i is also heavily involved in important processes such as neuronal development and maturation, the establishment of synaptic plasticity, and neural signal transduction (F, et al., 2024). Nearly half of all synaptic proteins and almost the entire myelin proteome are substrates for S-palmitoylation, which is the most common form of palmitoylation (F, et al., 2024; Jeong, et al., 2025). Some palmitoyltransferases, such as ZDHHC5 and ZDHHC9, are essential for oligodendrocyte maturation and myelin formation (F, et al., 2024). Owing to its pivotal role in synaptic plasticity, palmitoylation also assumes a crucial role in advanced cognitive functions (Buszka, et al., 2023). Recent studies have shown that the pathogenesis of multiple neurodegenerative diseases, including Alzheimer's disease (AD), Parkinson's disease (PD), and Amyotrophic Lateral Sclerosis (ALS), is closely associated with palmitoylation (Cho and Park, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). However, no studies have yet linked palmitoylation to multiple system atrophy.\u003c/p\u003e \u003cp\u003eIn recent years, bioinformatics and machine learning algorithms have developed rapidly. Machine learning algorithms can analyze complex and diverse data to identify optimal biomarkers and have therefore been widely applied in the medical field (Qi, et al., 2025). Bioinformatic analysis is a common analytical method for studying phenotypes and diseases; however, bioinformatics analyses regarding MSA are relatively scarce. From a bioinformatics perspective, we investigated the role of palmitoylation in MSA, identified palmitoylation-related biomarkers through various machine learning and logistic regression analyses, and discussed their functions and upstream regulatory mechanisms. This facilitates an understanding of MSA pathogenesis from a new perspective and provides a direction for the diagnosis and even personalized treatment of MSA.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Data acquisition\u003c/h2\u003e \u003cp\u003eThe datasets GSE199715 (n\u0026thinsp;=\u0026thinsp;94) and GSE199258 (n\u0026thinsp;=\u0026thinsp;38) were derived from bulk RNA sequencing data of cerebellar white matter from postmortem MSA patients in different brain banks. GSE199724 (n\u0026thinsp;=\u0026thinsp;12) consists of oligodendrocyte samples extracted from the cerebellar white matter tissue of postmortem MSA patients for bulk RNA sequencing (Piras, et al., 2020). We selected dataset GSE199715 as the discovery cohort to identify hub genes, used GSE199258 as the validation set, and employed GSE199724 to explore the expression of these genes in oligodendrocytes. All datasets above were obtained from the Gene Expression Omnibus (GEO) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.ncbi.nlm.nih.gov/geo\u003c/span\u003e\u003cspan address=\"http://www.ncbi.nlm.nih.gov/geo\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) .\u003c/p\u003e \u003cp\u003eThe snRNA-seq dataset (Syn52662231) utilized in this study was obtained from the Synapse database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.synapse.org/\u003c/span\u003e\u003cspan address=\"https://www.synapse.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). It consists of snRNA-seq data derived from postmortem prefrontal cortex tissues of individuals with Parkinson's disease (n\u0026thinsp;=\u0026thinsp;20), Parkinsonian-type multiple system atrophy (MSA-P, n\u0026thinsp;=\u0026thinsp;6), and healthy controls (HC, n\u0026thinsp;=\u0026thinsp;13). Data from the MSA-P and HC groups were extracted for analysis(Nido, et al., 2025).\u003c/p\u003e \u003cp\u003eA list of palmitoylation-regulated genes (PRGs), including 27 palmitoyltransferase genes and 9 depalmitoylation-related enzyme genes, was obtained from previous literature (Chen, et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Fan, et al., 2024; Lucas, et al., 2016). The 36 PRGs are shown in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Screening and analysis of DEGs\u003c/h2\u003e \u003cp\u003eIn this study, the \"DESeq2\" package in R was used to perform differential expression analysis on genes from MSA patients and healthy controls (HC) in GSE199715. Genes with an average expression level (counts)\u0026thinsp;\u0026le;\u0026thinsp;10 were filtered out. The criteria of |log2FoldChange| \u0026gt; 1 and adjusted P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were applied to screen for differentially expressed genes (DEGs). A heatmap of the DEGs was generated using the \"pheatmap\" R package, and a volcano plot was drawn using the \"ggpubr\" R package. We carried out Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of the DEGs using the \"clusterProfiler\" R package (Yu, et al., 2012). Enrichment terms meeting the criteria of P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and q-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered significant. The resulting biological functions and pathways were visualized using the \"ggplot2\" R package.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Identification of hub genes using machine learning models\u003c/h2\u003e \u003cp\u003eEight machine learning models were used to screen for palmitoylation-related core genes: Gradient Boosting Machine (GBM), Generalized Linear Model (GLM), Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbors (KNN), Neural Network (NNET), and Decision Tree (DT). We evaluated model performance using multiple metrics, including residuals, precision, recall, and the area under the curve (AUC). These models were used to assess the importance of PRGs, and the top 10 genes were ranked based on the feature importance scores of each model. Appropriate models were selected by integrating these performance metrics. The genes identified from the selected models were intersected with the DEGs using a Venn diagram to identify the hub genes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Screening for ZDHHC20-correlated genes\u003c/h2\u003e \u003cp\u003eMSA samples were divided into high- and low-expression groups based on the mean expression level of ZDHHC20 in the GSE199715 dataset. The \"DESeq2\" R package was used to perform differential analysis between these two groups. Genes meeting the criteria of |log2FoldChange| \u0026gt; 2 and adjusted P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were defined as DEGs. Concurrently, we conducted a Spearman correlation analysis between ZDHHC20 and all genes in the training dataset. Genes with a correlation coefficient (Cor)\u0026thinsp;\u0026gt;\u0026thinsp;0.8 and P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered highly correlated. The intersection of these highly correlated genes and the ZDHHC20-related DEGs was determined .\u003c/p\u003e \u003cp\u003eNext, we further screened these intersected genes using LASSO and SVM algorithms. We employed the \"glmnet\" package to conduct the LASSO algorithm and identify the penalty coefficient (λ) associated with the minimum partial likelihood deviance. The \"e1071\" R package was used to implement the SVM with 10-fold cross-validation to prevent overfitting (Leisch, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Genes identified by the intersection of LASSO and SVM results were advanced to univariate and multivariate logistic regression analyses. In the univariate analysis (using the \"survival\" R package), genes with a P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were included in the multivariate analysis to determine the final key genes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Gene set enrichment analysis (GSEA)\u003c/h2\u003e \u003cp\u003eTo explore the potential biological mechanisms and pathways of the hub palmitoylation-related genes (HPRGs), we conducted GSEA using the \"clusterProfiler\" R package. The \"c5.go.v2025.1.Hs.symbols.gmt\" (GO) and \"c2.cp.kegg_medicus.v2025.1.Hs.symbols.gmt\" (KEGG) gene sets were downloaded from the MsigDB database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.gsea-msigdb.org/gsea/msigdb\u003c/span\u003e\u003cspan address=\"http://www.gsea-msigdb.org/gsea/msigdb\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). A threshold of False Discovery Rate (FDR)\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6. Cellular deconvolution\u003c/h2\u003e \u003cp\u003eTo investigate the cellular composition of cerebellar tissue, we employed cell type deconvolution to quantify the proportions of CNS (central nervous system)-resident cells and infiltrating immune cells between the MSA and HC groups by utilizing the \"CIBERSORT\" package(Chen, et al., 2018). A signature matrix file encompassing six cell types ( oligodendrocytes, excitatory neurons, inhibitory neurons, astrocytes, endothelial cells, and microglia) in the CNS was used in our study(Wang, et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The results were visualized through the application of the \"ggplot2\" package. Correlations between genes and immune cells were computed using Spearman's rank correlation test and presented as heatmaps via the \"pheatmap\" package.