From Neural Connectopathy to a Therapeutic Path: An Integrated Multi-omics Framework Identifies a Causal Gene and Drug Candidates for Parkinson's Disease with mild cognitive impairment

From Neural Connectopathy to a Therapeutic Path: An Integrated Multi-omics Framework Identifies a Causal Gene and Drug Candidates for Parkinson's Disease with mild cognitive impairment | 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 Article From Neural Connectopathy to a Therapeutic Path: An Integrated Multi-omics Framework Identifies a Causal Gene and Drug Candidates for Parkinson's Disease with mild cognitive impairment Aydos Alemjan, Yuankai Gu, Li Wang, Wei Xu, Qing Ning, Lei Shen, and 5 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7937441/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Parkinson’s disease with mild cognitive impairment (PD-MCI) represents a devastating outcome for patients, yet its pathogenic drivers remain poorly understood, and predictive biomarkers are critically lacking. This study aims to establish an early diagnostic model for PD-MCI (from PD), delineate causal mechanisms and identify translational opportunities for PD-MCI. Methods We established an integrated analytical pipeline using the longitudinal Parkinson’s Progression Markers Initiative (PPMI) cohort. 1 Our framework combined plasma transcriptomics and cerebrospinal fluid proteomics with systems biology, machine learning, longitudinal proteomic trajectories analysis, clinical correlation analysis, single-nucleus RNA sequencing, Mendelian randomization, and computational drug repurposing. Results We identified a molecular signature of PD-MCI progression, revealing profound dysregulation of pathways governing neural connectivity—axon guidance, synaptogenesis, and neuron projection development—suggesting a "neural connectopathy" mechanism. Longitudinal proteomic trajectories analysis revealed that the pattern of dysregulated proteins changes years before PD-MCI diagnosis, underscoring an early pathological window. Gene-interaction network and protein-interaction network highlighted key hubs. We then implemented LASSO analysis to select key features and developed an interpretable machine learning model (Logistic Regression, transcripts + proteomics + demographics, AUC = 0.74) that accurately predicted cognitive decline, with SHAP analysis to interpret the order. Crucially, Mendelian randomization established putative causal drivers of PDD risk, and single-nucleus sequencing anchored these findings to the brain, showing altered expression of the causal candidate. specifically in excitatory neurons. Finally, computational drug repurposing nominated targeted candidates, including MEDRONIC ACID, ZILEUTON and ATENOLOL, for immediate therapeutic consideration. Conclusion Our study delivers an integrated translational advance by: ( 1 ) defining "neural connectopathy" as a core disease mechanism in PD-MCI; ( 2 ) establish an efficient machine learning model to early diagnose PD-MCI from PD; (3) providing a clinically actionable predictive tool, fingding causal genes and proposing a pathway to treatment via drug repurposing. This end-to-end framework establishes a foundation for early prediction, biological insight, and targeted intervention in PD-MCI. Biological sciences/Computational biology and bioinformatics Health sciences/Diseases Health sciences/Neurology Biological sciences/Neuroscience Parkinson’s disease with mild cognitive impairment Multi-omics Machine learning Mendelian Randomization Drug Repurposing Neural Connectopathy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Introduction Parkinson's disease (PD) is the second most common neurodegenerative disorder worldwide, characterized by both cardinal motor symptoms and numerous non-motor complications. 2–4 Among these, Parkinson's disease with mild cognitive impairment (PD-MCI) is one of the most debilitating and burdensome, affecting up to 80% of patients within 15 years of onset. 5–7 PD-MCI’s emergence signals severe quality-of-life decline for patients and caregivers and no effective disease-modifying therapies exist. 8,9 PD-MCI’s neuropathological hallmarks include the widespread alpha-synuclein aggregation (Lewy bodies), often coexisting with Alzheimer's disease-related pathologies (amyloid-β plaques, neurofibrillary tau tangles). 10 This indicates a complex and multifactorial etiology. 11 This complexity poses a major challenge: inability to accurately predict which PD patients will develop mild cognitive impairment or dementia, and incomplete understanding of underlying molecular mechanisms, hindering targeted prevention. 12 In recent years, high-throughput omics technologies have offered unprecedented opportunities to unravel neurodegenerative diseases’ molecular complexity. 13 Genome-wide association studies (GWAS) have identified many PD risk loci, but their ability to explain PD-MCI risk remains limited. 10,14,15 In contrast, transcriptomic and proteomic profiling of accessible biofluids (e.g., plasma, cerebrospinal fluid [CSF]) provides dynamic snapshots of pathological processes and holds promise for discovering diagnostic/prognostic biomarkers. 16 However, most studies to date are cross-sectional or focused on a single omics layer, failing to capture the temporal dynamics of disease progression and interactions between molecular levels. 17,18 A longitudinal, multi-omics approach is thus imperative to map the chronological molecular events leading to PD-MCI. Furthermore, conventional association studies using bulk tissue or biofluids cannot pinpoint the specific cellular origins of dysregulated molecules, a critical step for understanding mechanism. 19 Additionally, while omics analyses reveal correlations, they cannot establish causality. Mendelian randomization (MR), an epidemiological method using genetic variants as instrumental variables, has become a powerful tool to infer causal relationships between an exposure (e.g., gene expression) and outcomes (e.g., PD-MCI risk), reducing confounding and reverse causation. 20,21 Single-nucleus RNA sequencing (snRNA-seq) now enables unbiased analysis of cell-type-specific transcriptional changes in postmortem human brain, providing key spatial context for biomarker discoveries. 20,22,23 To address these gaps, we designed an integrative translational study using the large-scale, deeply phenotyped Parkinson's Progression Markers Initiative (PPMI) cohort. 24 We hypothesized that integrated multi-omics analysis (plasma transcriptomics + CSF proteomics) could identify core molecules/pathways driving PD-related cognitive decline. We further proposed that these findings could be integrated into a PD-MCI risk predictive model, tested for causality using MR, and validated in specific brain cell types via snRNA-seq. Finally, we aimed to translate these discoveries into therapeutic hypotheses through computational drug repurposing. 2. Results 2.1. Cohort Selection and an Integrative Analytical Framework for PD-MCI To systematically analyze the molecular mechanisms of PD-MCI, we used the longitudinal PPMI cohort. From 4115 initial participants, we selected 322 PD patients as the final analysis cohort, all of which with complete baseline multi-omics data (plasma RNA-seq and CSF proteomics) and 4-year longitudinal cognitive assessments (MoCA) (Fig. 1A, Table 1). We then designed and implemented a comprehensive analytical workflow, moving beyond simple association analysis to focus on predictive model development and clinical application. This framework integrated differential expression analysis, machine learning-based prediction, genetic causal inference, single-cell validation, and drug repurposing (Fig. 1B), establishing a translational research pipeline from biomarker discovery to therapeutic hypothesis. Table 1 Baseline characteristics of participants from the Parkinson's Progression Markers Initiative (PPMI) cohort, stratified by PD-MCI outcome. PD without MCI (n = 213) PD-MCI (n = 207) Total (n = 420) p-value RNA-seq 0.5114† No 32 (15.0%) 37 (17.9%) 69 (16.4%) Yes 181 (85.0%) 170 (82.1%) 351 (83.6%) Proteomics 0.2170† No 11 (5.2%) 18 (8.7%) 29 (6.9%) Yes 202 (94.8%) 189 (91.3%) 391 (93.1%) Sex 0.0143† Female 98 (46.0%) 70 (33.8%) 168 (40.0%) Male 115 (54.0%) 137 (66.2%) 252 (60.0%) Age 58.2 (51.7–64.5) 66.5 (59.8–70.7) 62.2 (55.1–68.7) < 0.001¶ Education Years 16.0 (14.0–18.0) 16.0 (13.5–18.0) 16.0 (14.0–18.0) 0.0646¶ REM hours 3.0 (2.0–5.0) 4.0 (2.0–7.0) 4.0 (2.0–6.0) 0.0285¶ PD Diagnosis Age 56.3 (49.6–62.5) 63.8 (58.0-68.8) 60.6 (53.4–66.5) < 0.001¶ †: chi-squre tests, ¶: Wilcoxon rank-sum tests. 2.2. A Multi-omics Landscape Links Neural Connectivity Pathways to PD-MCI Pathogenesis We first sought to define the molecular signature of PD-MCI progression. Differential expression analysis between PD patients who developed mild cognitive impairment (MCI) vs. PD patitents who d i d not identified 19 differential transcripts (5 upregulated) and 44 differential proteins (2 upregulated) (Fig. 2A, B). Notably, functional enrichment analysis of proteomics and RNA-seq showed similar patterns, where significant pathways were highly concentrated in biological processes critical for neuronal circuit integrity 15 Pathways like axon guidance, neuron projection development, and synapse organization were consistently and significantly enriched across both molecular layers (Fig. 2C-F). This convergent signal strongly suggested "neural connectopathy" as a PD-MCI pathology. Gene -Gene interaction network and Protein-protein interaction (PPI) network 25 analysis further refined these findings: it distilled the complex molecular landscape into a core functional module and identified key hub genes 13 , including SLC38A11, PLEKHH2, and protein network hubs such as COL6A1 and ITGAV which act as central regulators of this dysregulated network (Fig. 2E). (A-B) Volcano plots of differential expression analysis comparing PD-MCI patients and those with normal cognition in 4 years. (A) Plasma RNA-seq analysis. (B) CSF proteomics analysis.. (C-D) Gene Ontology (GO) enrichment analysis of biological processes. (C) GO terms significantly enriched among significant RNA. (D) GO terms significantly enriched among significant proteins. (E-F) Protein-Protein Interaction (PPI) network of core differentially expressed proteins. 2.3. Longitudinal Proteomic Trajectories Reveal Dynamic Biomarker Changes Years Before Clinical Diagnosis To map the temporal sequence of molecular events leading to cognitive impairment, we tracked the expression of significantly differential proteins (identified in baseline analysis) over the pre-diagnosis period (up to 4 years before reaching PD-MCI). Unsupervised k-means clustering of these longitudinal profiles showed that the PD-MCI proteomic landscape is not static. It evolves through distinct, coordinated dynamic patterns (Fig. 3B). We identified five major trajectory clusters (Cluster 1 to Cluster 5), each with a unique temporal signature (Fig. 3C). Notably, most differential proteins (65.9%) fell into clusters defined by early and mid-term changes (Cluster 3 and Cluster 4), indicating that key pathological processes start years before PD-MCI is clinically diagnosed (Fig. 3D). The trajectory of proteins shows a turning point around 3 years before MCI diagnosis, which may be related to the temporal characteristics of PD patients enrolled in the PPMI cohort and the selection of time points during follow-up. 26 Proteins in Cluster 2 and Cluster 5 exhibit a relatively sustained decrease or increase before MCI diagnosis, and these proteins are closely associated with mechanisms such as extracellular matrix remodeling, metabolic homeostasis, mitochondrial integrity, oxidative stress, and inflammation. Representative trajectories of individual proteins further highlighted this heterogeneity: some markers showing progressive elevation or decline, while others demonstrating an curved pattern (Fig. 3A). Collectively, these findings demonstrate that PD-MCI’s molecular pathophysiology begins long before clinical symptoms appear and follows a multi-phase temporal pattern, providing a critical window for early intervention. (A) Temporal expression patterns (Z-score normalized levels) of individual proteins (top 10 selected from LASSO). These trajectories illustrate the heterogeneous temporal dynamics, with some proteins exhibiting progressive elevation (e.g., Pro-inflammatory markers) and others showing early decline (e.g., Synaptic proteins). (B) Average protein expression trajectories for each cluster. Lines represent the mean Z-score of all proteins in a cluster, shading indicates the standard error of the mean. (C) Heatmap displaying expression patterns of all significantly differentiential protein. (D) Pie chart illustrating the proportion of dysregulated proteins across 5 trajectory clusters. 2.4. A Robust Predictive Model for PD-MCI from PD and its Interpretable Biological Drivers To translate our findings into a clinically actionable tool, we developed a machine learning framework to predict individual risk of cognitive decline. 27 LASSO regression first refined the differential expressed candidates to a parsimonious set of 6 transcripts and 7 proteins respectively (Fig. 4A, B). We constructed and compared models of three machine learning algorithms based on four variable patterns: only demographic factors (age, sex, education years); demographic factors plus transcriptomic targets; demographic factors plus proteomic targets; and demographic factors plus both transcriptomic and proteomic targets. These models were evaluated using 5-fold cross-validation. 27 Among the evaluated algorithms, the Logistic Regression model showed superior and robust predictive performance. (AUC = 0.7303, 0.7685, 0.7241, 0.7376), outperforming Random Forest (AUC = 0.6314, 0.7043, 0.6655, 0.6483) and XGBoost (AUC = 0.6440, 0.6674, 0.6385, 0.6938) (Supplementary Fig. 1, Fig. 4C-E). ROC curves were painted plotted for the model incorporating demographic factors, transcriptomic targets, and proteomic targets, comparing different machine learning algorithm (Fig. 4F). Importantly, SHAP analysis clarified the model's decision-making process across different variable patterns (Fig. 5A, 5B, 5C). This interpretability connects the model's predictive power back to specific biological insights, validating the relevance of our omics findings. (A-B) LASSO regression to select the most important biomarkers from DEG results. (A) Plasma transcripts; (B) CSF Proteomics. (C-E) Comparative performance of multiple machine learning algorithms. Box plots display the distribution of ROC-AUC score for Random Forest, XGBoost, and Logistic Regression models evaluated via 5-fold cross-validation. (C) Models based on plasma transcripts + demographics; (D) Models based on CSF proteomics + demographics; (E) Models based on plasma transcripts + CSF proteomics + demographics. (F) Overall model performance evaluation on the independent test set. ROC curves illustrate the discriminative ability of models based on plasma transcripts + CSF proteomics + demographics. (A-C) SHAP Analysis for Model Interpretability. The plot ranks top features by their mean absolute SHAP value, representing their overall importance in the optimal model's predictions. (A) CSF Proteomics + demographics; (B) Plasma Transcripts + demographics; (C) CSF Proteomics + Plasma Transcripts + demographics. 2.5 Cross-sectional Associations Link Candidate Proteins to Domain-Specific Cognitive Performance To confirm the clinical relevance of our identified Plasma RNA and CSF protein biomarkers, we addressed their cross-sectional associations with performance across key cognitive domains. Using multivariate linear regression models adjusting for age and years of education 28 , we visualized the standardized effect sizes as a heatmap (Fig. 6). This analysis revealed distinct, coherent association patterns between specific proteins and cognitive functions. For upregulated genes and proteins in the PD-MCI group identified in differential expression analysis, they are generally and stably negatively correlated with cognitive-related scores, and vice versa. The heatmap illustrates that PD-MCI molecular signatures are not unifrom but differentially linked to impairments in specific cognitive domains. This domain-specific mapping enhances the biological plausibility of our candidate biomarkers and offers a more nuanced understanding of how distinct molecular pathways may contribute to the heterogeneous cognitive profiles in PD-MCI. Heatmap shows standardized beta coefficients from multivariate linear regression models. Each cell represents the association between the features (X-axis) and a specific cognitive test score (Y-axis). The color scale follows the beta value, with p-values labeled as *: p < 0.05, **: p < 0.01. SDM TOTAL: Total score of Semantic Memory test. DVT_TOTAL_RECALL: Total recall score of Delayed Visual Test. DVT_SFTANIM: Semantic Fluency Test (animal category) in Delayed Visual Test. DVT_SDM: Semantic Memory in Delayed Visual Test. DVT_RECOG_RETENTION: Recognition retention in Delayed Visual Test. DVT_RECOG_DISC_INDEX: Recognition discrimination index in Delayed Visual Test. DVT_DELAYED_RECALL: Delayed recall in Delayed Visual Test. DVSD_SDM: Semantic Memory in Delayed Visual Recognition Test. DVS_SFTANIM: Semantic Fluency Test (animal category) in Visual Spatial Test. DVS_LNS: Letter-Number Sequencing in Visual Spatial Test. 2.6. Mendelian Randomization Establishes a Causal Link for a Novel Risk Gene While our analyses established strong associations, we used Two-Sample Mendelian Randomization to infer causality. 16 Using pQTLs and eQTLs as instrumental variables, we found that among the features selected via LASSO regression, LRPPRC, WDR1, EIF3H, PLEKHH2, SLC35G2, and UTS2 had eQTL-based instruments, while MMP12 and MSMB had pQTL-based instruments. Notably, WDR1 showed an association closest to statistical significance (though still p > 0.05) with an increased risk of PDD (Table 2). This causal relationship was robust in sensitivity analyses for pleiotropy and heterogeneity (Supplementary Fig. 2). This genetic evidence elevates WDR1 from a correlative biomarker to a putative causal risk factor, providing strong rationale for its prioritization in future functional studies. Table 2 Mendelian randomization analysis assessing the causal effect of feature levels on the risk of PDD. Gene eQTLs/pQTLs Method Odds Ratio (95% CI) P-value WDR1 eQTLs Inverse Variance Weighted 1.730 (0.938–3.190) 0.079 MR Egger 1.791 (0.520–6.163) 0.377 Weighted Median 1.929 (0.861–4.322) 0.110 Weighted Mode 2.015 (0.832–4.881) 0.149 Simple Mode 2.096 (0.585–7.511) 0.280 2.7. Single-nucleus Transcriptomics Uncovers Altered Cellular Ecology and Validates Key Targets in the PDD Brain After identifying predictive biomarkers in the periphery, we aimed to pinpoint their cellular origin within the pathological brain environment. Single-nucleus RNA sequencing of postmortem prefrontal cortex provided spatially resolved validation. 29,30 Integrated transcriptomic analysis revealed a distinct global shift in the cellular landscape of PDD (Fig. 7A-B, Supplementary Fig. 3A-B). We successfully resolved all major brain cell types, including excitatory neurons (35%), inhibitory neurons (22%), oligodendrocytes (26%), astrocytes (9%), microglia (3%), OPCs (4%), and endothelial cells (1%) (Supplementary Fig. 3,4C-D). Beyond global transcriptomic changes, we quantified significant alterations in cellular proportions, uncovering a profound dysregulation of the brain's cellular ecology in PDD. Most notably, we observed a marked reduction in oligodendrocytes concurrently with a significant increase in the proportion of microglia. This finding suggests that the pathophysiological process of PDD involves concurrent mechanisms of demyelination/white matter integrity loss and sustained neuroinflammation. Crucially, we then sought to anchor our top causal candidates from prior analyses to their specific cellular context, confirming their altered expression across multiple cell types in PDD (Fig. 7C). Additionally, CellChat further delineated the cell-cell interaction in the PDD group (Fig. 7D). This cell type-specific validation directly links our peripheral and genetic findings to the relevant cellular substrate in the brain, positioning dysfunction within excitatory neurons and inhibitory neurons as a key event in the PDD pathological cascade, potentially underlying the observed "connectopathy". (A) Integrated UMAP visualization of all nuclei from PDD and control (CTRL) cohorts. (B) UMAP plot colored by clinical diagnosis. (C) Heatmap of differential expression of predictor genes in different cell types between PDD and CTRL. (D) CellChat analysis of PDD and CTRL. 2.8. From Mechanism to Therapy: Proposing Drug Repurposing Candidates The ultimate goal of our translational pipeline was to identify potential interventions. Computational drug-target interaction analysis of feature genes selected from LASSO regression identified several promising candidates for repurposing 31 , including MEDRONIC ACID, ZILEUTON and ATENOLOL etc. (Supplementary Table 1). PheWAS analysis further depict the gene associated profiles of these candidate drug targets (Supplementary Fig. 5). This analysis directly translates our mechanistic insights into tangible therapeutic hypotheses, offering a shortcut to clinical testing for this devastating condition. 3. Discussion This study employed a multi-omics approach to identify predictive factors for Parkinson's disease with mild cognitive impairment (PD-MCI), integrating transcriptomic and proteomic data to test models and gain insights based on relevant biological markers. While further model refinement is necessary for future clinical translation, this work explores the practical application value of the findings, given the early presence of omics expression differences in these patient populations. Notably, the machine learning model established after screening differential features using LASSO regression, while improving prediction performance over models using only basic demographic factors (sex, age, education years), demonstrated limited enhancement. Transcriptomic features in this study contributed more to performance improvement compared to proteomic features. Interestingly, integrating transcriptomic with proteomic features did not yield a superior predictive effect but rather a performance compromise between the two. We posit several potential reasons for this. Firstly, the core reason for the weaker performance of proteomics may lie in the dynamic nature of proteins and detection characteristics. Early pathology underlying PD-MCI conversion might initially manifest as gene activation or suppression at the transcriptome level, with corresponding protein expression potentially lagging by weeks or even months. Furthermore, the longer half-life of proteins might hinder their ability to fully reflect PD-MCI conversion risk at baseline. Secondly, the narrower dynamic range and lower diversity of proteomic detection could lead to the masking of low-abundance key molecules, possibly rendering its detection capability relatively limited compared to transcriptomics. Additionally, the relatively limited sample size in this study further constrained effective quantification of low-abundance molecules by proteomics, where stability and data noise issues might be more pronounced. The compromised effect observed upon direct integration of the two omics datasets may primarily stem from signal conflict and data heterogeneity. Differences in the temporal scales of the two omics layers, coupled with the potential risk of overfitting in the transcriptomic data, might have introduced underlying interference and counteraction, preventing the anticipated complementary synergy. SHAP analysis indicated that age at PD diagnosis was a key predictor for MCI, consistent with previous studies, underscoring the importance of timely intervention, particularly in high-risk patients. Interestingly, longer education duration emerged as a protective factor for cognition in our study, possibly because higher cognitive reserve offers a greater buffer against decline. Furthermore, our study confirmed sex as a meaningful predictive variable, with males more frequently identified as MCI patients, suggesting sex-specific differences in cognitive impairment. Among the transcriptomic genes, PLEKHH2 demonstrated the highest predictive weight. Previous research has shown that PLEKHH2 is highly enriched in glomerular podocytes, where it stabilizes the cortical actin cytoskeleton by attenuating actin depolymerization. 32 The relationship between renal function and PD has garnered increasing research interest in recent years====. In oncology, PLEKHH2 has been found involved in ALK fusions in mesenchymal tumors such as epithelioid fibrous histiocytoma, dermatofibrosarcoma protuberans, and lung adenocarcinoma 33–35 . In non-small cell lung cancer, it exhibits enhanced cytoplasmic expression and significantly promotes tumor cell proliferation, migration, and invasion 36 . In our context, PLEKHH2 was significantly correlated with multiple test scores in visuospatial and delayed visual memory tests. Single-cell RNA data from PDD patients revealed its RNA expression was elevated in oligodendrocytes, microglia, and excitatory neurons, but downregulated in inhibitory neurons. These findings collectively suggest a potential role for PLEKHH2 in processes related to the extracellular matrix and cellular dynamics within the brain. Among proteins, CA1 (Carbonic Anhydrase 1) exhibited the highest predictive weight. The closely related CA3 was also among the significantly differentially expressed proteins. As a popular drug target, CA1 already has several classic clinically available drugs. Two FDA-approved, repurposed carbonic anhydrase inhibitors, acetazolamide and methazolamide, were found to correct proteomic abnormalities and effectively prevent early molecular pathologies related to cerebral amyloid angiopathy (CAA), maintain synaptic stability, and suppress neuroinflammation in Tg-SwDI mice carrying pathogenic mutations and exhibiting Alzheimer's disease (AD)-like cognitive deficits and severe CAA 37 . Pathway enrichment analyses of features significantly different between groups in both proteomics and transcriptomics pointed to alterations in key functions like axon guidance and synaptic transmission. Notably, the gene sets corresponding to these significant features from the two omics layers showed no overlap. Trajectory clustering analysis of proteomic data and correlation analysis with clinical indicators further suggested potential functional impacts of significant upregulation or downregulation of the relevant proteins. Due to the lack of GWAS data for PD-MCI and single-cell sequencing data specifically for PD-MCI, we utilized relevant data from PDD patients versus healthy controls for further validation of the machine learning-derived features, revealing causal genetic associations and enrichment patterns across different cell types. Finally, using the screened genes, we preliminarily identified potential drug candidates, which might offer insights for future therapeutic interventions in PD-MCI or PDD. Machine learning models demonstrated depth in identifying predictive features beyond traditional statistical methods, particularly in capturing complex non-linear relationships between variables. Based on baseline omics data, this study identified several robust predictors through multivariate association analysis. These biomarkers hold potential application value in precision medicine and routine clinical practice. Our findings may provide clinicians with novel decision-support tools, aiding in the early identification of high-risk patients and facilitating the clinical translation of timely intervention and personalized management strategies. However, this study has several limitations. First, the population representativeness of the study cohort is limited, and the availability of some key variables was uneven across sub-cohorts, potentially affecting the model's generalizability. Secondly, constrained by sample size and patient heterogeneity, the predictive stability of the model across different clinical settings requires further improvement. Furthermore, although the model showed good predictive performance in cross-study validation, a gap remains before practical clinical application, necessitating further optimization and validation through large-scale prospective studies. Despite these challenges, the machine learning approach based on multi-omics data lays an important foundation for building highly practical, digitally deployable predictive models for cognitive impairment in PD. Future research should focus on expanding, optimizing, and validating these predictive markers in broader and more diverse populations, while actively exploring their potential for guiding early intervention strategies, thereby advancing precision medicine in the field of neurodegenerative diseases. 4. Conclusion In conclusion, via a tightly integrated, multi-stage analysis, we have constructed an early diagnostic model for PD-MCI (from PD) and elucidated its pathological mechanisms using multiple approaches. Besides, we genetically linking biological predictors to PDD as causal risk factors, and have identified novel, high-priority targets for therapeutic development. Ultimately, this work offers a comprehensive resource guiding the transition from understanding PD-MCI mechanisms to targeted intervention for PD patients at riks of progressing to PD-MCI. 31 5. Materials and Methods 5.1. Study Cohort and Participants Data used in this study were obtained from the Parkinson's Progression Markers Initiative (PPMI) database ( www.ppmi-info.org/access-data-specimens/download-data , RRID:SCR_006431). 38 PPMI is an ongoing international, multicenter, prospective observational study designed to identify biomarkers of PD progression. 39,40 For this analysis, we included de-identified data from PD patients. All participants provided written informed consent, and the study was approved by the institutional review board at each participating site. 39,41 The primary analysis cohort consisted of PD patients with available baseline plasma RNA-seq data, baseline CSF proteomics data, and complete longitudinal cognitive assessments over 4 years. 31,42 Specifically, from the initial 4115 PPMI participants, we focused on 1454 PD patients. After applying inclusion criteria (availability of paired multi-omics data and 4-year cognitive follow-up), 322 patients constituted the final analysis cohort ( Fig. 1 A ). According to the Movement Disorder Society Level II criteria 43 , the Montreal Cognitive Assessment (MoCA), using a cutoff score of 26, was used as the primary cognitive test to define cognitive progression and diagnose PD-MCI. 44 Other cognitive tests were utilized to validate biomarker stability. 43 Detailed inclusion and exclusion criteria, as well as the participant selection flowchart, are provided in Fig. 1 A. Baseline demographic and clinical characteristics of the included participants are summarized in Table 1. 45 5.2. Multi-omics Data Acquisition and Preprocessing 5.2.1. Plasma RNA-seq According to the PPMI sequencing protocol, total RNA was extracted from plasma samples. Libraries were prepared and sequenced on an Illumina platform. 46 Raw sequencing reads (FASTQ files) were quality-controlled using FastQC (v0.11.9) and trimmed using Trimmomatic (v0.39). 47 Reads were aligned to the human reference genome (GRCh38) using STAR (v2.7.10a) 48 , and gene counts were quantified using featureCounts (v2.0.1). 49 Counts were normalized and transformed using the varianceStabilizingTransformation function in the DESeq2 R package (v1.34.0) 50 . 5.2.2. CSF Proteomics CSF samples were analyzed using SomaLogic’s SOMAscan platform, with the target molecules being protein biomarkers based on the Slow Off-rate Modified Aptamer (SOMAmer) technology. SOMAscan data were processed following SomaLogic's standard workflow "HybNormPlateScaleCal",which includes four core steps: Hybridization Normalization, Plate Scaling, Intra-plate Median Signal Normalization, and Calibration. The data processed through this workflow were used as analyzable raw data 51 . Protein intensities were log2-transformed, and quantile normalization was performed to correct for inter-sample variation . Batch effects were adjusted for using the ComBat function from the sva R package. 51,52 5.3. Statistical and Bioinformatic Analysis 5.3.1. Differential Expression Analysis Differential analysis of RNA-seq and proteomics data between progressors and non-progressors was performed using linear models in the limma R package (v3.50.3) 53,54 . The false discovery rate (FDR) 55 was controlled using the Benjamini-Hochberg method. Significance was set at |log2(fold change)| >0.3 and adjusted p-value < 0.1. 56,57 5.3.2. Functional Enrichment Analysis Gene Ontology (GO) biological process enrichment analysis was conducted on significantly dysregulated genes and proteins using the clusterProfiler R package (v4.2.2). 23,58,59 Terms with an FDR-adjusted p-value < 0.05 were considered significant. 60,61,62 5.3.3. Interaction Network Analysis A gene-gene interaction network was constructed using the genemania database. A protein-protein interaction (PPI) network was constructed using the STRING database (v11.5). 63,64,65 5.3.4. Longitudinal Proteomic Trajectories Analysis To delineate the temporal dynamics of protein expression in the progression of Parkinson's disease with mild cognitive impairment (PD-MCI) 3 , we performed trajectory modeling and unsupervised clustering. Protein expression levels were first standardized into Z-scores relative to a matched control group to mitigate inter-individual variability. The longitudinal trajectories of these Z-scores across the years leading to diagnosis were then modeled using Locally Estimated Scatterplot Smoothing (LOESS) regression, providing a continuous view of protein fluctuations. 66 Subsequently, we conducted unsupervised hierarchical clustering based on the Euclidean distance between the LOESS-predicted trajectories to group proteins with similar temporal patterns. The Ward.D2 method was employed as the agglomeration criterion to minimize within-cluster variance. This analysis robustly categorized the proteins into five distinct clusters. A protein was defined as dysregulated if its absolute Z-score exceeded the threshold of 0.3, as visualized in the trajectory heatmap. 5.3.5. Machine Learning Modeling Least absolute shrinkage and selection operator (LASSO) regression was applied for feature selection with optimal lambda. 67,68 The resulting features were used to train and compare three machine learning classifiers: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Logistic Regression (LR). 69,70,71 Model performance was evaluated using repeated (n = 5) 5-fold cross-validation. 72,73 The area under the receiver operating characteristic curve (AUC) was used as the primary metric for model comparison and selection. 45 The final model was interpreted using SHapley Additive exPlanations (SHAP) to assess the contribution of each feature to the predictions. 38,74,75 5.3.6. Clinical Correlation Analysis To evaluate the clinical relevance of the identified multi-omics features, we assessed their associations with key neuropsychological assessments and demographic factors. We performed multivariate linear regression analyses, treating the cognitive scores from a battery of tests as dependent variables. The independent variables comprised the baseline levels of candidate proteins and transcripts identified in differential expression analysis and LASSO regression,and all models were adjusted for critical covariates, including age and education years, to control for their potential confounding effects. 76 The results are presented as standardized Beta coefficients, quantifying the strength and direction of the associations. A positive Beta indicates that a higher omics feature level is associated with a better cognitive performance, while a negative Beta suggests an association with poorer cognitive function. 