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7. HPRG expression visualization, ROC analysis and correlation scatter plots\u003c/h2\u003e \u003cp\u003eThe \"ggplot2\" R package was used to visualize the differential expression of HPRGs. The \"pROC\" package was employed to plot receiver operating characteristic (ROC) curves, and the area under the curve (AUC) was calculated to evaluate diagnostic value. The \"ggplot2\" and \"ggpubr\" packages were used to generate correlation scatter plots.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.8. ceRNA and gene-TF regulatory network\u003c/h2\u003e \u003cp\u003eWe utilized TargetScan (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.targetscan.org/\u003c/span\u003e\u003cspan address=\"http://www.targetscan.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), miRTarBase (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://mirtarbase.cuhk.edu.cn/\u003c/span\u003e\u003cspan address=\"https://mirtarbase.cuhk.edu.cn/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), and miRDB (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://mirdb.org/\u003c/span\u003e\u003cspan address=\"http://mirdb.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) databases to predict the target microRNAs (miRNAs) of HPRGs. The ENCORI database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://rnasysu.com/encori/\u003c/span\u003e\u003cspan address=\"http://rnasysu.com/encori/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was employed to predict target long non-coding RNAs (lncRNAs) (Li, et al., 2014; Zhou KR).The NetworkAnalyst online tool (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.networkanalyst.ca/\u003c/span\u003e\u003cspan address=\"http://www.networkanalyst.ca/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) was used to analyze interactions between hub genes and transcription factors (TFs), with TF targets derived from the JASPAR database (Rauluseviciute, et al., 2024). Cytoscape software was used to build the competing endogenous RNA ceRNA and TF-gene interaction network.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.9. snRNA-seq analysis\u003c/h2\u003e \u003cp\u003eSingle-nucleus data from 6 MSA-P and 13 HC samples (Syn52662231) underwent processing using the \u0026ldquo;Seurat\u0026rdquo; package(Hao, et al., 2024). Quality control (QC) filtered cells based on following criteria: 500\u0026thinsp;\u0026lt;\u0026thinsp;nFeature_RNA\u0026thinsp;\u0026lt;\u0026thinsp;6000 and percent.mt\u0026thinsp;\u0026lt;\u0026thinsp;5. Data underwent normalization (\u0026ldquo;NormalizeData\u0026rdquo; function in Seurat) and subsequent identification of Highly Variable Genes (HVGs) via \u0026ldquo;FindVariableFeatures\u0026rdquo;. Scaling for downstream dimensionality reduction was performed using \u0026ldquo;ScaleData\u0026rdquo;. The optimal number of principal components (PCs) was determined using an elbow plot. Dimensions were selected based on two primary criteria: the cumulative explained variance exceeded 90% while the single PC contribution remained below 5%, and the variance difference between adjacent PCs fell below 0.1%. Clustering resolution was established using a clustree plot, selecting the resolution that exhibited the most stable downward clustering change. Cell clusters were then identified using Seurat's \u0026ldquo;FindClusters\u0026rdquo; function. Cell type annotation was performed using previously established cell markers(Mathys, et al., 2019). Nonlinear dimensionality reduction was then conducted using \u0026ldquo;RunUMAP\u0026rdquo;, with results visualized via the \u0026ldquo;DimPlot\u0026rdquo; function. Gene expression across cell types was displayed using \u0026ldquo;VlnPlot\u0026rdquo; and \u0026ldquo;DotPlot\u0026rdquo;, and the distribution of target genes across cells was visualized with \u0026ldquo;FeaturePlot\u0026rdquo; The selected core cell types subsequently underwent dimensionality reduction and clustering utilizing this identical procedure.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e2.10. Pseudotime analysis\u003c/h2\u003e \u003cp\u003eTo characterize the temporal trajectory and biomarker expression dynamics of core cell types, we conducted pseudotime trajectory analysis. Using the Monocle2 package(Qiu, et al., 2017; Qiu, et al., 2017; Trapnell, et al., 2014), a CellDataSet object was generated, with size factors and dispersion calculated via the \u0026ldquo;estimateSizeFactors\u0026rdquo; and \u0026ldquo;estimateDispersions\u0026rdquo; functions. Transcripts were filtered to retain those with average expression above 0.1 and dispersion exceeding model-fitted values. Subsequently, the DDRTree algorithm was conducted to reduce dimensionality, and cells were ordered along the pseudotime trajectory using these filtered genes. We visualized the single nucleus trajectory using the \u0026ldquo;plot_cell_trajectory\u0026rdquo; function. The dynamic changes in biomarker expression along pseudotime were also visualized. Ridge plots further illustrated cell density distributions across the pseudotime axis for both disease and control groups, as well as specific cell subpopulations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e2.11. Statistical analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed using R software (version 4.4.3). All parameter analyses were conducted using two-tailed tests. The Wilcoxon rank-sum test was employed to compare differences between the two groups. Spearman's rank correlation test was used for all correlation analyses. The Wald test was used to evaluate the significance of the logistic regression analysis. A P-value or FDR less than 0.05 was considered statistically significant.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Identification and functional enrichment of DEGs between MSA and HC\u003c/h2\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the workflow of the study. In the GSE199715 dataset, a principal component analysis (PCA) plot demonstrated a clear separation between the MSA and HC groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). Differential expression analysis identified 93 DEGs, including 81 upregulated and 12 downregulated genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). The GO enrichment analysis of these DEGs revealed enrichment in Biological Processes (BP) related to protein conformation and cell junctions; Cellular Components (CC) associated with cellular structure; and Molecular Functions (MF) involved in protease activity regulation (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD). KEGG enrichment analysis yielded only two significant pathways, suggesting low enrichment of these DEGs in gene sets related to KEGG pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Discovery of ZDHHC20 via machine learning\u003c/h2\u003e \u003cp\u003eEight machine learning algorithms were applied to rank the importance of 36 PRGs. Based on a comprehensive evaluation of model performance (AUC, residuals, precision, F1-score, and accuracy), GBM, LASSO, and SVM were selected as the optimal models (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, B, Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). GBM demonstrated the highest AUC, precision, and accuracy (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, B, E, Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). LASSO showed the second-highest AUC along with the smallest residual value (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, B, G, Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e), and SVM yielded the best recall value (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA, B, I, Table \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e). Each model ranked genes by feature importance (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). The top 10 genes and the ROC from each of the three selected models are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD\u0026ndash;I. The intersection of the 93 DEGs and the top 10 genes from these three models yielded a hub gene: ZDHHC20 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eJ).