5.3.7. Mendelian Randomization (MR) Protein quantitative trait locis (pQTLs) and expression quantitative trait loci (eQTLs) were sourced from the UKB-PPP and the eQTLGen Consortium. 77 Summary statistics for the outcome (PDD) were obtained from the FinnGen consortium (R9 release, finn-b-PD_DEMENTIA). 45,78 The TwoSampleMR R package (v0.5.7) 79 was used for MR analyses. Inverse-variance weighted (IVW) was the primary method, supplemented by MR-Egger, weighted median, and weighted mode methods. 20,80 Sensitivity analyses included Cochran's Q test for heterogeneity, the MR-Egger intercept test for horizontal pleiotropy, and leave-one-out analysis. 80,81 5.3.8. Single-nucleus RNA Sequencing (snRNA-seq) Analysis snRNA-seq data from GEO303823 82 were already processed using the Cell Ranger pipeline (v7.1.0). 76,83 Downstream analysis, including quality control, normalization, integration, clustering, and cell type annotation, was performed in Seurat (v4.3.0). 84,85 Differential expression and visualization of candidate genes were conducted. 78,79,20,81 For each cell type (those containing at least 3 cells in both groups) and gene combination (among the selected genes with expression records), the log2 fold change was calculated and the non-parametric Wilcoxon rank-sum test was employed to assess the significance of expression differences for each gene between the groups. The obtained raw p-values were adjusted for false discovery rate (FDR) using the Benjamini-Hochberg method. The differential expression results were visualized using a heatmap. Cell-cell communication analysis was performed to depict cell–cell interactions using CellChat and CellChatDB.human as a reference. 5.3.9. Drug-Target Interaction and Repurposing Analysis The DrugBank database (v5.1.9) was queried to identify known and investigational drugs targeting the prioritized hub proteins. Further, the DGIdb database was used for comprehensive drug-gene interaction mining. 86,87 Declarations Competing interests: All authors declare no financial or non-financial competing interests. Data Availability and Code Availability Processed data supporting the findings of this study are available within the paper and its supplementary information. Raw PPMI data are available upon registration and application through the PPMI website [https://www.ppmi-info.org/]. Analysis code will be made publicly available on GitHub upon publication. 39 Ethics Declaration The PPMI study was conducted in accordance with the Declaration of Helsinki and was approved by the institutional review board of all participating sites. All participants provided written informed consent. 39 Ethical approval for the use of postmortem snRNA-seq data was obtained by the original data providers, and this study complied with all terms of use. 39,82 Acknowledgement The authors would like to express their sincere gratitude to Wenbo Ji and Rui Zhang for their invaluable assistance in data collection and technical support during the course of this research. Author contributions AA : Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Visualization. YG : Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Visualization. LW : Investigation, Resources, Data Curation, Writing - Review & Editing. WX : Methodology, Software, Validation, Investigation. QN : Resources, Supervision, Project administration. LS : Software, Data Curation, Visualization. XF : Resources, Supervision. DH : Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration. JC : Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration. RX : Conceptualization, Methodology, Writing - Review & Editing, Supervision, Project administration. YY : Conceptualization, Resources, Writing - Review & Editing, Supervision, Project administration, Funding acquisition. Funding This work was supported by the National Natural Science Foundation of China, China (82471211), the Science and Technology Innovation Action Plan of the Shanghai Science and Technology Commission, China (23Y11906600). New Quality Clinical Specialties of High-end Medical Disciplinary Construction in Pudong New Area(2024-PWXZ-17) Key Disciplines of Shanghai East Hospital(2024-DFZD-003) Key Discipline of Shanghai Health System(2024ZDXK0002) References Assessment of Associations and Interventions in Movement Disorders: NSAID Use and Risk of Parkinsonian Markers in Genetic At-Risk Populations, and Examining Deep Brain Stimulation Surgery in Isolated and Acquired Dystonia - ProQuest. https://www.proquest.com/docview/3217274591. Postuma, R. B., Berg, D. & Stern, M. MDS clinical diagnostic criteria for Parkinson’s disease. Mov Disord 30 , 1591–1601 (2015). Abstract - 2020 - Movement Disorders - Wiley Online Library. https://movementdisorders.onlinelibrary.wiley.com/doi/10.1002/mds.28268. Feigin, V. L., Abajobir, A. A. & Abate, K. H. Global, regional, and national burden of neurological disorders during 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet Neurol 16 , 877–897 (2017). Aarsland, D. et al. Parkinson disease-associated cognitive impairment. Nat. Rev. Dis. Primer 7 , 47 (2021). Martinez-Eguiluz, M. et al. Machine Learning for Prediction of Cognitive Deterioration in Patients with Early Parkinson’s Disease. Appl. Sci. 14 , 8149 (2024). Hely, M. A., Reid, W. G. J., Adena, M. A., Halliday, G. M. & Morris, J. G. L. The Sydney multicenter study of Parkinson’s disease: the inevitability of dementia at 20 years. Mov. Disord. Off. J. Mov. Disord. Soc. 23 , 837–844 (2008). Weintraub, D. et al. Management of psychiatric and cognitive complications in Parkinson’s disease. BMJ 379 , e068718 (2022). Role of Gut-Liver-Brain Axis in Neuropsychiatric Symptoms of Parkinson’s Disease - ProQuest. https://www.proquest.com/docview/3217835689. He, P. et al. The association of CSF biomarkers and cognitive decline with choroid plexus volume in early Parkinson’s disease. Parkinsonism Relat. Disord. 120 , (2024). Irwin, D. J., Grossman, M. & Weintraub, D. Neuropathological and genetic correlates of survival and dementia onset in synucleinopathies: a retrospective analysis. Lancet Neurol 16 , 55–65 (2017). Harvey, J., Rekkas, V. & Svenningsson, P. Machine learning-based prediction of cognitive outcomes in de novo Parkinson’s disease. NPJ Park. Dis 8 , (2022). Karayel, O., Virreira Winter, S. & Padmanabhan, S. Proteome profiling of cerebrospinal fluid reveals biomarker candidates for Parkinson’s disease. Cell Rep Med 3 , (2022). Nalls, M. A., Blauwendraat, C. & Vallerga, C. L. Identification of novel risk loci, causal insights, and heritable risk for Parkinson’s disease: a meta-analysis of genome-wide association studies. Lancet Neurol 18 , 1091–1102 (2019). Liu, G., Peng, J. & Liao, Z. Genome-wide survival study identifies a novel synaptic locus and polygenic score for cognitive progression in Parkinson’s disease. Nat Genet 53 , 787–793 (2021). Sun, B. B., Chiou, J. & Traylor, M. Plasma proteomic associations with genetics and health in the UK Biobank. Nature 622 , 329–338 (2023). Cui, M., Cheng, C. & Zhang, L. High-throughput proteomics: a methodological mini-review. Lab. Investig. J. Tech. Methods Pathol. 102 , 1170–1181 (2022). Systematic Review of Proteomics in Age-Related Macular Degeneration and Pathway Analysis of Significant Protein Changes - Ophthalmology Science. https://www.ophthalmologyscience.org/article/S2666-9145(25)00091-0/fulltext. Chia, R. et al. A plasma proteomics-based candidate biomarker panel predictive of amyotrophic lateral sclerosis. Nat. Med. https://doi.org/10.1038/s41591-025-03890-6 (2025) doi:10.1038/s41591-025-03890-6. Burgess, S. & Thompson, S. G. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur. J. Epidemiol. 32 , 377–389 (2017). Zhao, W. et al. Risk associated circulating biomarkers S100A3 identified in congenital heart disease-associated pulmonary arterial hypertension. Intractable Rare Dis. Res. 14 , 36–45 (2025). Mapping the dynamics of epigenetic adaptation | bioRxiv. https://www.biorxiv.org/content/10.1101/2023.07.10.548368v1.full. A Population-scale Single-cell Spatial Transcriptomic Atlas of the Human Cortex | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.10.13.681959v1. Wang, J. et al. Impact of Y chromosome loss on the risk of Parkinson’s disease and progression. eBioMedicine 117 , (2025). Jeong, D., Koo, B., Oh, M., Kim, T.-B. & Kim, S. GOAT: Gene-level biomarker discovery from multi-Omics data using graph ATtention neural network for eosinophilic asthma subtype. Bioinformatics 39 , btad582 (2023). Development and Validation of PARCOMS Composite Scales for Assessing Disease Progression and Treatment Effects in Parkinson’s Disease | Neurology and Therapy. https://link.springer.com/article/10.1007/s40120-025-00771-5. Sood, M., Ktena, S. I. & Wenzel, F. Generating synthetic patient data using Bayesian networks. J Am Med Inf. Assoc 27 , 1255–1267 (2020). Wang, Z. et al. Cerebrospinal fluid proteomics identification of biomarkers for amyloid and tau PET stages. Cell Rep. Med. 6 , 102031 (2025). Habib, N. et al. Massively parallel single-nucleus RNA-seq with DroNc-seq. Nat. Methods 14 , 955–958 (2017). Jiang, Z. Q., Wu, J. J. & Liu, J. Characterization of a pathogenic variant in GBA for Parkinson’s disease with mild cognitive impairment patients. Mol Brain 13 , (2020). Walker, K. A., Chen, J. & Zhang, J. Proteomics analysis of plasma from middle-aged adults identifies protein markers of dementia risk in later life. Sci Transl Med 15 , (2023). Perisic, L. et al. Plekhh2, a novel podocyte protein downregulated in human focal segmental glomerulosclerosis, is involved in matrix adhesion and actin dynamics. Kidney Int. 82 , 1071–1083 (2012). DeSimone, M. S. et al. Epithelioid Fibrous Histiocytoma Is on a Continuum With Superficial ALK -rearranged Myxoid Spindle Cell Neoplasm : A Clinicopathologic Series of 35 Cases Including Alternate RET and NTRK3 Fusions. Am. J. Surg. Pathol. 48 , 1568–1579 (2024). Agrawal, S. et al. ALK-rearranged, CD34-positive spindle cell neoplasms resembling dermatofibrosarcoma protuberans: a study of seven cases. Histopathology 85 , 649–659 (2024). Coppock, J. D. et al. S100 and CD34 Expressing Mesenchymal Neoplasm With Rare PLEKHH2::ALK Fusion and Response to ALK Inhibition. Am. J. Surg. Pathol. 46 , 1309–1313 (2022). Wang, R. et al. PLEKHH2 binds β-arrestin1 through its FERM domain, activates FAK/PI3K/AKT phosphorylation, and promotes the malignant phenotype of non-small cell lung cancer. Cell Death Dis. 13 , 858 (2022). Carlsen, J., Fossati, S., Østergaard, L., Gutiérrez-Jiménez, E. & Palmfeldt, J. Cerebral proteome adaptations to amyloid angiopathy are prevented by carbonic anhydrase inhibitors. Alzheimers Dement. J. Alzheimers Assoc. 21 , e70122 (2025). Fabrizio, C., Termine, A. & Caltagirone, C. Transcriptomics profiling of Parkinson’s disease progression subtypes reveals distinctive patterns of gene expression. 2023.10.11.23296884 Preprint at https://doi.org/10.1101/2023.10.11.23296884 (2023). Marek, K., Chowdhury, S. & Siderowf, A. The Parkinson’s progression markers initiative (PPMI) - establishing a PD biomarker cohort. Ann Clin Transl Neurol 5 , 1460–1477 (2018). Associations between Sleep Disorders and Impulsive-Compulsive Behaviors in Parkinson’s Disease: A Prospective Cohort Study | Neuroepidemiology | Karger Publishers. https://karger.com/ned/article/58/4/256/895414/Associations-between-Sleep-Disorders-and-Impulsive. Machine learning-based classification of dual fluorescence signals reveals muscle stem cell fate transitions in response to regenerative niche factors | npj Regenerative Medicine. https://www.nature.com/articles/s41536-023-00277-4. Loo, R. T. J. et al. Multi-cohort machine learning identifies predictors of cognitive impairment in Parkinson’s disease. NPJ Digit. Med. 8 , 482 (2025). Litvan, I. et al. Diagnostic criteria for mild cognitive impairment in Parkinson’s disease: Movement Disorder Society Task Force guidelines. Mov. Disord. Off. J. Mov. Disord. Soc. 27 , 349–356 (2012). Xing, Y. et al. Regional Neural Activity Changes in Parkinson’s Disease-Associated Mild Cognitive Impairment and Cognitively Normal Patients. Neuropsychiatr. Dis. Treat. 17 , 2697–2706 (2021). Frontiers | Machine learning-based prediction of carotid intima–media thickness progression: a three-year prospective cohort study. https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1593662/full. Guduric-Fuchs, J. et al. Selective extracellular vesicle-mediated export of an overlapping set of microRNAs from multiple cell types. BMC Genomics 13 , 357 (2012). Mapping the dynamics of epigenetic adaptation | bioRxiv. https://www.biorxiv.org/content/10.1101/2023.07.10.548368v1.full. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinforma. Oxf. Engl. 29 , 15–21 (2013). Epigenomic and transcriptomic analyses reveal cnidocyte specialization in a sea anemone | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.10.06.680633v1. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15 , 550 (2014). Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostat. Oxf. Engl. 8 , 118–127 (2007). Full article: Copper associates with differential methylation in placentae from two US birth cohorts. https://www.tandfonline.com/doi/full/10.1080/15592294.2019.1661211. Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43 , e47 (2015). Changes in the miRNA-mRNA Regulatory Network Precede Motor Symptoms in a Mouse Model of Multiple System Atrophy: Clinical Implications | PLOS One. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150705. Glickman, M. E., Rao, S. R. & Schultz, M. R. False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies. J. Clin. Epidemiol. 67 , 850–857 (2014). CSF proteomics in autosomal dominant Alzheimer’s disease highlights parallels with sporadic disease | Brain | Oxford Academic. https://academic.oup.com/brain/article/146/11/4495/7205393. DUX4-stimulated genes define an antiviral defense program in human placental trophoblasts | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.01.22.634317v2. Yu, G., Wang, L.-G., Han, Y. & He, Q.-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics J. Integr. Biol. 16 , 284–287 (2012). Ashburner, M. et al. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet. 25 , 25–29 (2000). Kanehisa, M., Goto, S., Furumichi, M., Tanabe, M. & Hirakawa, M. KEGG for representation and analysis of molecular networks involving diseases and drugs. Nucleic Acids Res 38 , (2010). Milacic, M., Beavers, D. & Conley, P. The Reactome Pathway Knowledgebase 2024. Nucleic Acids Res 52 , (2024). Li, J. et al. Functional study of RNA alternative splicing in zebrafish adaptation to low temperature. Comp. Biochem. Physiol. Part D Genomics Proteomics 101656 (2025) doi:10.1016/j.cbd.2025.101656. Shannon, P., Markiel, A. & Ozier, O. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13 , 2498–2504 (2003). Szklarczyk, D. et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic Acids Res. 51 , D638–D646 (2023). Oh, H. et al. Cytokine-Related Genes and Inflammatory Profiles as Potential Biomarkers in Major Depressive Disorder. Psychiatry Investig. 22 , 858–869 (2025). Gan, Y.-H. et al. Large-scale proteomic analyses of incident Parkinson’s disease reveal new pathophysiological insights and potential biomarkers. Nat. Aging 5 , 642–657 (2025). Friedman, J., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J. Stat. Softw. 33 , 1–22 (2010). Identifying pyroptosis-related genes as novel therapeutic targets in diabetic foot ulceration | Diabetology & Metabolic Syndrome | Full Text. https://dmsjournal.biomedcentral.com/articles/10.1186/s13098-025-01880-9. Breiman, L. Random Forests. Mach Learn 45 , 5–32 (2001). XGBoost | Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. https://dl.acm.org/doi/10.1145/2939672.2939785. Nega Tarekegn, A., Alaya Cheikh, F., Sajjad, M. & Ullah, M. Towards Detecting Freezing of Gait Events Using Wearable Sensors and Genetic Programming. in Artificial Intelligence and Soft Computing (eds Rutkowski, L. et al.) 274–285 (Springer Nature Switzerland, Cham, 2023). doi:10.1007/978-3-031-42505-9_24. Kuhn, M. Building predictive models in R using the caret package. J Stat Softw 28 , 1–26 (2008). Can MRI-based multivariate gray matter volumetric distance predict motor progression and classify slow versus fast progressors in Parkinson’s disease? | medRxiv. https://www.medrxiv.org/content/10.1101/2022.07.25.22278012v1. Lundberg, S. M. & Lee, S. I. A Unified Approach to Interpreting Model Predictions. in Advances in Neural Information Processing Systems 30 (NIPS vol. 2017 4765–4774 (2017). An artificial intelligence interpretable tool to predict risk of deep vein thrombosis after endovenous thermal ablation - Journal of Vascular Surgery: Venous and Lymphatic Disorders. https://www.jvsvenous.org/article/S2213-333X(25)00088-5/fulltext. Zhong, J., Li, M., Dai, Z. & Wan, J. Multi-omics reveals changes in astrocyte fatty acid metabolism during early stages of Alzheimer’s disease. Preprint at https://doi.org/10.21203/rs.3.rs-4976800/v1 (2024). Abstract - 2020 - Movement Disorders - Wiley Online Library. https://movementdisorders.onlinelibrary.wiley.com/doi/10.1002/mds.28268. Kurki, M. I., Karjalainen, J. & Palta, P. FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 613 , 508–518 (2023). Hemani, G., Zheng, J. & Elsworth, B. The MR-Base platform supports systematic causal inference across the human phenome. Elife 7:e34408 , (2018). Causal association of circulating immune cells and lymphoma: A Mendelian randomization study. https://www.degruyterbrill.com/document/doi/10.1515/med-2024-0984/html. Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50 , 693–698 (2018). Chia, S.-Y. et al. Single-nucleus transcriptomics reveals a distinct microglial state and increased MSR1-mediated phagocytosis as common features across dementia subtypes. Genome Med. 17 , 92 (2025). BRAIN Initiative Cell Census Network (BICCN). A multimodal cell census and atlas of the mammalian primary motor cortex. Nature 598 , 86–102 (2021). Zheng, G. X. Y. et al. Massively parallel digital transcriptional profiling of single cells. Nat. Commun. 