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Identification of AK5 as the key ZDHHC20-correlated gene\u003c/h2\u003e \u003cp\u003eTo find genes related to ZDHHC20, we stratified MSA samples into ZDHHC20-high and ZDHHC20-low groups based on the mean expression levels. Differential analysis between these groups identified 142 upregulated and 1,222 downregulated genes (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eA, S1B). Concurrently, correlation analysis identified 197 genes highly correlated with ZDHHC20, with the threshold of correlation coefficient (Cor)\u0026thinsp;\u0026gt;\u0026thinsp;0.8 and P\u0026thinsp;\u0026lt;\u0026thinsp;0.05. The intersection of these two gene sets (ZDHHC20-related DEGs and highly correlated genes) yielded 29 key genes (Table S3), excluding ZDHHC20 itself (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eC). Functional enrichment analysis showed that these 29 genes were enriched in terms related to vesicle transport, such as Golgi apparatus subcompartment, focal adhesion, and exocytic vesicle (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003eTo further refine these 29 genes, we applied LASSO and SVM algorithms. The LASSO analysis results showed the penalty parameter associated with the minimum partial likelihood deviance (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA, \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). Ultimately, 14 genes were successfully selected. To achieve more stable and reliable results and prevent overfitting, the SVM algorithm employed 10-fold cross-validation. As depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD, when the top 11 genes are selected, the accuracy reaches its optimal value and the error is minimized. The intersection of the LASSO and SVM results yielded 10 genes for logistic regression (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE). Univariate analysis pinpointed five genes (SLCO1A2, PDGFRA, TF, CDKN1C, and AK5) that were associated with MSA (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Intriguingly, all of these genes exhibited an odds ratio (OR)\u0026thinsp;\u0026lt;\u0026thinsp;1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eE). In the subsequent multivariate analysis, only AK5 remained significantly associated with MSA status (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05), suggesting it may be a protective factor (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eF). A scatter plot confirmed a strong positive correlation between ZDHHC20 and AK5 (R\u0026thinsp;=\u0026thinsp;0.88, P\u0026thinsp;\u0026lt;\u0026thinsp;2.2e-16) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eG). Violin plots showed that both ZDHHC20 and AK5 were significantly downregulated in the MSA group (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH), consistent with the heatmap in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC. ROC curve analysis demonstrated good diagnostic performance for both ZDHHC20 (AUC\u0026thinsp;=\u0026thinsp;0.729) and AK5 (AUC\u0026thinsp;=\u0026thinsp;0.718) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eI). We finally defined ZDHHC20 and AK5 as hub palmitoylation-related genes (HPRGs) for subsequent analysis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e3.4. GSEA reveals similar enrichment profiles for ZDHHC20 and AK5\u003c/h2\u003e \u003cp\u003eSingle-gene GSEA was performed to explore the functions of HPRGs. In the GO analysis, pathways involving establishment of mitochondrion localization microtubule mediated, NADH dehydrogenase activity, protein targeting to lysosome, oxidoreductase activity acting on NADPH quinone or similar compound as acceptor, and oxidoreduction driven active transmembrane transporter activity were enriched in the ZDHHC20 high-expression group, while keratin filament, olfactory receptor activity, sensory perception of smell, sensory perception of chemical stimulus, and detection of chemical stimulus were enriched in the low-expression group (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA). AK5 showed a similar enrichment profile: establishment of mitochondrion localization microtubule mediated, oxidoreductase activity acting on NADPH quinone or similar compound as acceptor, regulation of mitochondrial fission, mitochondrial fission, and NADPH dehydrogenase activity were activated in the high-expression group, whereas keratin filament, olfactory receptor activity, sensory perception of smell, sensory perception of chemical stimulus and CUL2 ring ubiquitin ligase complex were suppressed (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC). The KEGG pathways significantly enriched for ZDHHC20 and AK5 were almost identical, including dynein recruitment to the kinetochore, mitochondrial complex UCP1 in thermogenesis, electron transfer in complex I, mutation caused aberrant ABETA to electron transfer in complex I, mutation caused aberrant SNCA to electron transfer in complex I, mutation caused aberrant TDP43 to electron transfer in complex I, and mutation inactivated PINK1 to electron transfer in complex I (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD). The \u0026ldquo;translation initiation\u0026rdquo; pathway was enriched only for ZDHHC20 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB). These terms are predominantly associated with energy metabolism, suggesting that the downregulation of HPRG might be linked to dysregulated intracellular energy metabolism.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e3.5. Decreased oligodendrocyte abundance in MSA and its positive correlation with HPRG expression\u003c/h2\u003e \u003cp\u003eIn this study, the \"CIBERSORT\" package was used to assess the cellular composition between MSA and control groups. Figure\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA depicts a box plot showing differences in the proportions of six cell types between MSA samples and controls. It reveals that the proportions of endothelial cells and inhibitory neurons were significantly higher in MSA, while oligodendrocytes were less prevalent in MSA (P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Subsequently, we calculated the correlations between HPRGs and cell types (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB). Oligodendrocytes showed positive correlations with both ZDHHC20 and AK5, whereas excitatory neurons displayed negative correlations. Additionally, scatter plots depicting the correlations between HPRGs and cells were generated (Fig.s 6C\u0026ndash;6D).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.6. Validation and subtype-specific expression patterns of ZDHHC20 and AK5 in MSA\u003c/h2\u003e \u003cp\u003eWe used the independent validation set GSE199258 to confirm our findings. In this dataset, ZDHHC20 and AK5 were significantly downregulated in the MSA group (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eA), consistent with our initial findings (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eH). ROC analysis validated their good diagnostic performance (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eB). The strong coefficient was also confirmed (R\u0026thinsp;=\u0026thinsp;0.77, P\u0026thinsp;=\u0026thinsp;2.1e-08, Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eC). Together, these findings suggest that the differential expression of HPRGs is not coincidental.\u003c/p\u003e \u003cp\u003eWe next explored HPRG expression in MSA clinical subtypes. Compared to the HC group, the expression of ZDHHC20 and AK5 was significantly downregulated in both the MSA cerebellar type (MSA-C) and the parkinsonian type (MSA-P). No significant difference was observed between the two clinical subtypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eD, \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eE).\u003c/p\u003e \u003cp\u003eSubsequently, we analyzed the expression in pathological subtypes. ZDHHC20 was significantly decreased in the olivopontocerebellar atrophy (OPCA) and striatonigral degeneration-olivopontocerebellar atrophy (SND-OPCA) subtypes compared to HC, but not in the striatonigral degeneration (SND) subtype (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eF). Furthermore, ZDHHC20 expression in the OPCA group was significantly lower than in the SND group, while there was no significant difference between SND-OPCA and either OPCA or SND (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eF). AK5 expression patterns were highly similar: expression was significantly downregulated in the OPCA and SND-OPCA subtypes compared to HC, but not in the SND subtype (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eG). AK5 expression in the SND subtype was significantly higher than that in the OPCA and SND-OPCA subtypes, and there was no significant difference between OPCA and SND-OPCA (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eG).