8 , 14049 (2017). Selenoprotein P Deficiency Drives Hepatocellular Carcinoma Progression via Induction of Neutrophil Senescence and Immunosuppressive Microenvironment | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.06.24.661430v1. Wishart, D. S. et al. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 46 , D1074–D1082 (2018). Nowell, J., Blunt, E., Gupta, D. & Edison, P. Antidiabetic agents as a novel treatment for Alzheimer’s and Parkinson’s disease. Ageing Res Rev 89 , (2023). Additional Declarations No competing interests reported. Supplementary Files Supplementarymaterial.docx Cite Share Download PDF Status: Posted Version 1 posted 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-7937441","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":545840060,"identity":"5d18546d-1f8d-4aab-8b7d-bf7ffbcf9606","order_by":0,"name":"Aydos Alemjan","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Aydos","middleName":"","lastName":"Alemjan","suffix":""},{"id":545840062,"identity":"c6a9cde7-e074-4bf9-934a-7586fe5cbca2","order_by":1,"name":"Yuankai Gu","email":"","orcid":"","institution":"Changhai Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yuankai","middleName":"","lastName":"Gu","suffix":""},{"id":545840064,"identity":"f2ad62b7-6508-484c-aee7-4c51a1ff1449","order_by":2,"name":"Li Wang","email":"","orcid":"","institution":"Shanghai Changzheng Hospital","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Wang","suffix":""},{"id":545840065,"identity":"7288d8ec-6aaa-4134-a818-6c988a814c84","order_by":3,"name":"Wei Xu","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Xu","suffix":""},{"id":545840067,"identity":"5154e005-90eb-47ba-a95f-2d2af6ddf4e0","order_by":4,"name":"Qing Ning","email":"","orcid":"","institution":"Shanghai Changzheng Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qing","middleName":"","lastName":"Ning","suffix":""},{"id":545840068,"identity":"cd6f95a4-5d4d-47c5-ad83-bbe4b4db8468","order_by":5,"name":"Lei Shen","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Shen","suffix":""},{"id":545840069,"identity":"5d441c2d-7238-48a1-acef-5b5a5dbeefe0","order_by":6,"name":"Xuelin Feng","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xuelin","middleName":"","lastName":"Feng","suffix":""},{"id":545840070,"identity":"6c07f805-925e-4657-8701-939c8296cfda","order_by":7,"name":"Dongya Huang","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Dongya","middleName":"","lastName":"Huang","suffix":""},{"id":545840071,"identity":"691b4a59-567b-40c0-b4f8-0c36c4d245cf","order_by":8,"name":"Juxiang Chen","email":"","orcid":"","institution":"Changhai Hospital","correspondingAuthor":false,"prefix":"","firstName":"Juxiang","middleName":"","lastName":"Chen","suffix":""},{"id":545840072,"identity":"df5a767a-6c3b-4c66-ab55-9ef4f6b5f4cd","order_by":9,"name":"Ran Xiong","email":"","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":false,"prefix":"","firstName":"Ran","middleName":"","lastName":"Xiong","suffix":""},{"id":545840073,"identity":"15e82a0e-22f1-4ae6-ab8b-833d05eb504b","order_by":10,"name":"You Yin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYDACZjApUd/P3tj48AMJWiwYZ/YcbjaWIMGuCsYNN9LbBHiIUWtwnPeYxMc2CWaDmw/bGCQY7OR0GwhpOcyXJjmzTYJN8nZi24MChmRjswMEtfCYSfO2SfDw3U5sN5BgOJC4jVgtEgw3DwI1kqLFQOAGI5FaJA/zGFvOOCeRINmTCAxkAyL8wnf+jOGND2V1Cfzsxx8+/FBhJ0dQi8IBBhakCDQgoBwE5BsYmIlLJqNgFIyCUTByAQAF9kBCTEn1OQAAAABJRU5ErkJggg==","orcid":"","institution":"Shanghai East Hospital","correspondingAuthor":true,"prefix":"","firstName":"You","middleName":"","lastName":"Yin","suffix":""}],"badges":[],"createdAt":"2025-10-24 06:38:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7937441/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7937441/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96160054,"identity":"8ab4a600-8f2c-4231-8e8e-f98e4538ca95","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5889518,"visible":true,"origin":"","legend":"","description":"","filename":"FromNeuralConnectopathytoaTherapeuticPathAnIntegratedMultiomicsFrameworkIdentifiesaCausalGeneandDrugCandidatesforParkinsonsDiseasewithmildcognitiveimpairment.docx","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/34857bda26869ff52336bcb3.docx"},{"id":96249223,"identity":"0c3523d3-6083-4e0a-b544-5b352995783c","added_by":"auto","created_at":"2025-11-19 07:30:50","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12960,"visible":true,"origin":"","legend":"","description":"","filename":"c507685531b24c1886bc7b06c277e2c8.json","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/dc34758aa490946c98461f71.json"},{"id":96249963,"identity":"fc8e7d3b-b39c-432c-a472-e0a7de2e1970","added_by":"auto","created_at":"2025-11-19 07:36:53","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1487439,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/92ef6ebf5f2730b5b97c9ce1.docx"},{"id":96250235,"identity":"ce8ec07b-90d0-48ba-81b7-8e8481e4d0ab","added_by":"auto","created_at":"2025-11-19 07:37:46","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":127883,"visible":true,"origin":"","legend":"","description":"","filename":"c507685531b24c1886bc7b06c277e2c81enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/95a188171ca8a22f56b1ceef.xml"},{"id":96251001,"identity":"cd2df82b-9562-4918-8fad-14d0095fef12","added_by":"auto","created_at":"2025-11-19 07:39:12","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":345240,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/20dad58b00a76a54e6853a66.jpeg"},{"id":96160065,"identity":"59e6c244-41ea-41e3-b605-5a23d3750555","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":266555,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/acdd772b58317dab630de1ff.jpeg"},{"id":96160066,"identity":"2dc9717d-5f8d-4f24-86c8-7672f655879d","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":499199,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/f85c94d1352783c6ffecab0d.jpeg"},{"id":96252230,"identity":"6f0f36d3-e00e-44e7-a9c3-489743b2793f","added_by":"auto","created_at":"2025-11-19 07:40:41","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":137854,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/85a23c8132c50df0cf2db8bd.png"},{"id":96249175,"identity":"9cc7de8c-c717-4c24-86e1-48243e9247fa","added_by":"auto","created_at":"2025-11-19 07:30:32","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":848972,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/88ccd3e59bd00ed4bf20b476.jpeg"},{"id":96252012,"identity":"5406e5ab-dfa1-42df-bc3b-400c3ee4a8c1","added_by":"auto","created_at":"2025-11-19 07:40:19","extension":"jpeg","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":776607,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/189c7ea6b3caecc311fb30fb.jpeg"},{"id":96160082,"identity":"8bb4b1bc-ced9-4023-9ca2-19c9ad2d1b0e","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":685310,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/1250c5d1555b2588c1501dc9.jpeg"},{"id":96160074,"identity":"a5e20ee5-56e5-4c11-834b-4108e5bfcded","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":461524,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/2bd076fb30aa15142cbf625f.jpeg"},{"id":96160075,"identity":"cf3b8e46-05de-4c11-899c-ffcfb3a6191d","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":547159,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/d1e3341c4e6dbcb886c6cb41.jpeg"},{"id":96160072,"identity":"c914f93b-cba7-43ab-b466-024577b2d1a5","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":641084,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/7bed2d0ce42658885ecba7d9.jpeg"},{"id":96160081,"identity":"71904b8e-62f2-40f8-9992-b52f47ea98e2","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":59542,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/785a6c5edbfc99de37b5a9e9.png"},{"id":96160076,"identity":"578c7708-db56-422d-bd34-9e0837ace78d","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"jpeg","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":489844,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/1e9c8fbde92679e87584e9a2.jpeg"},{"id":96250047,"identity":"d6df1458-a176-4039-b432-86c93846d3bb","added_by":"auto","created_at":"2025-11-19 07:37:12","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":164513,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/9f3c9781ba7130b5ef9d6afa.png"},{"id":96250266,"identity":"1cde5f92-40cd-4a93-b667-6e6323ebeaa3","added_by":"auto","created_at":"2025-11-19 07:37:54","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":79370,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/67af0ff2f1d7be80d3974a13.png"},{"id":96160091,"identity":"e8cb15c6-7128-499b-9bdc-f1d17223e91c","added_by":"auto","created_at":"2025-11-18 08:46:40","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":193906,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/63f8aedcb397cdc80c51ffca.png"},{"id":96160068,"identity":"9b53732f-9416-4ea7-99d1-5a97ba486b1d","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31116,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/deda67fce5ff3ab1a1e03e16.png"},{"id":96160092,"identity":"a8dc2df9-ed8a-4ed2-a1e1-c2019818fe46","added_by":"auto","created_at":"2025-11-18 08:46:40","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":196949,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/26ca5ab6adb94b355f3dfa19.png"},{"id":96160079,"identity":"335292e7-69be-499a-913e-bca2261b6e28","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":197234,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/1a61767b8d3cc34f9b27e0fd.png"},{"id":96160083,"identity":"a8e38e1e-ff0f-451c-9142-17b789ec9017","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":147807,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/c9f03d8e0082391c0302af49.png"},{"id":96160085,"identity":"e292bdba-15af-4365-94b1-ee1c15a28580","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":110659,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/e16eccdd27b8bc4f026819a9.png"},{"id":96250013,"identity":"a17fcd50-da10-485d-9cc8-d7fa36828b24","added_by":"auto","created_at":"2025-11-19 07:37:07","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149227,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/fa825d7d4deb8acf75961e9c.png"},{"id":96251063,"identity":"5793bdf8-21ef-4e96-b76d-8bc2e26f9315","added_by":"auto","created_at":"2025-11-19 07:39:16","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":216355,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/8840f48873cdaaf0d9a01889.png"},{"id":96160088,"identity":"5b6394de-3ad0-44f3-8aad-ff6de3bf28c3","added_by":"auto","created_at":"2025-11-18 08:46:40","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15324,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/b02534da9682645fc59d4ded.png"},{"id":96249388,"identity":"9e2362e1-1667-4a30-9793-98ca9a3e71a2","added_by":"auto","created_at":"2025-11-19 07:33:20","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66671,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/1a8ccc8135ebd498d099dccc.png"},{"id":96252129,"identity":"6d42a23f-8f4f-4d72-b143-4670fd67fd1b","added_by":"auto","created_at":"2025-11-19 07:40:29","extension":"xml","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":128457,"visible":true,"origin":"","legend":"","description":"","filename":"c507685531b24c1886bc7b06c277e2c81structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/68d226478546abcca8de8d6c.xml"},{"id":96160087,"identity":"d6d68f64-6193-4c68-88a5-417319aec49c","added_by":"auto","created_at":"2025-11-18 08:46:40","extension":"html","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":150185,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/2909e032243dbcad86addc16.html"},{"id":96249335,"identity":"66f0e973-de20-4030-a4ca-01df84a55470","added_by":"auto","created_at":"2025-11-19 07:33:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":410391,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStudy flowchart and integrated analytical framework for identifying/validating protein and transcriptomic biomarkers linked to cognitive decline in PD.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Participant inclusion flowchart of the PPMI cohort. From the total PPMI cohort (n=4115), we first identified PD patients (n=1454). The final analytical cohort consisted of 322 PD patients with complete baseline data (including CSF proteomics, plasma RNA-seq) and 4-year longitudinal cognitive assessments. HC: Healthy Control; SWEDD: Scans Without Evidence of Dopaminergic Deficit.\u003c/p\u003e\n\u003cp\u003e(B) Schematic of the multi-modal bioinformatic and statistical analysis pipeline. The analysis began with standard preprocessing of proteomic and transcriptomic data, followed by differential expression analysis to identify candidate biomarkers. These candidates underwent time-sequencing analysis (to explore dynamic changes), pathway enrichment analysis (to decipher biological processes), and LASSO regression (for feature selection in predictive models). These features were used to build a Machine learning model, with SHAP analysis for interpretability and clinical features correlation analysis to establish clinical relevance. Mendelian Randomization (MR) was employed to infer causality, and single-cell sequencing analysis was conducted for cellular localization; together, these steps guided the prioritization of putative Drug Targets.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/a153999678cd181bcfe0aa8d.png"},{"id":96160057,"identity":"e9bdb340-28e0-46c3-96eb-b1f3485b3412","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":817322,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIntegrated multi-omics analyses identify key molecules and pathways linked to cognitive impairment in PD.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A-B) Volcano plots of differential expression analysis comparing PD-MCI patients and those with normal cognition in 4 years. (A) Plasma RNA-seq analysis. (B) CSF proteomics analysis.. (C-D) Gene Ontology (GO) enrichment analysis of biological processes. (C) GO terms significantly enriched among significant RNA. (D) GO terms significantly enriched among significant proteins. (E-F) Protein-Protein Interaction (PPI) network of core differentially expressed proteins.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/a8325ef6106a6327105bce57.png"},{"id":96160058,"identity":"868f87a3-2319-4fec-9bdc-1b67861a9706","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":759257,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLongitudinal proteomic trajectories reveal dynamic CSF proteomic biomarker signatures preceding cognitive impairment in PD-CI.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Temporal expression patterns (Z-score normalized levels) of individual proteins (top 10 selected from LASSO). These trajectories illustrate the heterogeneous temporal dynamics, with some proteins exhibiting progressive elevation (e.g., Pro-inflammatory markers) and others showing early decline (e.g., Synaptic proteins). (B) Average protein expression trajectories for each cluster. Lines represent the mean Z-score of all proteins in a cluster, shading indicates the standard error of the mean. (C) Heatmap displaying expression patterns of all significantly differentiential protein. (D) Pie chart illustrating the proportion of dysregulated proteins across 5 trajectory clusters.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/11aec02d66f6c04fe13f48b2.png"},{"id":96251980,"identity":"65cd0f67-d350-424e-811a-82dad38a556c","added_by":"auto","created_at":"2025-11-19 07:40:16","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":742019,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConstruction and Comparative Evaluation of Machine Learning Models for Predicting Cognitive Impairment in Parkinson's Disease\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A-B) LASSO regression to select the most important biomarkers from DEG results. (A) Plasma transcripts; (B) CSF Proteomics. (C-E) Comparative performance of multiple machine learning algorithms. Box plots display the distribution of ROC-AUC score for Random Forest, XGBoost, and Logistic Regression models evaluated via 5-fold cross-validation. (C) Models based on plasma transcripts + demographics; (D) Models based on CSF proteomics + demographics; (E) Models based on plasma transcripts + CSF proteomics + demographics. (F) Overall model performance evaluation on the independent test set. ROC curves illustrate the discriminative ability of models based on plasma transcripts + CSF proteomics + demographics.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/14cfd83fb8db89a590c350e9.png"},{"id":96251137,"identity":"5bbdcb3e-d346-48e1-a215-5936c4daf64c","added_by":"auto","created_at":"2025-11-19 07:39:24","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":457251,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP Interpretation of Machine Learning Models for Predicting Cognitive Impairment in Parkinson's Disease\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A-C) SHAP Analysis for Model Interpretability. The plot ranks top features by their mean absolute SHAP value, representing their overall importance in the optimal model's predictions. (A) CSF Proteomics + demographics; (B) Plasma Transcripts + demographics; (C) CSF Proteomics + Plasma Transcripts + demographics.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/aa35b79834d478ee89c81037.png"},{"id":96160064,"identity":"994203c7-8e88-444e-b1c1-824c3ac06e2c","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":544255,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCross-sectional associations between candidate protein biomarkers and domain-specific cognitive scores.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHeatmap shows standardized beta coefficients from multivariate linear regression models. Each cell represents the association between the features (X-axis) and a specific cognitive test score (Y-axis). The color scale follows the beta value, with p-values labeled as *: p \u0026lt; 0.05, **: p \u0026lt; 0.01. SDM TOTAL: Total score of Semantic Memory test. DVT_TOTAL_RECALL: Total recall score of Delayed Visual Test. DVT_SFTANIM: Semantic Fluency Test (animal category) in Delayed Visual Test. DVT_SDM: Semantic Memory in Delayed Visual Test. DVT_RECOG_RETENTION: Recognition retention in Delayed Visual Test. DVT_RECOG_DISC_INDEX: Recognition discrimination index in Delayed Visual Test. DVT_DELAYED_RECALL: Delayed recall in Delayed Visual Test.\u003c/p\u003e\n\u003cp\u003eDVSD_SDM: Semantic Memory in Delayed Visual Recognition Test. DVS_SFTANIM: Semantic Fluency Test (animal category) in Visual Spatial Test. DVS_LNS: Letter-Number Sequencing in Visual Spatial Test.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/a1a0362864105431483db44b.png"},{"id":96250887,"identity":"089d76d7-f0b4-4c12-b1ca-35ccb54fb3a6","added_by":"auto","created_at":"2025-11-19 07:39:05","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":592080,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-nucleus RNA sequencing of human brain tissue reveals altered cellular community structure in PDD.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Integrated UMAP visualization of all nuclei from PDD and control (CTRL) cohorts. (B) UMAP plot colored by clinical diagnosis. (C) Heatmap of differential expression of predictor genes in different cell types between PDD and CTRL. (D) CellChat analysis of PDD and CTRL.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/d2ec29af9aa45a41700eb7dd.png"},{"id":98435444,"identity":"441f6b33-0bb3-4430-8a35-5473ab2dab6d","added_by":"auto","created_at":"2025-12-17 16:53:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6215072,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/462df8b9-aa6b-4580-8803-2af65c1071b3.pdf"},{"id":96160060,"identity":"2c42d9bb-9e42-4e6c-8057-c09de4617e9b","added_by":"auto","created_at":"2025-11-18 08:46:39","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1487439,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7937441/v1/68bf1713fd177edbb1fd0d55.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"From Neural Connectopathy to a Therapeutic Path: An Integrated Multi-omics Framework Identifies a Causal Gene and Drug Candidates for Parkinson's Disease with mild cognitive impairment","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eParkinson's disease (PD) is the second most common neurodegenerative disorder worldwide, characterized by both cardinal motor symptoms and numerous non-motor complications.\u003csup\u003e2\u0026ndash;4\u003c/sup\u003e Among these, Parkinson's disease with mild cognitive impairment (PD-MCI) is one of the most debilitating and burdensome, affecting up to 80% of patients within 15 years of onset.\u003csup\u003e5\u0026ndash;7\u003c/sup\u003e PD-MCI\u0026rsquo;s emergence signals severe quality-of-life decline for patients and caregivers and no effective disease-modifying therapies exist.\u003csup\u003e8,9\u003c/sup\u003e PD-MCI\u0026rsquo;s neuropathological hallmarks include the widespread alpha-synuclein aggregation (Lewy bodies), often coexisting with Alzheimer's disease-related pathologies (amyloid-β plaques, neurofibrillary tau tangles).\u003csup\u003e10\u003c/sup\u003e This indicates a complex and multifactorial etiology.\u003csup\u003e11\u003c/sup\u003e This complexity poses a major challenge: inability to accurately predict which PD patients will develop mild cognitive impairment or dementia, and incomplete understanding of underlying molecular mechanisms, hindering targeted prevention.\u003csup\u003e12\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eIn recent years, high-throughput omics technologies have offered unprecedented opportunities to unravel neurodegenerative diseases\u0026rsquo; molecular complexity.\u003csup\u003e13\u003c/sup\u003e Genome-wide association studies (GWAS) have identified many PD risk loci, but their ability to explain PD-MCI risk remains limited.\u003csup\u003e10,14,15\u003c/sup\u003e In contrast, transcriptomic and proteomic profiling of accessible biofluids (e.g., plasma, cerebrospinal fluid [CSF]) provides dynamic snapshots of pathological processes and holds promise for discovering diagnostic/prognostic biomarkers.\u003csup\u003e16\u003c/sup\u003e However, most studies to date are cross-sectional or focused on a single omics layer, failing to capture the temporal dynamics of disease progression and interactions between molecular levels.\u003csup\u003e17,18\u003c/sup\u003e A longitudinal, multi-omics approach is thus imperative to map the chronological molecular events leading to PD-MCI.\u003c/p\u003e\u003cp\u003eFurthermore, conventional association studies using bulk tissue or biofluids cannot pinpoint the specific cellular origins of dysregulated molecules, a critical step for understanding mechanism.\u003csup\u003e19\u003c/sup\u003e Additionally, while omics analyses reveal correlations, they cannot establish causality. Mendelian randomization (MR), an epidemiological method using genetic variants as instrumental variables, has become a powerful tool to infer causal relationships between an exposure (e.g., gene expression) and outcomes (e.g., PD-MCI risk), reducing confounding and reverse causation.\u003csup\u003e20,21\u003c/sup\u003e Single-nucleus RNA sequencing (snRNA-seq) now enables unbiased analysis of cell-type-specific transcriptional changes in postmortem human brain, providing key spatial context for biomarker discoveries.\u003csup\u003e20,22,23\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eTo address these gaps, we designed an integrative translational study using the large-scale, deeply phenotyped Parkinson's Progression Markers Initiative (PPMI) cohort.\u003csup\u003e24\u003c/sup\u003e We hypothesized that integrated multi-omics analysis (plasma transcriptomics\u0026thinsp;+\u0026thinsp;CSF proteomics) could identify core molecules/pathways driving PD-related cognitive decline. We further proposed that these findings could be integrated into a PD-MCI risk predictive model, tested for causality using MR, and validated in specific brain cell types via snRNA-seq.\u0026nbsp;Finally, we aimed to translate these discoveries into therapeutic hypotheses through computational drug repurposing.\u003c/p\u003e"},{"header":"2. Results","content":"\u003cdiv id=\"Sec3\"\u003e\n \u003ch2\u003e2.1. Cohort Selection and an Integrative Analytical Framework for PD-MCI\u003c/h2\u003e\n \u003cp\u003eTo systematically analyze the molecular mechanisms of PD-MCI, we used the longitudinal PPMI cohort. From 4115 initial participants, we selected 322 PD patients as the final analysis cohort, all of which with complete baseline multi-omics data (plasma RNA-seq and CSF proteomics) and 4-year longitudinal cognitive assessments (MoCA) (Fig. 1A, Table 1). We then designed and implemented a comprehensive analytical workflow, moving beyond simple association analysis to focus on predictive model development and clinical application. This framework integrated differential expression analysis, machine learning-based prediction, genetic causal inference, single-cell validation, and drug repurposing (Fig. 1B), establishing a translational research pipeline from biomarker discovery to therapeutic hypothesis.\u003c/p\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 1\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eBaseline characteristics of participants from the Parkinson's Progression Markers Initiative (PPMI) cohort, stratified by PD-MCI outcome.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePD without MCI (n = 213)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePD-MCI (n = 207)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal (n = 420)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ep-value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRNA-seq\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.5114†\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e32 (15.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37 (17.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e69 (16.4%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e181 (85.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e170 (82.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e351 (83.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProteomics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.2170†\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11 (5.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e18 (8.7%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e29 (6.9%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e202 (94.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e189 (91.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e391 (93.1%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0143†\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e98 (46.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70 (33.8%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e168 (40.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e115 (54.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e137 (66.2%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e252 (60.0%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e58.2 (51.7–64.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e66.5 (59.8–70.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e62.2 (55.1–68.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt; 0.001¶\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEducation Years\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16.0 (14.0–18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16.0 (13.5–18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16.0 (14.0–18.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0646¶\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eREM hours\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.0 (2.0–5.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.0 (2.0–7.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.0 (2.0–6.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.0285¶\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePD Diagnosis Age\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e56.3 (49.6–62.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e63.8 (58.0-68.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60.6 (53.4–66.5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt; 0.001¶\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003c/p\u003e\n \u003cp\u003e†: chi-squre tests, ¶: Wilcoxon rank-sum tests.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\"\u003e\n \u003ch2\u003e2.2. A Multi-omics Landscape Links Neural Connectivity Pathways to PD-MCI Pathogenesis\u003c/h2\u003e\n \u003cp\u003eWe first sought to define the molecular signature of PD-MCI progression. Differential expression analysis between PD patients who developed mild cognitive impairment (MCI) vs. PD patitents who d\u003cstrong\u003ei\u003c/strong\u003ed not identified 19 differential transcripts (5 upregulated) and 44 differential proteins (2 upregulated) (Fig. 2A, B). Notably, functional enrichment analysis of proteomics and RNA-seq showed similar patterns, where significant pathways were highly concentrated in biological processes critical for neuronal circuit integrity\u003csup\u003e15\u003c/sup\u003e Pathways like axon guidance, neuron projection development, and synapse organization were consistently and significantly enriched across both molecular layers (Fig. 2C-F). This convergent signal strongly suggested \"neural connectopathy\" as a PD-MCI pathology. Gene -Gene interaction network and Protein-protein interaction (PPI) network\u003csup\u003e25\u003c/sup\u003e analysis further refined these findings: it distilled the complex molecular landscape into a core functional module and identified key hub genes\u003csup\u003e13\u003c/sup\u003e, including SLC38A11, PLEKHH2, and protein network hubs such as COL6A1 and ITGAV which act as central regulators of this dysregulated network (Fig. 2E).\u003c/p\u003e\n \u003cp\u003e(A-B) Volcano plots of differential expression analysis comparing PD-MCI patients and those with normal cognition in 4 years. (A) Plasma RNA-seq analysis. (B) CSF proteomics analysis.. (C-D) Gene Ontology (GO) enrichment analysis of biological processes. (C) GO terms significantly enriched among significant RNA. (D) GO terms significantly enriched among significant proteins. (E-F) Protein-Protein Interaction (PPI) network of core differentially expressed proteins.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\"\u003e\n \u003ch2\u003e2.3. Longitudinal Proteomic Trajectories Reveal Dynamic Biomarker Changes Years Before Clinical Diagnosis\u003c/h2\u003e\n \u003cp\u003eTo map the temporal sequence of molecular events leading to cognitive impairment, we tracked the expression of significantly differential proteins (identified in baseline analysis) over the pre-diagnosis period (up to 4 years before reaching PD-MCI). Unsupervised k-means clustering of these longitudinal profiles showed that the PD-MCI proteomic landscape is not static. It evolves through distinct, coordinated dynamic patterns (Fig. 3B). We identified five major trajectory clusters (Cluster 1 to Cluster 5), each with a unique temporal signature (Fig. 3C).\u003c/p\u003e\n \u003cp\u003eNotably, most differential proteins (65.9%) fell into clusters defined by early and mid-term changes (Cluster 3 and Cluster 4), indicating that key pathological processes start years before PD-MCI is clinically diagnosed (Fig. 3D). The trajectory of proteins shows a turning point around 3 years before MCI diagnosis, which may be related to the temporal characteristics of PD patients enrolled in the PPMI cohort and the selection of time points during follow-up.\u003csup\u003e26\u003c/sup\u003e Proteins in Cluster 2 and Cluster 5 exhibit a relatively sustained decrease or increase before MCI diagnosis, and these proteins are closely associated with mechanisms such as extracellular matrix remodeling, metabolic homeostasis, mitochondrial integrity, oxidative stress, and inflammation. Representative trajectories of individual proteins further highlighted this heterogeneity: some markers showing progressive elevation or decline, while others demonstrating an curved pattern (Fig. 3A). Collectively, these findings demonstrate that PD-MCI’s molecular pathophysiology begins long before clinical symptoms appear and follows a multi-phase temporal pattern, providing a critical window for early intervention.\u003c/p\u003e\n \u003cp\u003e(A) Temporal expression patterns (Z-score normalized levels) of individual proteins (top 10 selected from LASSO). These trajectories illustrate the heterogeneous temporal dynamics, with some proteins exhibiting progressive elevation (e.g., Pro-inflammatory markers) and others showing early decline (e.g., Synaptic proteins). (B) Average protein expression trajectories for each cluster. Lines represent the mean Z-score of all proteins in a cluster, shading indicates the standard error of the mean. (C) Heatmap displaying expression patterns of all significantly differentiential protein. (D) Pie chart illustrating the proportion of dysregulated proteins across 5 trajectory clusters.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\"\u003e\n \u003ch2\u003e2.4. A Robust Predictive Model for PD-MCI from PD and its Interpretable Biological Drivers\u003c/h2\u003e\n \u003cp\u003eTo translate our findings into a clinically actionable tool, we developed a machine learning framework to predict individual risk of cognitive decline.\u003csup\u003e27\u003c/sup\u003e LASSO regression first refined the differential expressed candidates to a parsimonious set of 6 transcripts and 7 proteins respectively (Fig. 4A, B). We constructed and compared models of three machine learning algorithms based on four variable patterns: only demographic factors (age, sex, education years); demographic factors plus transcriptomic targets; demographic factors plus proteomic targets; and demographic factors plus both transcriptomic and proteomic targets. These models were evaluated using 5-fold cross-validation.\u003csup\u003e27\u003c/sup\u003e Among the evaluated algorithms, the Logistic Regression model showed superior and robust predictive performance. (AUC = 0.7303, 0.7685, 0.7241, 0.7376), outperforming Random Forest (AUC = 0.6314, 0.7043, 0.6655, 0.6483) and XGBoost (AUC = 0.6440, 0.6674, 0.6385, 0.6938) (Supplementary Fig. 1, Fig. 4C-E). ROC curves were painted plotted for the model incorporating demographic factors, transcriptomic targets, and proteomic targets, comparing different machine learning algorithm (Fig. 4F). Importantly, SHAP analysis clarified the model's decision-making process across different variable patterns (Fig. 5A, 5B, 5C). This interpretability connects the model's predictive power back to specific biological insights, validating the relevance of our omics findings.\u003c/p\u003e\n \u003cp\u003e(A-B) LASSO regression to select the most important biomarkers from DEG results. (A) Plasma transcripts; (B) CSF Proteomics. (C-E) Comparative performance of multiple machine learning algorithms. Box plots display the distribution of ROC-AUC score for Random Forest, XGBoost, and Logistic Regression models evaluated via 5-fold cross-validation. (C) Models based on plasma transcripts + demographics; (D) Models based on CSF proteomics + demographics; (E) Models based on plasma transcripts + CSF proteomics + demographics. (F) Overall model performance evaluation on the independent test set. ROC curves illustrate the discriminative ability of models based on plasma transcripts + CSF proteomics + demographics.\u003c/p\u003e\n \u003cp\u003e(A-C) SHAP Analysis for Model Interpretability. The plot ranks top features by their mean absolute SHAP value, representing their overall importance in the optimal model's predictions. (A) CSF Proteomics + demographics; (B) Plasma Transcripts + demographics; (C) CSF Proteomics + Plasma Transcripts + demographics.