\u003c/p\u003e \u003cp\u003eFinally, we conducted a further exploration of the expression status of HPRGs in oligodendrocyte samples. Both genes showed a non-significant downward trend in MSA samples (Fig.. 7H).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e3.7. Construction and analysis of upstream regulatory networks for ZDHHC20 and AK5\u003c/h2\u003e \u003cp\u003eTarget miRNAs for HPRGs were predicted using the miRTarBase, miRDB, and TargetScan databases. Eleven of these miRNAs had predictable target lncRNAs in the ENCORI database. Eight miRNAs (hsa-miR-653-5p, hsa-miR-548o-3p, hsa-miR-6509-3p, hsa-miR-2114-5p, hsa-miR-30d-5p, hsa-miR-642b-5p, hsa-miR-3200-5p, and hsa-miR-20b-5p) were associated with ZDHHC20, and three (hsa-miR-134-5p, hsa-miR-642b-3p, and hsa-miR-342-3p) were related to AK5. We predicted 35 common lncRNAs from these two groups of miRNAs (Table S4) and constructed a ceRNA network (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eA). Based on this network, we were able to gain a rough understanding of the common miRNAs shared between two genes and lncRNAs. The size and color of the lncRNA nodes, along with the size of their respective labels, represent the number of miRNAs with which the lncRNAs are associated (denoted as \u0026ldquo;degree\u0026rdquo;). We also investigated upstream transcription factors (TFs) using the NetworkAnalysis platform. The resulting TF-gene interaction network identified GATA2 as a common TF for both ZDHHC20 and AK5 (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e3.8. Single-nucleus RNA-seq analysis reveals oligodendrocyte downregulation of HPRGs in MSA\u003c/h2\u003e \u003cp\u003eQuality control yielded 165,577 cells and 24,214 genes (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eA). Subsequent dimensionality reduction utilized the top 2,000 HVGs (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eB). Twenty-four PCs were selected (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eC), and a resolution of 1 was determined for cell clustering based on results shown in Fig. S3. This process classified 50 cell subpopulations (Fig. \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003eD), which were annotated into eight major cell types: excitatory neurons (65487), inhibitory neurons (23765), astrocytes (18636), oligodendrocytes (42310), microglia (5096), oligodendrocyte progenitor cells (OPCs) (7153), endothelial cells (1601), and pericytes (1529) (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eA, Fig. S4). Fig.s 9B and Fig. S5A illustrate the composition and distribution of these cell types in the MSA and HC groups, respectively. Analysis of HPRG expression patterns revealed that ZDHHC20 was significantly expressed across all cell types except pericytes, whereas AK5 expression was restricted to oligodendrocytes, excitatory neurons, and inhibitory neurons (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eC). Crucially, both genes exhibited peak expression in oligodendrocytes (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eC, D), suggesting pivotal roles within this lineage. Within the dot plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eC), both HPRG expression levels and the proportion of expressing cells were reduced in the MSA group relative to controls in oligodendrocytes, corroborating our findings in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eH. Distribution plots (Fig. S5B, S5C) showed that ZDHHC20 and AK5 expression was generally high and densely concentrated in oligodendrocytes, consistent with Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eC and D. Consequently, oligodendrocytes were selected for subsequent analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e3.9. MSA exhibits a distinct oligodendroglial composition compared to HC\u003c/h2\u003e \u003cp\u003eAnalysis commenced with 42,310 oligodendrocytes. Dimensionality reduction and clustering leveraged the top 2,000 HVGs and 19 PCs, ultimately identifying 19 clusters at a resolution of 0.5 (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003eA, Fig. S6). This clustering revealed significant compositional shifts between the multiple system atrophy (MSA) and healthy control (HC) groups (Fig.s 10B and 10C). For instance, six clusters (1, 7, 10, 13, 16, 18) were nearly exclusive to MSA, whereas nine clusters (0, 2, 4, 5, 8, 9, 11, 12, 14) predominated in HC (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003eB). HPRG expression patterns across these 19 subpopulations are detailed in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003eD.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e3.10. Abnormal terminal differentiation and dynamic downregulation of HPRGs along the pseudo-temporal trajectory in MSA\u003c/h2\u003e \u003cp\u003ePseudotime analysis was conducted on a random subset of 20,000 microglial cells (N\u0026thinsp;=\u0026thinsp;42,310) using 1,418 HVGs for dimensionality reduction (Fig. S7A). Cells were subsequently ordered along a computed developmental trajectory, with color intensity indicating differentiation time (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eA). Seven distinct differentiation states were identified along the trajectory (Fig. S7B). Concurrently, the relative expression levels of ZDHHC20 and AK5 consistently decreased with increasing pseudotime (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eB). Group distribution densities showed divergent, opposing trends (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eC): HC cells were concentrated in the early and middle pseudotime stages, whereas MSA cells demonstrated significant late-stage accumulation. This pattern suggests that oligodendrocytes in MSA patients may progress toward an abnormal terminal differentiation state. Visualization of subpopulation distribution further revealed that MSA-specific clusters were primarily restricted to the mid-to-late stages (excluding cluster 18, which occupied the early-to-mid phases). Conversely, HC-specific clusters (e.g., clusters 1, 2, 4, 9, 11, 14) were concentrated in the early and middle stages, with cluster 8 representing an exception, located predominantly in the late stage (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eD, Fig. S7C). Collectively, these findings indicate that the relative suppression of ZDHHC20 and AK5 expression in MSA oligodendrocytes is associated with pathological alterations during disease onset and progression.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eMultiple System Atrophy (MSA), a sporadic, adult-onset, rapidly progressive, and ultimately fatal neurodegenerative disease, presents significant challenges in both diagnosis and treatment (Ndayisaba, et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Poewe, et al., 2022; Wan, et al., 2023). Therefore, it is necessary to identify effective diagnostic markers and therapeutic targets. Palmitoylation, one of the most common protein post-translational modifications, plays an indispensable role in neuronal development, synaptic plasticity, neural signal transduction, and neuroinflammation (B and Talwar, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; F, et al., 2024). Abnormal palmitoylation is often associated with neurodegenerative disorders (Cho and Park, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Currently, the relationship between palmitoylation and MSA remains unclear, and systematic analyses are lacking. Consequently, we performed a bioinformatics analysis to examine the expression patterns of palmitoylation-related genes in MSA and their potential functions. In this study, by integrating transcriptomic data, we identified two palmitoylation-associated genes, ZDHHC20 and AK5, and discovered a strong correlation between them; these two genes may serve as promising biomarkers. Our findings may contribute to elucidating and refining the pathogenesis of MSA.