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\"\u003e\n \u003ch2\u003e2.5 Cross-sectional Associations Link Candidate Proteins to Domain-Specific Cognitive Performance\u003c/h2\u003e\n \u003cp\u003eTo confirm the clinical relevance of our identified Plasma RNA and CSF protein biomarkers, we addressed their cross-sectional associations with performance across key cognitive domains. Using multivariate linear regression models adjusting for age and years of education\u003csup\u003e28\u003c/sup\u003e, we visualized the standardized effect sizes as a heatmap (Fig. 6).\u003c/p\u003e\n \u003cp\u003eThis analysis revealed distinct, coherent association patterns between specific proteins and cognitive functions. For upregulated genes and proteins in the PD-MCI group identified in differential expression analysis, they are generally and stably negatively correlated with cognitive-related scores, and vice versa. The heatmap illustrates that PD-MCI molecular signatures are not unifrom but differentially linked to impairments in specific cognitive domains. This domain-specific mapping enhances the biological plausibility of our candidate biomarkers and offers a more nuanced understanding of how distinct molecular pathways may contribute to the heterogeneous cognitive profiles in PD-MCI.\u003c/p\u003e\n \u003cp\u003eHeatmap shows standardized beta coefficients from multivariate linear regression models. Each cell represents the association between the features (X-axis) and a specific cognitive test score (Y-axis). The color scale follows the beta value, with p-values labeled as *: p \u0026lt; 0.05, **: p \u0026lt; 0.01. SDM TOTAL: Total score of Semantic Memory test. DVT_TOTAL_RECALL: Total recall score of Delayed Visual Test. DVT_SFTANIM: Semantic Fluency Test (animal category) in Delayed Visual Test. DVT_SDM: Semantic Memory in Delayed Visual Test. DVT_RECOG_RETENTION: Recognition retention in Delayed Visual Test. DVT_RECOG_DISC_INDEX: Recognition discrimination index in Delayed Visual Test. DVT_DELAYED_RECALL: Delayed recall in Delayed Visual Test.\u003c/p\u003e\n \u003cp\u003eDVSD_SDM: Semantic Memory in Delayed Visual Recognition Test. DVS_SFTANIM: Semantic Fluency Test (animal category) in Visual Spatial Test. DVS_LNS: Letter-Number Sequencing in Visual Spatial Test.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\"\u003e\n \u003ch2\u003e2.6. Mendelian Randomization Establishes a Causal Link for a Novel Risk Gene\u003c/h2\u003e\n \u003cp\u003eWhile our analyses established strong associations, we used Two-Sample Mendelian Randomization to infer causality.\u003csup\u003e16\u003c/sup\u003e Using pQTLs and eQTLs as instrumental variables, we found that among the features selected via LASSO regression, LRPPRC, WDR1, EIF3H, PLEKHH2, SLC35G2, and UTS2 had eQTL-based instruments, while MMP12 and MSMB had pQTL-based instruments. Notably, WDR1 showed an association closest to statistical significance (though still p \u0026gt; 0.05) with an increased risk of PDD (Table 2). This causal relationship was robust in sensitivity analyses for pleiotropy and heterogeneity (Supplementary Fig. 2). This genetic evidence elevates WDR1 from a correlative biomarker to a putative causal risk factor, providing strong rationale for its prioritization in future functional studies.\u003c/p\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eMendelian randomization analysis assessing the causal effect of feature levels on the risk of PDD.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eGene\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eeQTLs/pQTLs\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMethod\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eOdds Ratio (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP-value\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eWDR1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eeQTLs\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInverse Variance Weighted\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.730 (0.938–3.190)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.079\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMR Egger\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.791 (0.520–6.163)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.377\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eWeighted Median\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1.929 (0.861–4.322)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.110\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eWeighted Mode\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.015 (0.832–4.881)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.149\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSimple Mode\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2.096 (0.585–7.511)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.280\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\"\u003e\n \u003ch2\u003e2.7. Single-nucleus Transcriptomics Uncovers Altered Cellular Ecology and Validates Key Targets in the PDD Brain\u003c/h2\u003e\n \u003cp\u003eAfter identifying predictive biomarkers in the periphery, we aimed to pinpoint their cellular origin within the pathological brain environment. Single-nucleus RNA sequencing of postmortem prefrontal cortex provided spatially resolved validation.\u003csup\u003e29,30\u003c/sup\u003e Integrated transcriptomic analysis revealed a distinct global shift in the cellular landscape of PDD (Fig. 7A-B, Supplementary Fig.\u0026nbsp;3A-B). We successfully resolved all major brain cell types, including excitatory neurons (35%), inhibitory neurons (22%), oligodendrocytes (26%), astrocytes (9%), microglia (3%), OPCs (4%), and endothelial cells (1%) (Supplementary Fig.\u0026nbsp;3,4C-D).\u003c/p\u003e\n \u003cp\u003eBeyond global transcriptomic changes, we quantified significant alterations in cellular proportions, uncovering a profound dysregulation of the brain's cellular ecology in PDD. Most notably, we observed a marked reduction in oligodendrocytes concurrently with a significant increase in the proportion of microglia. This finding suggests that the pathophysiological process of PDD involves concurrent mechanisms of demyelination/white matter integrity loss and sustained neuroinflammation.\u003c/p\u003e\n \u003cp\u003eCrucially, we then sought to anchor our top causal candidates from prior analyses to their specific cellular context, confirming their altered expression across multiple cell types in PDD (Fig. 7C). Additionally, CellChat further delineated the cell-cell interaction in the PDD group (Fig. 7D). This cell type-specific validation directly links our peripheral and genetic findings to the relevant cellular substrate in the brain, positioning dysfunction within excitatory neurons and inhibitory neurons as a key event in the PDD pathological cascade, potentially underlying the observed \"connectopathy\".\u003c/p\u003e\n \u003cp\u003e(A) Integrated UMAP visualization of all nuclei from PDD and control (CTRL) cohorts. (B) UMAP plot colored by clinical diagnosis. (C) Heatmap of differential expression of predictor genes in different cell types between PDD and CTRL. (D) CellChat analysis of PDD and CTRL.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\"\u003e\n \u003ch2\u003e2.8. From Mechanism to Therapy: Proposing Drug Repurposing Candidates\u003c/h2\u003e\n \u003cp\u003eThe ultimate goal of our translational pipeline was to identify potential interventions. Computational drug-target interaction analysis of feature genes selected from LASSO regression identified several promising candidates for repurposing\u003csup\u003e31\u003c/sup\u003e, including MEDRONIC ACID, ZILEUTON and ATENOLOL etc. (Supplementary Table\u0026nbsp;1). PheWAS analysis further depict the gene associated profiles of these candidate drug targets (Supplementary Fig.\u0026nbsp;5). This analysis directly translates our mechanistic insights into tangible therapeutic hypotheses, offering a shortcut to clinical testing for this devastating condition.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Discussion","content":"\u003cp\u003eThis study employed a multi-omics approach to identify predictive factors for Parkinson's disease with mild cognitive impairment (PD-MCI), integrating transcriptomic and proteomic data to test models and gain insights based on relevant biological markers. While further model refinement is necessary for future clinical translation, this work explores the practical application value of the findings, given the early presence of omics expression differences in these patient populations.\u003c/p\u003e\u003cp\u003eNotably, the machine learning model established after screening differential features using LASSO regression, while improving prediction performance over models using only basic demographic factors (sex, age, education years), demonstrated limited enhancement. Transcriptomic features in this study contributed more to performance improvement compared to proteomic features. Interestingly, integrating transcriptomic with proteomic features did not yield a superior predictive effect but rather a performance compromise between the two. We posit several potential reasons for this. Firstly, the core reason for the weaker performance of proteomics may lie in the dynamic nature of proteins and detection characteristics. Early pathology underlying PD-MCI conversion might initially manifest as gene activation or suppression at the transcriptome level, with corresponding protein expression potentially lagging by weeks or even months. Furthermore, the longer half-life of proteins might hinder their ability to fully reflect PD-MCI conversion risk at baseline. Secondly, the narrower dynamic range and lower diversity of proteomic detection could lead to the masking of low-abundance key molecules, possibly rendering its detection capability relatively limited compared to transcriptomics. Additionally, the relatively limited sample size in this study further constrained effective quantification of low-abundance molecules by proteomics, where stability and data noise issues might be more pronounced. The compromised effect observed upon direct integration of the two omics datasets may primarily stem from signal conflict and data heterogeneity. Differences in the temporal scales of the two omics layers, coupled with the potential risk of overfitting in the transcriptomic data, might have introduced underlying interference and counteraction, preventing the anticipated complementary synergy.\u003c/p\u003e\u003cp\u003eSHAP analysis indicated that age at PD diagnosis was a key predictor for MCI, consistent with previous studies, underscoring the importance of timely intervention, particularly in high-risk patients. Interestingly, longer education duration emerged as a protective factor for cognition in our study, possibly because higher cognitive reserve offers a greater buffer against decline. Furthermore, our study confirmed sex as a meaningful predictive variable, with males more frequently identified as MCI patients, suggesting sex-specific differences in cognitive impairment.\u003c/p\u003e\u003cp\u003eAmong the transcriptomic genes, \u003cem\u003ePLEKHH2\u003c/em\u003e demonstrated the highest predictive weight. Previous research has shown that \u003cem\u003ePLEKHH2\u003c/em\u003e is highly enriched in glomerular podocytes, where it stabilizes the cortical actin cytoskeleton by attenuating actin depolymerization.\u003csup\u003e32\u003c/sup\u003e The relationship between renal function and PD has garnered increasing research interest in recent years====. In oncology, \u003cem\u003ePLEKHH2\u003c/em\u003e has been found involved in \u003cem\u003eALK\u003c/em\u003e fusions in mesenchymal tumors such as epithelioid fibrous histiocytoma, dermatofibrosarcoma protuberans, and lung adenocarcinoma\u003csup\u003e33\u0026ndash;35\u003c/sup\u003e. In non-small cell lung cancer, it exhibits enhanced cytoplasmic expression and significantly promotes tumor cell proliferation, migration, and invasion\u003csup\u003e36\u003c/sup\u003e. In our context, \u003cem\u003ePLEKHH2\u003c/em\u003e was significantly correlated with multiple test scores in visuospatial and delayed visual memory tests. Single-cell RNA data from PDD patients revealed its RNA expression was elevated in oligodendrocytes, microglia, and excitatory neurons, but downregulated in inhibitory neurons. These findings collectively suggest a potential role for \u003cem\u003ePLEKHH2\u003c/em\u003e in processes related to the extracellular matrix and cellular dynamics within the brain.\u003c/p\u003e\u003cp\u003eAmong proteins, CA1 (Carbonic Anhydrase 1) exhibited the highest predictive weight. The closely related CA3 was also among the significantly differentially expressed proteins. As a popular drug target, CA1 already has several classic clinically available drugs. Two FDA-approved, repurposed carbonic anhydrase inhibitors, acetazolamide and methazolamide, were found to correct proteomic abnormalities and effectively prevent early molecular pathologies related to cerebral amyloid angiopathy (CAA), maintain synaptic stability, and suppress neuroinflammation in Tg-SwDI mice carrying pathogenic mutations and exhibiting Alzheimer's disease (AD)-like cognitive deficits and severe CAA\u003csup\u003e37\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003ePathway enrichment analyses of features significantly different between groups in both proteomics and transcriptomics pointed to alterations in key functions like axon guidance and synaptic transmission. Notably, the gene sets corresponding to these significant features from the two omics layers showed no overlap. Trajectory clustering analysis of proteomic data and correlation analysis with clinical indicators further suggested potential functional impacts of significant upregulation or downregulation of the relevant proteins. Due to the lack of GWAS data for PD-MCI and single-cell sequencing data specifically for PD-MCI, we utilized relevant data from PDD patients versus healthy controls for further validation of the machine learning-derived features, revealing causal genetic associations and enrichment patterns across different cell types. Finally, using the screened genes, we preliminarily identified potential drug candidates, which might offer insights for future therapeutic interventions in PD-MCI or PDD.\u003c/p\u003e\u003cp\u003eMachine learning models demonstrated depth in identifying predictive features beyond traditional statistical methods, particularly in capturing complex non-linear relationships between variables. Based on baseline omics data, this study identified several robust predictors through multivariate association analysis. These biomarkers hold potential application value in precision medicine and routine clinical practice. Our findings may provide clinicians with novel decision-support tools, aiding in the early identification of high-risk patients and facilitating the clinical translation of timely intervention and personalized management strategies.\u003c/p\u003e\u003cp\u003eHowever, this study has several limitations. First, the population representativeness of the study cohort is limited, and the availability of some key variables was uneven across sub-cohorts, potentially affecting the model's generalizability. Secondly, constrained by sample size and patient heterogeneity, the predictive stability of the model across different clinical settings requires further improvement. Furthermore, although the model showed good predictive performance in cross-study validation, a gap remains before practical clinical application, necessitating further optimization and validation through large-scale prospective studies.\u003c/p\u003e\u003cp\u003eDespite these challenges, the machine learning approach based on multi-omics data lays an important foundation for building highly practical, digitally deployable predictive models for cognitive impairment in PD. Future research should focus on expanding, optimizing, and validating these predictive markers in broader and more diverse populations, while actively exploring their potential for guiding early intervention strategies, thereby advancing precision medicine in the field of neurodegenerative diseases.\u003c/p\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eIn conclusion, via a tightly integrated, multi-stage analysis, we have constructed an early diagnostic model for PD-MCI (from PD) and elucidated its pathological mechanisms using multiple approaches. Besides, we genetically linking biological predictors to PDD as causal risk factors, and have identified novel, high-priority targets for therapeutic development. Ultimately, this work offers a comprehensive resource guiding the transition from understanding PD-MCI mechanisms to targeted intervention for PD patients at riks of progressing to PD-MCI.\u003csup\u003e31\u003c/sup\u003e\u003c/p\u003e"},{"header":"5. Materials and Methods","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e5.1. Study Cohort and Participants\u003c/h2\u003e\u003cp\u003eData used in this study were obtained from the Parkinson's Progression Markers Initiative (PPMI) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.ppmi-info.org/access-data-specimens/download-data\u003c/span\u003e\u003c/span\u003e, RRID:SCR_006431).\u003csup\u003e38\u003c/sup\u003e PPMI is an ongoing international, multicenter, prospective observational study designed to identify biomarkers of PD progression.\u003csup\u003e39,40\u003c/sup\u003e For this analysis, we included de-identified data from PD patients. All participants provided written informed consent, and the study was approved by the institutional review board at each participating site.\u003csup\u003e39,41\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eThe primary analysis cohort consisted of PD patients with available baseline plasma RNA-seq data, baseline CSF proteomics data, and complete longitudinal cognitive assessments over 4 years.\u003csup\u003e31,42\u003c/sup\u003e \u003cb\u003eSpecifically, from the initial 4115 PPMI participants, we focused on 1454 PD patients. After applying inclusion criteria (availability of paired multi-omics data and 4-year cognitive follow-up), 322 patients constituted the final analysis cohort (\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA\u003cb\u003e).\u003c/b\u003e According to the Movement Disorder Society Level II criteria \u003csup\u003e43\u003c/sup\u003e, the Montreal Cognitive Assessment (MoCA), using a cutoff score of 26, was used as the primary cognitive test to define cognitive progression and diagnose PD-MCI.\u003csup\u003e44\u003c/sup\u003e Other cognitive tests were utilized to validate biomarker stability. \u003csup\u003e43\u003c/sup\u003eDetailed inclusion and exclusion criteria, as well as the participant selection flowchart, are provided in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA. Baseline demographic and clinical characteristics of the included participants are summarized in Table\u0026nbsp;1.\u003csup\u003e45\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e5.2. Multi-omics Data Acquisition and Preprocessing\u003c/h2\u003e\u003cdiv id=\"Sec16\" class=\"Section3\"\u003e\u003ch2\u003e5.2.1. Plasma RNA-seq\u003c/h2\u003e\u003cp\u003eAccording to the PPMI sequencing protocol, total RNA was extracted from plasma samples. Libraries were prepared and sequenced on an Illumina platform.\u003csup\u003e46\u003c/sup\u003e Raw sequencing reads (FASTQ files) were quality-controlled using FastQC (v0.11.9) and trimmed using Trimmomatic (v0.39).\u003csup\u003e47\u003c/sup\u003e Reads were aligned to the human reference genome (GRCh38) using STAR (v2.7.10a)\u003csup\u003e48\u003c/sup\u003e, and gene counts were quantified using featureCounts (v2.0.1).\u003csup\u003e49\u003c/sup\u003e Counts were normalized and transformed using the varianceStabilizingTransformation function in the DESeq2 R package (v1.34.0)\u003csup\u003e50\u003c/sup\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section3\"\u003e\u003ch2\u003e5.2.2. CSF Proteomics\u003c/h2\u003e\u003cp\u003eCSF samples were analyzed using SomaLogic\u0026rsquo;s SOMAscan platform, with the target molecules being protein biomarkers based on the Slow Off-rate Modified Aptamer (SOMAmer) technology. SOMAscan data were processed following SomaLogic's standard workflow \"HybNormPlateScaleCal\",which includes four core steps: Hybridization Normalization, Plate Scaling, Intra-plate Median Signal Normalization, and Calibration. The data processed through this workflow were used as analyzable raw data\u003csup\u003e51\u003c/sup\u003e. Protein intensities were log2-transformed, and \u003cb\u003equantile normalization was performed to correct for inter-sample variation\u003c/b\u003e. Batch effects were adjusted for using the ComBat function from the sva R package.\u003csup\u003e51,52\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e5.3. Statistical and Bioinformatic Analysis\u003c/h2\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e5.3.1. Differential Expression Analysis\u003c/h2\u003e\u003cp\u003eDifferential analysis of RNA-seq and proteomics data between progressors and non-progressors was performed using linear models in the limma R package (v3.50.3)\u003csup\u003e53,54\u003c/sup\u003e. The false discovery rate (FDR)\u003csup\u003e55\u003c/sup\u003e was controlled using the Benjamini-Hochberg method. Significance was set at |log2(fold change)| \u0026gt;0.3 and adjusted \u003cem\u003ep-value\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.1.\u003csup\u003e56,57\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section3\"\u003e\u003ch2\u003e5.3.2. Functional Enrichment Analysis\u003c/h2\u003e\u003cp\u003eGene Ontology (GO) biological process enrichment analysis was conducted on significantly dysregulated genes and proteins using the clusterProfiler R package (v4.2.2).\u003csup\u003e23,58,59\u003c/sup\u003e Terms with an FDR-adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered significant.\u003csup\u003e60,61,62\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section3\"\u003e\u003ch2\u003e\u003cb\u003e5.3.3. Interaction Network Analysis\u003c/b\u003e\u003c/h2\u003e\u003cp\u003eA gene-gene interaction network was constructed using the genemania database. A protein-protein interaction (PPI) network was constructed using the STRING database (v11.5).\u003csup\u003e63,64,65\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section3\"\u003e\u003ch2\u003e5.3.4. Longitudinal Proteomic Trajectories Analysis\u003c/h2\u003e\u003cp\u003eTo delineate the temporal dynamics of protein expression in the progression of Parkinson's disease with mild cognitive impairment (PD-MCI)\u003csup\u003e3\u003c/sup\u003e, we performed trajectory modeling and unsupervised clustering. Protein expression levels were first standardized into Z-scores relative to a matched control group to mitigate inter-individual variability. The longitudinal trajectories of these Z-scores across the years leading to diagnosis were then modeled using Locally Estimated Scatterplot Smoothing (LOESS) regression, providing a continuous view of protein fluctuations.\u003csup\u003e66\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eSubsequently, we conducted unsupervised hierarchical clustering based on the Euclidean distance between the LOESS-predicted trajectories to group proteins with similar temporal patterns. The Ward.D2 method was employed as the agglomeration criterion to minimize within-cluster variance. This analysis robustly categorized the proteins into five distinct clusters. A protein was defined as dysregulated if its absolute Z-score exceeded the threshold of 0.3, as visualized in the trajectory heatmap.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec23\" class=\"Section3\"\u003e\u003ch2\u003e5.3.5. Machine Learning Modeling\u003c/h2\u003e\u003cp\u003eLeast absolute shrinkage and selection operator (LASSO) regression was applied for feature selection with optimal lambda.\u003csup\u003e67,68\u003c/sup\u003e The resulting features were used to train and compare three machine learning classifiers: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Logistic Regression (LR).\u003csup\u003e69,70,71\u003c/sup\u003e Model performance was evaluated using repeated (n\u0026thinsp;=\u0026thinsp;5) 5-fold cross-validation.\u003csup\u003e72,73\u003c/sup\u003eThe area under the receiver operating characteristic curve (AUC) was used as the primary metric for model comparison and selection.\u003csup\u003e45\u003c/sup\u003e The final model was interpreted using SHapley Additive exPlanations (SHAP) to assess the contribution of each feature to the predictions.\u003csup\u003e38,74,75\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec24\" class=\"Section3\"\u003e\u003ch2\u003e5.3.6. Clinical Correlation Analysis\u003c/h2\u003e\u003cp\u003eTo evaluate the clinical relevance of the identified multi-omics features, we assessed their associations with key neuropsychological assessments and demographic factors. We performed multivariate linear regression analyses, treating the cognitive scores from a battery of tests as dependent variables. The independent variables comprised the baseline levels of candidate proteins and transcripts identified in differential expression analysis and LASSO regression,and all models were adjusted for critical covariates, including age and education years, to control for their potential confounding effects.\u003csup\u003e76\u003c/sup\u003e The results are presented as standardized Beta coefficients, quantifying the strength and direction of the associations. A positive Beta indicates that a higher omics feature level is associated with a better cognitive performance, while a negative Beta suggests an association with poorer cognitive function.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec25\" class=\"Section3\"\u003e\u003ch2\u003e5.3.7. Mendelian Randomization (MR)\u003c/h2\u003e\u003cp\u003eProtein quantitative trait locis (pQTLs) and expression quantitative trait loci (eQTLs) were sourced from the UKB-PPP and the eQTLGen Consortium.\u003csup\u003e77\u003c/sup\u003e Summary statistics for the outcome (PDD) were obtained from the FinnGen consortium (R9 release, finn-b-PD_DEMENTIA).\u003csup\u003e45,78\u003c/sup\u003e The TwoSampleMR R package (v0.5.7)\u003csup\u003e79\u003c/sup\u003e was used for MR analyses. Inverse-variance weighted (IVW) was the primary method, supplemented by MR-Egger, weighted median, and weighted mode methods.\u003csup\u003e20,80\u003c/sup\u003e Sensitivity analyses included Cochran's Q test for heterogeneity, the MR-Egger intercept test for horizontal pleiotropy, and leave-one-out analysis.\u003csup\u003e80,81\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec26\" class=\"Section3\"\u003e\u003ch2\u003e5.3.8. Single-nucleus RNA Sequencing (snRNA-seq) Analysis\u003c/h2\u003e\u003cp\u003esnRNA-seq data from GEO303823\u003csup\u003e82\u003c/sup\u003e were already processed using the Cell Ranger pipeline (v7.1.0).\u003csup\u003e76,83\u003c/sup\u003e Downstream analysis, including quality control, normalization, integration, clustering, and cell type annotation, was performed in Seurat (v4.3.0).\u003csup\u003e84,85\u003c/sup\u003e Differential expression and visualization of candidate genes were conducted.\u003csup\u003e78,79,20,81\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eFor each cell type (those containing at least 3 cells in both groups) and gene combination (among the selected genes with expression records), the log2 fold change was calculated and the non-parametric Wilcoxon rank-sum test was employed to assess the significance of expression differences for each gene between the groups. The obtained raw p-values were adjusted for false discovery rate (FDR) using the Benjamini-Hochberg method. The differential expression results were visualized using a heatmap. Cell-cell communication analysis was performed to depict cell\u0026ndash;cell interactions using CellChat and CellChatDB.human as a reference.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec27\" class=\"Section3\"\u003e\u003ch2\u003e5.3.9. Drug-Target Interaction and Repurposing Analysis\u003c/h2\u003e\u003cp\u003eThe DrugBank database (v5.1.9) was queried to identify known and investigational drugs targeting the prioritized hub proteins. Further, the DGIdb database was used for comprehensive drug-gene interaction mining.\u003csup\u003e86,87\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e All authors declare no financial or non-financial competing interests.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eData Availability and Code Availability\u003c/strong\u003e\u003cbr\u003e Processed data supporting the findings of this study are available within the paper and its supplementary information. Raw PPMI data are available upon registration and application through the PPMI website [https://www.ppmi-info.org/]. Analysis code will be made publicly available on GitHub upon publication.\u003csup\u003e39\u003c/sup\u003e\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eEthics Declaration\u003c/strong\u003e\u003cbr\u003e The PPMI study was conducted in accordance with the Declaration of Helsinki and was approved by the institutional review board of all participating sites. All participants provided written informed consent.\u003csup\u003e39\u003c/sup\u003e Ethical approval for the use of postmortem snRNA-seq data was obtained by the original data providers, and this study complied with all terms of use.\u003csup\u003e39,82\u003c/sup\u003e\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAcknowledgement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to express their sincere gratitude to Wenbo Ji and Rui Zhang for their invaluable assistance in data collection and technical support during the course of this research.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAA\u003c/strong\u003e: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Visualization. \u003cstrong\u003eYG\u003c/strong\u003e: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data Curation, Writing - Original Draft, Visualization. \u003cstrong\u003eLW\u003c/strong\u003e: Investigation, Resources, Data Curation, Writing - Review \u0026amp; Editing. \u003cstrong\u003eWX\u003c/strong\u003e: Methodology, Software, Validation, Investigation. \u003cstrong\u003eQN\u003c/strong\u003e: Resources, Supervision, Project administration. \u003cstrong\u003eLS\u003c/strong\u003e: Software, Data Curation, Visualization. \u003cstrong\u003eXF\u003c/strong\u003e: Resources, Supervision. \u003cstrong\u003eDH\u003c/strong\u003e: Conceptualization, Methodology, Writing - Review \u0026amp; Editing, Supervision, Project administration. \u003cstrong\u003eJC\u003c/strong\u003e: Conceptualization, Methodology, Writing - Review \u0026amp; Editing, Supervision, Project administration. \u003cstrong\u003eRX\u003c/strong\u003e: Conceptualization, Methodology, Writing - Review \u0026amp; Editing, Supervision, Project administration. \u003cstrong\u003eYY\u003c/strong\u003e: Conceptualization, Resources, Writing - Review \u0026amp; Editing, Supervision, Project administration, Funding acquisition.\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Natural Science Foundation of China, China (82471211), the Science and Technology Innovation Action Plan of the Shanghai Science and Technology Commission, China (23Y11906600).\u003c/p\u003e\n\u003cp\u003eNew Quality Clinical Specialties of High-end Medical Disciplinary Construction in Pudong New Area(2024-PWXZ-17)\u003c/p\u003e\n\u003cp\u003eKey Disciplines of Shanghai East Hospital(2024-DFZD-003)\u003c/p\u003e\n\u003cp\u003eKey Discipline of Shanghai Health System(2024ZDXK0002)\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAssessment of Associations and Interventions in Movement Disorders: NSAID Use and Risk of Parkinsonian Markers in Genetic At-Risk Populations, and Examining Deep Brain Stimulation Surgery in Isolated and Acquired Dystonia - ProQuest. https://www.proquest.com/docview/3217274591.\u003c/li\u003e\n\u003cli\u003ePostuma, R. B., Berg, D. \u0026amp; Stern, M. MDS clinical diagnostic criteria for Parkinson\u0026rsquo;s disease. \u003cem\u003eMov Disord\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1591\u0026ndash;1601 (2015).\u003c/li\u003e\n\u003cli\u003eAbstract - 2020 - Movement Disorders - Wiley Online Library. https://movementdisorders.onlinelibrary.wiley.com/doi/10.1002/mds.28268.\u003c/li\u003e\n\u003cli\u003eFeigin, V. L., Abajobir, A. A. \u0026amp; Abate, K. H. Global, regional, and national burden of neurological disorders during 1990\u0026ndash;2015: a systematic analysis for the Global Burden of Disease Study 2015. \u003cem\u003eLancet Neurol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 877\u0026ndash;897 (2017).\u003c/li\u003e\n\u003cli\u003eAarsland, D. \u003cem\u003eet al.\u003c/em\u003e Parkinson disease-associated cognitive impairment. \u003cem\u003eNat. Rev. Dis. Primer\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 47 (2021).\u003c/li\u003e\n\u003cli\u003eMartinez-Eguiluz, M. \u003cem\u003eet al.\u003c/em\u003e Machine Learning for Prediction of Cognitive Deterioration in Patients with Early Parkinson\u0026rsquo;s Disease. \u003cem\u003eAppl. Sci.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 8149 (2024).\u003c/li\u003e\n\u003cli\u003eHely, M. A., Reid, W. G. J., Adena, M. A., Halliday, G. M. \u0026amp; Morris, J. G. L. The Sydney multicenter study of Parkinson\u0026rsquo;s disease: the inevitability of dementia at 20 years. \u003cem\u003eMov. Disord. Off. J. Mov. Disord. Soc.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 837\u0026ndash;844 (2008).\u003c/li\u003e\n\u003cli\u003eWeintraub, D. \u003cem\u003eet al.\u003c/em\u003e Management of psychiatric and cognitive complications in Parkinson\u0026rsquo;s disease. \u003cem\u003eBMJ\u003c/em\u003e \u003cstrong\u003e379\u003c/strong\u003e, e068718 (2022).\u003c/li\u003e\n\u003cli\u003eRole of Gut-Liver-Brain Axis in Neuropsychiatric Symptoms of Parkinson\u0026rsquo;s Disease - ProQuest. https://www.proquest.com/docview/3217835689.\u003c/li\u003e\n\u003cli\u003eHe, P. \u003cem\u003eet al.\u003c/em\u003e The association of CSF biomarkers and cognitive decline with choroid plexus volume in early Parkinson\u0026rsquo;s disease. \u003cem\u003eParkinsonism Relat. Disord.\u003c/em\u003e \u003cstrong\u003e120\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eIrwin, D. J., Grossman, M. \u0026amp; Weintraub, D. Neuropathological and genetic correlates of survival and dementia onset in synucleinopathies: a retrospective analysis. \u003cem\u003eLancet Neurol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 55\u0026ndash;65 (2017).\u003c/li\u003e\n\u003cli\u003eHarvey, J., Rekkas, V. \u0026amp; Svenningsson, P. Machine learning-based prediction of cognitive outcomes in de novo Parkinson\u0026rsquo;s disease. \u003cem\u003eNPJ Park. Dis\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eKarayel, O., Virreira Winter, S. \u0026amp; Padmanabhan, S. Proteome profiling of cerebrospinal fluid reveals biomarker candidates for Parkinson\u0026rsquo;s disease. \u003cem\u003eCell Rep Med\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eNalls, M. A., Blauwendraat, C. \u0026amp; Vallerga, C. L. Identification of novel risk loci, causal insights, and heritable risk for Parkinson\u0026rsquo;s disease: a meta-analysis of genome-wide association studies. \u003cem\u003eLancet Neurol\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1091\u0026ndash;1102 (2019).\u003c/li\u003e\n\u003cli\u003eLiu, G., Peng, J. \u0026amp; Liao, Z. Genome-wide survival study identifies a novel synaptic locus and polygenic score for cognitive progression in Parkinson\u0026rsquo;s disease. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 787\u0026ndash;793 (2021).\u003c/li\u003e\n\u003cli\u003eSun, B. B., Chiou, J. \u0026amp; Traylor, M. Plasma proteomic associations with genetics and health in the UK Biobank. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e622\u003c/strong\u003e, 329\u0026ndash;338 (2023).