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003epalmitoylation, the most prevalent form of palmitoylation, is catalyzed by a family of enzymes known as palmitoyltransferases (PATs). This enzyme family typically harbors a conserved cysteine-rich zinc finger-like domain that encompasses the aspartate-histidine-histidine-cysteine (Asp-His-His-Cys, DHHC) sequence. Consequently, this family of enzymes is designated as the ZDHHC protein family (Buszka, et al., 2023; Fan, et al., 2024). These enzymes function by covalently attaching a 16-carbon saturated fatty acid, palmitic acid, to the cysteine residues of substrate proteins via thioester bonds, playing a pivotal role in diverse biological processes (Ko and Dixon, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). ZDHHC20, a member of ZDHHCs, has been predominantly studied in the context of tumors and the immune system. For example, one study showed that ZDHHC20 regulates cellular sensitivity to ferroptosis by palmitoylating GPX4 that is a core inhibitor of ferroptosis. Inducing ferroptosis is associated with the suppression of tumor occurrence and metastasis (Huang, et al., 2025). Inhibition of ZDHHC20 can lead to hyperactivation of the EGFR signaling pathway, thereby increasing tumor sensitivity to specific targeted therapies. Interferon-induced transmembrane protein 3 (IFITM3) is regulated by palmitoylation, and ZDHHC20 can significantly enhance the antiviral activity of IFITM3 (McMichael, et al., 2017). ZDHHC20 also plays a critical role in regulating the function of the T-cell co-stimulatory molecule CD80 (Lu, et al., 2024). Despite the fact that ZDHHC20 assumes a pivotal role in tumorigenesis and immune regulation, data from The Human Protein Atlas (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.proteinatlas.org\u003c/span\u003e\u003cspan address=\"https://www.proteinatlas.org\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) suggest that it is also prominently expressed in brain tissue. Furthermore, research has found that ZDHHC20 is considerably expressed in oligodendrocytes (Jeong, et al., 2025). Recently, numerous Mendelian Randomization (MR) studies have established associations between it and a variety of psychiatric disorders (Guo, et al., 2025). These findings suggest that ZDHHC20 may play an important role in the brain. This study is the first to find that ZDHHC20 expression is downregulated in MSA patients, suggesting that it may be a potential protective factor in MSA.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAdenylate kinases (AKs) are phosphotransferase enzymes widely present in organisms. They are responsible for catalyzing the reversible interconversion of adenine nucleotides and play a crucial role in maintaining adenine nucleotide balance and energy homeostasis (Fujisawa, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Siddiqui, et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Adenylate kinase 5 (AK5) was first discovered in brain tissue in 1999 and was considered to be specifically expressed in the brain (Van Rompay, et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e1999\u003c/span\u003e). Since then, elevated expression of AK5 has been detected in the neurons of the cerebral cortex and hippocampus. These brain regions are closely related to higher cognitive functions, learning, and memory (Siddiqui, et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). As a member of the AK family, AK5 participates not only in nucleotide conversion and energy metabolism but is also associated with the pathogenesis of various diseases, including cancer, asthma, diabetes, and neurological disorders (Fujisawa, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Siddiqui, et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Notably, AK5 has been found to be significantly downregulated in the substantia nigra of patients with mid- to late-stage Parkinson's disease (PD) (Garcia-Esparcia, et al., 2015). This phenomenon has also been observed in the entorhinal cortex and frontal cortex of patients with late-stage Alzheimer's disease (AD) (Ansoleaga, et al., 2015). A recent study indicated that AK5 downregulation may be associated with the onset of temporal lobe epilepsy (Lai, et al., 2016). By analyzing AK5 expression levels in the cerebellar white matter of MSA patients and healthy controls, we also observed this downregulation phenomenon.\u003c/p\u003e \u003cp\u003eZDHHC20 and AK5 exhibit overlapping functional pathways. AK5 is predominantly associated with energy metabolism, with significant enrichment of these pathways observed in the AK5 high-expression group\u0026mdash;consistent with its established biological role (Siddiqui, et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Notablely, the KEGG biological pathways enriched for ZDHHC20 and AK5 were nearly identical. Most of these pathways are related to mitochondria and respiratory chain complex I. Furthermore, GO analysis of ZDHHC20 identified many energy metabolism-related pathways, some of which also appeared in the AK5 enrichment results. These findings suggest a potential functional interaction between ZDHHC20 and AK5. While the precise mechanism remains to be elucidated, palmitoylation is known to play a critical role in the establishment and maintenance of normal neurological function. (B and Talwar, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; F, et al., 2024). Therefore, we speculate that the palmitoylation activity of ZDHHC20 may influence the expression of AK5 through certain mechanisms, such as by regulating transcription factors. Furthermore, mitochondria are the \"powerhouses\" of the cell, and their dysfunction is a common feature of many neurodegenerative diseases (Mantle and Hargreaves, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This pathological feature may also involve in the pathogenesis of MSA(Krismer, et al., 2024; Wan, et al., 2023). Our results suggest that ZDHHC20 and AK5 may be involved in pathological processes in MSA similar to those in the aforementioned neurodegenerative diseases (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB, \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD). Consequently, we propose that downregulation of ZDHHC20 and\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAK5 may exacerbate MSA progression by impairing mitochondrial function.\u003c/p\u003e \u003cp\u003eIn our study, the overall expression levels of ZDHHC20 and AK5 were significantly downregulated in the cerebellar white matter tissues of MSA patients, and a positive correlation was observed between the expression of these two genes. In human white matter tissue, oligodendrocytes are the predominant cell type, and these two markers exhibit considerable expression levels in oligodendrocytes (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eD). We speculate that their low expression may result from extensive loss and death of oligodendrocytes in the affected tissues of late-stage MSA samples, while their high correlation may be related to the degree of oligodendrocyte loss\u0026mdash;though the possibility of co-expression between them cannot be ruled out. Analyses of various MSA subtypes in the study also indirectly support this notion. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eD\u0026mdash;G, the differential expression results of the two core genes across clinical and pathological subtypes of MSA are not entirely consistent. Specifically, the expression levels of HPRGs are significantly decreased in both clinical phenotypes of MSA compared to the control group. However, this trend is not observed in MSA-SNP and subtypes, whereas decreased expression is observed in MSA-OPCA and SND-OPCA. Generally, MSA-P and MSA-C correspond pathologically to SND and OPCA, respectively. (Poewe, et al., 2022). However, a growing body of pathological research indicates that, in addition to MSA subtypes dominated by either SND or OPCA pathology, a considerable number of cases present as mixed-type MSA with comparable pathological severity of SND and OPCA (Jellinger, et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Ozawa, et al., 2004). This suggests that pathological changes in MSA are not confined to a single brain region but become more widespread as the disease progresses(Ndayisaba, et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Wan, et al., 2023). Moreover, results from the single-nucleus analysis (including dot plots shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eC and expression changes along pseudotime presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eB) consistently show widespread downregulation of HPRGs in oligodendrocytes of MSA patients. Consistent findings were also observed in our analysis of the GSE199724 dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eH). This indicates that the overall low expression of HPRGs in tissues is not solely due to the extensive loss of oligodendrocytes but may also result from suppressed expression within dysfunctional oligodendrocytes. These findings further support the potential of HPRGs as diagnostic markers.\u003c/p\u003e \u003cp\u003eOligodendrocytes are the earliest and most severely affected cell type in MSA, playing a central role in the onset and progression of the disease(Hsiao, et al., 2023; Poewe, et al., 2022). The accumulation of misfolded α-synuclein within oligodendrocytes underlies the formation of GCIs(Schweighauser, et al., 2020; Wan, et al., 2023). Oligodendrocytes subjected to prolonged high GCI burden exhibit structural and functional impairments, often accompanied by demyelination and subsequent neuronal loss(Ndayisaba, et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Poewe, et al., 2022). In the early to middle stages of the disease, GCI density shows a strong positive correlation with the extent of neuronal loss and disease duration(Ozawa, et al., 2004). However, in advanced-stage MSA cases, this correlation reverses. Multiple studies have reported a decline in GCI density in severely atrophic brain regions, primarily due to the widespread death of host oligodendrocytes(Jellinger, et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Tanaka, et al., 2025). Various mechanisms contribute to this cell death: when intracellular protein load exceeds a certain threshold, or when mitochondrial dysfunction compromises the cell\u0026rsquo;s capacity to sustain basic metabolic activities, oligodendrocytes may undergo apoptosis or necrosis, leading to the release of GCIs. These insoluble fibrils may remain in the extracellular matrix for a short period but are ultimately phagocytosed and removed by microglia, accompanied by severe neuroinflammation(Tanaka, et al., 2025). In MSA patients, oligodendrocyte precursor cells (OPCs) fail to differentiate into mature myelin-forming cells to repair the damage, a process inhibited by mechanisms such as the toxic effects of α-synuclein on OPCs(Hsiao, et al., 2023; May, et al., 2014). Consequently, in end-stage tissue, severe demyelination and neuronal loss may be observed, yet the visible GCI burden is lower than in less affected regions. In our cell deconvolution analysis of MSA cerebellar white matter transcriptomic data, we observed a reduction in oligodendrocytes (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA). Both ZDHHC20 and AK5 showed positive correlations with oligodendrocyte proportion (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eB), and their downregulation may indirectly reflects the extensive loss of oligodendrocytes in the terminal stage of MSA.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe findings of this study have been discussed in detail above and are of significant importance. However, this study also has several limitations. First, although the ROC curves of HPRGs in the training and validation sets demonstrated good diagnostic performance (AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.7), we did not construct a diagnostic model. If these markers can be combined with other influencing factors to build a diagnostic model in the future, it may lead to even better diagnostic outcomes. Second, the bulk RNA-seq data and single-nucleus data used in this study were derived from different tissue sources. The prefrontal cortex of advanced MSA-P patients may exhibit a certain degree of involvement, although the extent of atrophy is generally mild(Poewe, et al., 2022; Tanaka, et al., 2025). While we observed low expression of these biomarkers in oligodendrocytes in such tissues, the expression of HPRGs in severely affected tissues (such as cerebellar white matter) remains unclear. Additionally, all analyzed samples were obtained from end-stage MSA patients, and the expression patterns of these genes in early-stage patients require further investigation. Third, the current use of HPRGs as diagnostic markers is limited, as obtaining corresponding brain tissue from MSA patients for diagnosis is highly invasive. Therefore, it is necessary to identify other tissues\u0026mdash;such as blood, cerebrospinal fluid, or even skin\u0026mdash;that may reflect such expression patterns. Fourth, due to the scarcity of publicly available samples for MSA, all bulk RNA-seq data used in our study were derived from a single study. Fortunately, this study comprised three independent cohorts, with samples sourced from different brain banks worldwide, exhibiting significant batch effects. This provided a basis for selecting training and validation sets. Moreover, MSA patients are extremely rare, and obtaining postmortem brain tissue from them is particularly challenging. As a result, experimental validation was not conducted in this study. If conditions permit in the future, further experimental validation should be carried out to verify the reliability of these findings. Fifth, the differential expression results of HPRGs in oligodendrocytes showed a non-significant downward trend, which may be related to the small sample size of oligodendrocytes (n\u0026thinsp;=\u0026thinsp;12). A small sample size may lead to discrepancies between the results and expectations. In addition, heterogeneity in factors such as study populations, sample processing time and methods, and various clinical variables (e.g., disease duration, age at death, comorbid conditions) may introduce bias.\u003c/p\u003e"},{"header":"5. Conclusions","content":"\u003cp\u003eBy applying multiple machine learning algorithms and logistic regression analysis, we identified ZDHHC20 and its highly co-expressed gene, AK5. We verified their significant downregulation in MSA and demonstrated their strong performance as potential diagnostic markers. Furthermore, we explored their associated biological functions and potential upstream regulators, and elucidated their distribution and expression across various cell types at the single-nucleus level. We also simulated alterations in their expression levels during the progression of MSA. The findings provide novel insights into the investigation of the pathogenesis of MSA and concurrently offer innovative alternatives for the diagnosis of MSA.\u003c/p\u003e \u003cp\u003e \u003cb\u003eAuthorship Contributions\u003c/b\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eZhipeng Lu\u003c/b\u003e: Writing - original draft,Writing - review \u0026amp; editing, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization. \u003cb\u003ePu Fang\u003c/b\u003e: Writing - review \u0026amp; editing, Project administration, Supervision, Funding acquisition. \u003cb\u003eZhongqi Li\u003c/b\u003e: Writing - review \u0026amp; editing, Methodology, Formal analysis. \u003cb\u003eZhibiao Yin\u003c/b\u003e: Writing - review, Supervision, Methodology. \u003cb\u003eJialong Liu\u003c/b\u003e: Writing - review \u0026amp; editing.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAlzheimer's disease\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAK\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAdenylate kinase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eALS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eAmyotrophic lateral sclerosis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArea under the curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCBD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCorticobasal degeneration\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eceRNA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCompeting endogenous RNA\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCor\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eCorrelation coefficient\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDEG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDifferentially expressed gene\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDecision tree\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eEMG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eElectromyography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFDR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eFalse discovery rate\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGBM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGradient boosting machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGCI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGlial cytoplasmic inclusion\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGEO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGene expression omnibus\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGLM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eGeneralized linear model\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHPRG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHub palmitoylation-related gene\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHVG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eHighly Variable Gene\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKNN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eK-nearest neighbors\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLeast absolute shrinkage and selection operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMendelian randomization\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMSA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMultiple system atrophy\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMSA-C\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMSA cerebellar type\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMSA-P\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMSA parkinsonian type\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNNET\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNeural network\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOPC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eOligodendrocyte progenitor cell\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOPCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eOlivopontocerebellar atrophy\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eOR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eOdds ratio\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePAT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePalmitoyltransferase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePrincipal component analysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePD\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eParkinson's disease\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePRG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePalmitoylation-related gene\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePSP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eProgressive supranuclear palsy\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eQC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eQuality control\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eRandom forest\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSAOA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSporadic adult-onset ataxia of unknown etiology\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSND\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eStriatonigral degeneration\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSNCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSynuclein, alpha\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSVM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSupport vector machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTFs\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eTranscription factors\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eαSyn\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eα-synuclein\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthorship Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eZhipeng Lu:\u0026nbsp;\u003c/strong\u003eWriting - original draft,Writing - review \u0026amp; editing, Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization. \u003cstrong\u003ePu Fang:\u0026nbsp;\u003c/strong\u003eWriting - review \u0026amp; editing, Project administration, Supervision, Funding acquisition. \u003cstrong\u003eZhongqi Li:\u003c/strong\u003e Writing - review \u0026amp; editing, Methodology, Formal analysis. \u003cstrong\u003eZhibiao Yin:\u0026nbsp;\u003c/strong\u003eWriting - review, Supervision, Methodology. \u003cstrong\u003eJialong Liu:\u0026nbsp;\u003c/strong\u003eWriting - review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe express our sincere gratitude to all the authors who actively participated in this research. In particular, we are indebted to the corresponding author for his invaluable assistance in facilitating the progress of this study. Simultaneously, we are profoundly thankful to the GEO and Synapse database and to all the contributors who uploaded the relevant transcriptomic data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e: \u003cem\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e: \u003cem\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eEthics approval: Not applicable.\u003c/p\u003e\n\u003cp\u003eConsent to Participate: Not applicable.\u003c/p\u003e\n\u003cp\u003eConsent for Publication: Not applicable. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e The authors declare no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAnsoleaga B et al (2015) Deregulation of purine metabolism in Alzheimer's disease. Neurobiol Aging 36(1):68\u0026ndash;80\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eB SP, Talwar P (2025) Influence of palmitoylation in axonal transport mechanisms in neurodegenerative diseases. Front Cell Neurosci 19:1613379\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBuszka A et al (2023) S-Palmitoylation of Synaptic Proteins in Neuronal Plasticity in Normal and Pathological Brains. Cells ;12(3)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen B et al (2018) Profiling Tumor Infiltrating Immune Cells with CIBERSORT. Methods Mol Biol 1711:243\u0026ndash;259\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen Y, Li Y, Wu L (2024) Protein S-palmitoylation modification: implications in tumor and tumor immune microenvironment. Front Immunol 15:1337478\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCho E, Park M (2016) Palmitoylation in Alzheimer's disease and other neurodegenerative diseases. Pharmacol Res 111:133\u0026ndash;151\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eF SM et al (2024) Mechanisms and functions of protein S-acylation. Nat Rev Mol Cell Biol 25(6):488\u0026ndash;509\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFan Z et al (2024) Modulators for palmitoylation of proteins and small molecules. Eur J Med Chem 271:116408\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFujisawa K (2023) Regulation of Adenine Nucleotide Metabolism by Adenylate Kinase Isozymes: Physiological Roles and Diseases. Int J Mol Sci ;24(6)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarcia-Esparcia P et al (2015) Purine metabolism gene deregulation in Parkinson's disease. Neuropathol Appl Neurobiol 41(7):926\u0026ndash;940\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuo Q et al (2025) Multi-Omics Mendelian Randomization Identifies a DNA Methylation-ZDHHC20-Immune Axis Associated With Schizophrenia Risk. Brain Behav 15(8):e70722\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHao Y et al (2024) Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 42(2):293\u0026ndash;304\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHsiao JT et al (2023) Role of Oligodendrocyte Lineage Cells in Multiple System Atrophy. Cells ;12(5)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang B et al (2025) Palmitoylation-dependent regulation of GPX4 suppresses ferroptosis. Nat Commun 16(1):867\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJellinger KA, Seppi K, Wenning GK (2005) Grading of neuropathology in multiple system atrophy: proposal for a novel scale. Mov Disord 20(Suppl 12):S29\u0026ndash;36\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeong HK et al (2025) ,\u003cem\u003e. Micro-scale control of oligodendrocyte morphology and myelination by the intellectual disability-linked protein acyltransferase ZDHHC9. Elife ;13.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo PJ, Dixon SJ (2018) Protein palmitoylation and cancer. EMBO Rep ;19(10)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKrismer F et al (2024) Multiple system atrophy: advances in pathophysiology, diagnosis, and treatment. Lancet Neurol 23(12):1252\u0026ndash;1266\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLai Y et al (2016) Down-regulation of adenylate kinase 5 in temporal lobe epilepsy patients and rat model. J Neurol Sci 366:20\u0026ndash;26\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeisch DM (2024) a.E.D.a.K.H.a.A.W.a.F. e1071: Misc Functions of the Department of Statistics, Probability Theory Group (Formerly: E1071), TU Wien. Arxiv preprint arXiv 1(1):548\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi JH et al (2014) starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res 42(Database issue):D92\u0026ndash;97\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu B et al (2024) zDHHC20-driven S-palmitoylation of CD80 is required for its costimulatory function. Acta Pharmacol Sin 45(6):1214\u0026ndash;1223\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLucas C et al (2016) Yeast Gup1(2) Proteins Are Homologues of the Hedgehog Morphogens Acyltransferases HHAT(L): Facts and Implications. J Dev Biol ;4(4)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMantle D, Hargreaves IP (2022) Mitochondrial Dysfunction and Neurodegenerative Disorders: Role of Nutritional Supplementation. Int J Mol Sci ;23(20)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMathys H et al (2019) Single-cell transcriptomic analysis of Alzheimer's disease. Nature 570(7761):332\u0026ndash;337\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMay VE et al (2014) α-Synuclein impairs oligodendrocyte progenitor maturation in multiple system atrophy. Neurobiol Aging 35(10):2357\u0026ndash;2368\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcMichael TM et al (2017) The palmitoyltransferase ZDHHC20 enhances interferon-induced transmembrane protein 3 (IFITM3) palmitoylation and antiviral activity. J Biol Chem 292(52):21517\u0026ndash;21526\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNdayisaba A, Halliday GM, Khurana V (2025) Multiple System Atrophy: Pathology, Pathogenesis, and Path Forward. Annu Rev Pathol 20(1):245\u0026ndash;273\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNido GS et al (2025) Single-nucleus transcriptomics reveals disease- and pathology-specific signatures in α-synucleinopathies. Brain 148(5):1588\u0026ndash;1603\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOzawa T et al (2004) The spectrum of pathological involvement of the striatonigral and olivopontocerebellar systems in multiple system atrophy: clinicopathological correlations. Brain 127(Pt 12):2657\u0026ndash;2671\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiras IS et al (2020) Transcriptional profiling of multiple system atrophy cerebellar tissue highlights differences between the parkinsonian and cerebellar sub-types of the disease. Acta Neuropathol Commun 8(1):76\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePoewe W et al (2022) Multiple system atrophy. Nat Rev Dis Primers 8(1):56\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQi X et al (2025) Integrated Analysis of Ferroptosis- and Cellular Senescence-Related Biomarkers in Atherosclerosis Based on Machine Learning and Single-Cell Sequencing Data. J Inflamm Res 18:9283\u0026ndash;9305\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu X et al (2017) Single-cell mRNA quantification and differential analysis with Census. Nat Methods 14(3):309\u0026ndash;315\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQiu X et al (2017) Reversed graph embedding resolves complex single-cell trajectories. Nat Methods 14(10):979\u0026ndash;982\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRauluseviciute I et al (2024), \u003cem\u003eJASPAR. : 20th anniversary of the open-access database of transcription factor binding profiles. Nucleic Acids Res 2024;52(D1):D174\u0026ndash;d182.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchweighauser M et al (2020) Structures of α-synuclein filaments from multiple system atrophy. Nature 585(7825):464\u0026ndash;469\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiddiqui MS, Shahi MH, Castresana JS (2024) The role of the adenylate kinase 5 gene in various diseases and cancer. J Clin Transl Sci 8(1):e96\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTanaka MT et al (2025) Pathological and Molecular Insights into the Early Stage of Multiple System Atrophy. Cells ;14(24)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrapnell C et al (2014) The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat Biotechnol 32(4):381\u0026ndash;386\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVan Rompay AR, Johansson M, Karlsson A (1999) Identification of a novel human adenylate kinase. cDNA cloning, expression analysis, chromosome localization and characterization of the recombinant protein. Eur J Biochem 261(2):509\u0026ndash;517\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWan L et al (2023) Multidimensional biomarkers for multiple system atrophy: an update and future directions. Transl Neurodegener 12(1):38\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang J, Devlin B, Roeder K (2020) Using multiple measurements of tissue to estimate subject- and cell-type-specific gene expression. Bioinformatics 36(3):782\u0026ndash;788\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWenning GK et al (2004) Multiple system atrophy. Lancet Neurol 3(2):93\u0026ndash;103\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWenning GK et al (2022) The Movement Disorder Society Criteria for the Diagnosis of Multiple System Atrophy. Mov Disord 37(6):1131\u0026ndash;1148\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu G et al (2012) clusterProfiler: an R package for comparing biological themes among gene clusters. Omics 16(5):284\u0026ndash;287\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang H et al (2024) ZDHHC20-mediated S-palmitoylation of YTHDF3 stabilizes MYC mRNA to promote pancreatic cancer progression. Nat Commun 15(1):4642\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou KR, Liu HJ, Zheng S, Liu WJ SR, An Encyclopedic Regulatory and Functional Atlas of RNA Interactomes\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-molecular-neuroscience","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jomn","sideBox":"Learn more about [Journal of Molecular Neuroscience](https://www.springer.com/journal/12031)","snPcode":"12031","submissionUrl":"https://submission.nature.com/new-submission/12031/3","title":"Journal of Molecular Neuroscience","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Multiple System Atrophy (MSA), Palmitoylation, Machine learning, Energy metabolism, Oligodendrocytes","lastPublishedDoi":"10.21203/rs.3.rs-8711800/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8711800/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eMultiple system atrophy (MSA) is a fatal neurodegenerative disorder lacking effective diagnostic tools. While protein palmitoylation is crucial for neuronal function, its specific role in MSA pathogenesis remains unexplored.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe integrated bulk and single-nucleus RNA sequencing (snRNA-seq) data from postmortem MSA brain tissues. Eight machine learning algorithms were utilized to screen palmitoylation-related genes. Downstream analyses, including functional enrichment, cellular deconvolution, and pseudotime trajectory inference, were then conducted.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eZDHHC20 and its highly correlated gene, AK5, were identified as hub genes. Both demonstrated significant downregulation in MSA, particularly within the cerebellar white matter. Functional enrichment analysis linked this expression pattern to mitochondrial dysfunction and impaired energy metabolism. Furthermore, snRNA-seq revealed that ZDHHC20 and AK5 are predominantly expressed in oligodendrocytes and are progressively suppressed during the abnormal terminal differentiation trajectory observed in MSA.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eZDHHC20 and AK5 represent promising diagnostic biomarkers for MSA. These findings highlight the potential role of palmitoylation in MSA pathogenesis, providing new insights into the diagnosis and treatment of MSA.\u003c/p\u003e","manuscriptTitle":"Transcriptomic Analysis and Multiple Machine Learning Approaches Identify ZDHHC20 and Its Highly Correlated Gene AK5 as Diagnostic Markers in Multiple System Atrophy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-13 13:13:16","doi":"10.21203/rs.3.rs-8711800/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-21T05:50:15+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-16T01:59:54+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"258956518509350162708367380079060170473","date":"2026-02-10T14:16:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"11298155029851581091865944062865251167","date":"2026-02-10T09:26:26+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-02-10T08:24:33+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-29T12:14:31+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-29T12:13:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Molecular Neuroscience","date":"2026-01-27T14:20:43+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-molecular-neuroscience","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jomn","sideBox":"Learn more about [Journal of Molecular Neuroscience](https://www.springer.com/journal/12031)","snPcode":"12031","submissionUrl":"https://submission.nature.com/new-submission/12031/3","title":"Journal of Molecular Neuroscience","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"49a34e77-5fdf-453b-aee1-0fe44a8909bb","owner":[],"postedDate":"February 13th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-08T03:53:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-13 13:13:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8711800","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8711800","identity":"rs-8711800","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}