\u003c/li\u003e\n\u003cli\u003eCui, M., Cheng, C. \u0026amp; Zhang, L. High-throughput proteomics: a methodological mini-review. \u003cem\u003eLab. Investig. J. Tech. Methods Pathol.\u003c/em\u003e \u003cstrong\u003e102\u003c/strong\u003e, 1170\u0026ndash;1181 (2022).\u003c/li\u003e\n\u003cli\u003eSystematic Review of Proteomics in Age-Related Macular Degeneration and Pathway Analysis of Significant Protein Changes - Ophthalmology Science. https://www.ophthalmologyscience.org/article/S2666-9145(25)00091-0/fulltext.\u003c/li\u003e\n\u003cli\u003eChia, R. \u003cem\u003eet al.\u003c/em\u003e A plasma proteomics-based candidate biomarker panel predictive of amyotrophic lateral sclerosis. \u003cem\u003eNat. Med.\u003c/em\u003e https://doi.org/10.1038/s41591-025-03890-6 (2025) doi:10.1038/s41591-025-03890-6.\u003c/li\u003e\n\u003cli\u003eBurgess, S. \u0026amp; Thompson, S. G. Interpreting findings from Mendelian randomization using the MR-Egger method. \u003cem\u003eEur. J. Epidemiol.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 377\u0026ndash;389 (2017).\u003c/li\u003e\n\u003cli\u003eZhao, W. \u003cem\u003eet al.\u003c/em\u003e Risk associated circulating biomarkers S100A3 identified in congenital heart disease-associated pulmonary arterial hypertension. \u003cem\u003eIntractable Rare Dis. Res.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 36\u0026ndash;45 (2025).\u003c/li\u003e\n\u003cli\u003eMapping the dynamics of epigenetic adaptation | bioRxiv. https://www.biorxiv.org/content/10.1101/2023.07.10.548368v1.full.\u003c/li\u003e\n\u003cli\u003eA Population-scale Single-cell Spatial Transcriptomic Atlas of the Human Cortex | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.10.13.681959v1.\u003c/li\u003e\n\u003cli\u003eWang, J. \u003cem\u003eet al.\u003c/em\u003e Impact of Y chromosome loss on the risk of Parkinson\u0026rsquo;s disease and progression. \u003cem\u003eeBioMedicine\u003c/em\u003e \u003cstrong\u003e117\u003c/strong\u003e, (2025).\u003c/li\u003e\n\u003cli\u003eJeong, D., Koo, B., Oh, M., Kim, T.-B. \u0026amp; Kim, S. GOAT: Gene-level biomarker discovery from multi-Omics data using graph ATtention neural network for eosinophilic asthma subtype. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e39\u003c/strong\u003e, btad582 (2023).\u003c/li\u003e\n\u003cli\u003eDevelopment and Validation of PARCOMS Composite Scales for Assessing Disease Progression and Treatment Effects in Parkinson\u0026rsquo;s Disease | Neurology and Therapy. https://link.springer.com/article/10.1007/s40120-025-00771-5.\u003c/li\u003e\n\u003cli\u003eSood, M., Ktena, S. I. \u0026amp; Wenzel, F. Generating synthetic patient data using Bayesian networks. \u003cem\u003eJ Am Med Inf. Assoc\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 1255\u0026ndash;1267 (2020).\u003c/li\u003e\n\u003cli\u003eWang, Z. \u003cem\u003eet al.\u003c/em\u003e Cerebrospinal fluid proteomics identification of biomarkers for amyloid and tau PET stages. \u003cem\u003eCell Rep. Med.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 102031 (2025).\u003c/li\u003e\n\u003cli\u003eHabib, N. \u003cem\u003eet al.\u003c/em\u003e Massively parallel single-nucleus RNA-seq with DroNc-seq. \u003cem\u003eNat. Methods\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 955\u0026ndash;958 (2017).\u003c/li\u003e\n\u003cli\u003eJiang, Z. Q., Wu, J. J. \u0026amp; Liu, J. Characterization of a pathogenic variant in GBA for Parkinson\u0026rsquo;s disease with mild cognitive impairment patients. \u003cem\u003eMol Brain\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eWalker, K. A., Chen, J. \u0026amp; Zhang, J. Proteomics analysis of plasma from middle-aged adults identifies protein markers of dementia risk in later life. \u003cem\u003eSci Transl Med\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003ePerisic, L. \u003cem\u003eet al.\u003c/em\u003e Plekhh2, a novel podocyte protein downregulated in human focal segmental glomerulosclerosis, is involved in matrix adhesion and actin dynamics. \u003cem\u003eKidney Int.\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 1071\u0026ndash;1083 (2012).\u003c/li\u003e\n\u003cli\u003eDeSimone, M. S. \u003cem\u003eet al.\u003c/em\u003e Epithelioid Fibrous Histiocytoma Is on a Continuum With Superficial ALK -rearranged Myxoid Spindle Cell Neoplasm : A Clinicopathologic Series of 35 Cases Including Alternate RET and NTRK3 Fusions. \u003cem\u003eAm. J. Surg. Pathol.\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 1568\u0026ndash;1579 (2024).\u003c/li\u003e\n\u003cli\u003eAgrawal, S. \u003cem\u003eet al.\u003c/em\u003e ALK-rearranged, CD34-positive spindle cell neoplasms resembling dermatofibrosarcoma protuberans: a study of seven cases. \u003cem\u003eHistopathology\u003c/em\u003e \u003cstrong\u003e85\u003c/strong\u003e, 649\u0026ndash;659 (2024).\u003c/li\u003e\n\u003cli\u003eCoppock, J. D. \u003cem\u003eet al.\u003c/em\u003e S100 and CD34 Expressing Mesenchymal Neoplasm With Rare PLEKHH2::ALK Fusion and Response to ALK Inhibition. \u003cem\u003eAm. J. Surg. Pathol.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 1309\u0026ndash;1313 (2022).\u003c/li\u003e\n\u003cli\u003eWang, R. \u003cem\u003eet al.\u003c/em\u003e PLEKHH2 binds \u0026beta;-arrestin1 through its FERM domain, activates FAK/PI3K/AKT phosphorylation, and promotes the malignant phenotype of non-small cell lung cancer. \u003cem\u003eCell Death Dis.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 858 (2022).\u003c/li\u003e\n\u003cli\u003eCarlsen, J., Fossati, S., \u0026Oslash;stergaard, L., Guti\u0026eacute;rrez-Jim\u0026eacute;nez, E. \u0026amp; Palmfeldt, J. Cerebral proteome adaptations to amyloid angiopathy are prevented by carbonic anhydrase inhibitors. \u003cem\u003eAlzheimers Dement. J. Alzheimers Assoc.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, e70122 (2025).\u003c/li\u003e\n\u003cli\u003eFabrizio, C., Termine, A. \u0026amp; Caltagirone, C. Transcriptomics profiling of Parkinson\u0026rsquo;s disease progression subtypes reveals distinctive patterns of gene expression. 2023.10.11.23296884 Preprint at https://doi.org/10.1101/2023.10.11.23296884 (2023).\u003c/li\u003e\n\u003cli\u003eMarek, K., Chowdhury, S. \u0026amp; Siderowf, A. The Parkinson\u0026rsquo;s progression markers initiative (PPMI) - establishing a PD biomarker cohort. \u003cem\u003eAnn Clin Transl Neurol\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 1460\u0026ndash;1477 (2018).\u003c/li\u003e\n\u003cli\u003eAssociations between Sleep Disorders and Impulsive-Compulsive Behaviors in Parkinson\u0026rsquo;s Disease: A Prospective Cohort Study | Neuroepidemiology | Karger Publishers. https://karger.com/ned/article/58/4/256/895414/Associations-between-Sleep-Disorders-and-Impulsive.\u003c/li\u003e\n\u003cli\u003eMachine learning-based classification of dual fluorescence signals reveals muscle stem cell fate transitions in response to regenerative niche factors | npj Regenerative Medicine. https://www.nature.com/articles/s41536-023-00277-4.\u003c/li\u003e\n\u003cli\u003eLoo, R. T. J. \u003cem\u003eet al.\u003c/em\u003e Multi-cohort machine learning identifies predictors of cognitive impairment in Parkinson\u0026rsquo;s disease. \u003cem\u003eNPJ Digit. Med.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 482 (2025).\u003c/li\u003e\n\u003cli\u003eLitvan, I. \u003cem\u003eet al.\u003c/em\u003e Diagnostic criteria for mild cognitive impairment in Parkinson\u0026rsquo;s disease: Movement Disorder Society Task Force guidelines. \u003cem\u003eMov. Disord. Off. J. Mov. Disord. Soc.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 349\u0026ndash;356 (2012).\u003c/li\u003e\n\u003cli\u003eXing, Y. \u003cem\u003eet al.\u003c/em\u003e \u0026lt;p\u0026gt;Regional Neural Activity Changes in Parkinson\u0026amp;rsquo;s Disease-Associated Mild Cognitive Impairment and Cognitively Normal Patients\u0026lt;/p\u0026gt;. \u003cem\u003eNeuropsychiatr. Dis. Treat.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 2697\u0026ndash;2706 (2021).\u003c/li\u003e\n\u003cli\u003eFrontiers | Machine learning-based prediction of carotid intima\u0026ndash;media thickness progression: a three-year prospective cohort study. https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1593662/full.\u003c/li\u003e\n\u003cli\u003eGuduric-Fuchs, J. \u003cem\u003eet al.\u003c/em\u003e Selective extracellular vesicle-mediated export of an overlapping set of microRNAs from multiple cell types. \u003cem\u003eBMC Genomics\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 357 (2012).\u003c/li\u003e\n\u003cli\u003eMapping the dynamics of epigenetic adaptation | bioRxiv. https://www.biorxiv.org/content/10.1101/2023.07.10.548368v1.full.\u003c/li\u003e\n\u003cli\u003eDobin, A. \u003cem\u003eet al.\u003c/em\u003e STAR: ultrafast universal RNA-seq aligner. \u003cem\u003eBioinforma. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 15\u0026ndash;21 (2013).\u003c/li\u003e\n\u003cli\u003eEpigenomic and transcriptomic analyses reveal cnidocyte specialization in a sea anemone | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.10.06.680633v1.\u003c/li\u003e\n\u003cli\u003eLove, M. I., Huber, W. \u0026amp; Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. \u003cem\u003eGenome Biol.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 550 (2014).\u003c/li\u003e\n\u003cli\u003eJohnson, W. E., Li, C. \u0026amp; Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. \u003cem\u003eBiostat. Oxf. Engl.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 118\u0026ndash;127 (2007).\u003c/li\u003e\n\u003cli\u003eFull article: Copper associates with differential methylation in placentae from two US birth cohorts. https://www.tandfonline.com/doi/full/10.1080/15592294.2019.1661211.\u003c/li\u003e\n\u003cli\u003eRitchie, M. E. \u003cem\u003eet al.\u003c/em\u003e limma powers differential expression analyses for RNA-sequencing and microarray studies. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, e47 (2015).\u003c/li\u003e\n\u003cli\u003eChanges in the miRNA-mRNA Regulatory Network Precede Motor Symptoms in a Mouse Model of Multiple System Atrophy: Clinical Implications | PLOS One. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0150705.\u003c/li\u003e\n\u003cli\u003eGlickman, M. E., Rao, S. R. \u0026amp; Schultz, M. R. False discovery rate control is a recommended alternative to Bonferroni-type adjustments in health studies. \u003cem\u003eJ. Clin. Epidemiol.\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 850\u0026ndash;857 (2014).\u003c/li\u003e\n\u003cli\u003eCSF proteomics in autosomal dominant Alzheimer\u0026rsquo;s disease highlights parallels with sporadic disease | Brain | Oxford Academic. https://academic.oup.com/brain/article/146/11/4495/7205393.\u003c/li\u003e\n\u003cli\u003eDUX4-stimulated genes define an antiviral defense program in human placental trophoblasts | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.01.22.634317v2.\u003c/li\u003e\n\u003cli\u003eYu, G., Wang, L.-G., Han, Y. \u0026amp; He, Q.-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. \u003cem\u003eOmics J. Integr. Biol.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 284\u0026ndash;287 (2012).\u003c/li\u003e\n\u003cli\u003eAshburner, M. \u003cem\u003eet al.\u003c/em\u003e Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 25\u0026ndash;29 (2000).\u003c/li\u003e\n\u003cli\u003eKanehisa, M., Goto, S., Furumichi, M., Tanabe, M. \u0026amp; Hirakawa, M. KEGG for representation and analysis of molecular networks involving diseases and drugs. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, (2010).\u003c/li\u003e\n\u003cli\u003eMilacic, M., Beavers, D. \u0026amp; Conley, P. The Reactome Pathway Knowledgebase 2024. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e52\u003c/strong\u003e, (2024).\u003c/li\u003e\n\u003cli\u003eLi, J. \u003cem\u003eet al.\u003c/em\u003e Functional study of RNA alternative splicing in zebrafish adaptation to low temperature. \u003cem\u003eComp. Biochem. Physiol. Part D Genomics Proteomics\u003c/em\u003e 101656 (2025) doi:10.1016/j.cbd.2025.101656.\u003c/li\u003e\n\u003cli\u003eShannon, P., Markiel, A. \u0026amp; Ozier, O. Cytoscape: a software environment for integrated models of biomolecular interaction networks. \u003cem\u003eGenome Res\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 2498\u0026ndash;2504 (2003).\u003c/li\u003e\n\u003cli\u003eSzklarczyk, D. \u003cem\u003eet al.\u003c/em\u003e The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, D638\u0026ndash;D646 (2023).\u003c/li\u003e\n\u003cli\u003eOh, H. \u003cem\u003eet al.\u003c/em\u003e Cytokine-Related Genes and Inflammatory Profiles as Potential Biomarkers in Major Depressive Disorder. \u003cem\u003ePsychiatry Investig.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 858\u0026ndash;869 (2025).\u003c/li\u003e\n\u003cli\u003eGan, Y.-H. \u003cem\u003eet al.\u003c/em\u003e Large-scale proteomic analyses of incident Parkinson\u0026rsquo;s disease reveal new pathophysiological insights and potential biomarkers. \u003cem\u003eNat. Aging\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 642\u0026ndash;657 (2025).\u003c/li\u003e\n\u003cli\u003eFriedman, J., Hastie, T. \u0026amp; Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. \u003cem\u003eJ. Stat. Softw.\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1\u0026ndash;22 (2010).\u003c/li\u003e\n\u003cli\u003eIdentifying pyroptosis-related genes as novel therapeutic targets in diabetic foot ulceration | Diabetology \u0026amp; Metabolic Syndrome | Full Text. https://dmsjournal.biomedcentral.com/articles/10.1186/s13098-025-01880-9.\u003c/li\u003e\n\u003cli\u003eBreiman, L. Random Forests. \u003cem\u003eMach Learn\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 5\u0026ndash;32 (2001).\u003c/li\u003e\n\u003cli\u003eXGBoost | Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. https://dl.acm.org/doi/10.1145/2939672.2939785.\u003c/li\u003e\n\u003cli\u003eNega Tarekegn, A., Alaya Cheikh, F., Sajjad, M. \u0026amp; Ullah, M. Towards Detecting Freezing of Gait Events Using Wearable Sensors and Genetic Programming. in \u003cem\u003eArtificial Intelligence and Soft Computing\u003c/em\u003e (eds Rutkowski, L. et al.) 274\u0026ndash;285 (Springer Nature Switzerland, Cham, 2023). doi:10.1007/978-3-031-42505-9_24.\u003c/li\u003e\n\u003cli\u003eKuhn, M. Building predictive models in R using the caret package. \u003cem\u003eJ Stat Softw\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1\u0026ndash;26 (2008).\u003c/li\u003e\n\u003cli\u003eCan MRI-based multivariate gray matter volumetric distance predict motor progression and classify slow versus fast progressors in Parkinson\u0026rsquo;s disease? | medRxiv. https://www.medrxiv.org/content/10.1101/2022.07.25.22278012v1.\u003c/li\u003e\n\u003cli\u003eLundberg, S. M. \u0026amp; Lee, S. I. A Unified Approach to Interpreting Model Predictions. in \u003cem\u003eAdvances in Neural Information Processing Systems 30 (NIPS\u003c/em\u003e vol. 2017 4765\u0026ndash;4774 (2017).\u003c/li\u003e\n\u003cli\u003eAn artificial intelligence interpretable tool to predict risk of deep vein thrombosis after endovenous thermal ablation - Journal of Vascular Surgery: Venous and Lymphatic Disorders. https://www.jvsvenous.org/article/S2213-333X(25)00088-5/fulltext.\u003c/li\u003e\n\u003cli\u003eZhong, J., Li, M., Dai, Z. \u0026amp; Wan, J. Multi-omics reveals changes in astrocyte fatty acid metabolism during early stages of Alzheimer\u0026rsquo;s disease. Preprint at https://doi.org/10.21203/rs.3.rs-4976800/v1 (2024).\u003c/li\u003e\n\u003cli\u003eAbstract - 2020 - Movement Disorders - Wiley Online Library. https://movementdisorders.onlinelibrary.wiley.com/doi/10.1002/mds.28268.\u003c/li\u003e\n\u003cli\u003eKurki, M. I., Karjalainen, J. \u0026amp; Palta, P. FinnGen provides genetic insights from a well-phenotyped isolated population. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e613\u003c/strong\u003e, 508\u0026ndash;518 (2023).\u003c/li\u003e\n\u003cli\u003eHemani, G., Zheng, J. \u0026amp; Elsworth, B. The MR-Base platform supports systematic causal inference across the human phenome. \u003cem\u003eElife\u003c/em\u003e \u003cstrong\u003e7:e34408\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eCausal association of circulating immune cells and lymphoma: A Mendelian randomization study. https://www.degruyterbrill.com/document/doi/10.1515/med-2024-0984/html.\u003c/li\u003e\n\u003cli\u003eVerbanck, M., Chen, C. Y., Neale, B. \u0026amp; Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e50\u003c/strong\u003e, 693\u0026ndash;698 (2018).\u003c/li\u003e\n\u003cli\u003eChia, S.-Y. \u003cem\u003eet al.\u003c/em\u003e Single-nucleus transcriptomics reveals a distinct microglial state and increased MSR1-mediated phagocytosis as common features across dementia subtypes. \u003cem\u003eGenome Med.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 92 (2025).\u003c/li\u003e\n\u003cli\u003eBRAIN Initiative Cell Census Network (BICCN). A multimodal cell census and atlas of the mammalian primary motor cortex. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e598\u003c/strong\u003e, 86\u0026ndash;102 (2021).\u003c/li\u003e\n\u003cli\u003eZheng, G. X. Y. \u003cem\u003eet al.\u003c/em\u003e Massively parallel digital transcriptional profiling of single cells. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 14049 (2017).\u003c/li\u003e\n\u003cli\u003eSelenoprotein P Deficiency Drives Hepatocellular Carcinoma Progression via Induction of Neutrophil Senescence and Immunosuppressive Microenvironment | bioRxiv. https://www.biorxiv.org/content/10.1101/2025.06.24.661430v1.\u003c/li\u003e\n\u003cli\u003eWishart, D. S. \u003cem\u003eet al.\u003c/em\u003e DrugBank 5.0: a major update to the DrugBank database for 2018. \u003cem\u003eNucleic Acids Res.\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, D1074\u0026ndash;D1082 (2018).\u003c/li\u003e\n\u003cli\u003eNowell, J., Blunt, E., Gupta, D. \u0026amp; Edison, P. Antidiabetic agents as a novel treatment for Alzheimer\u0026rsquo;s and Parkinson\u0026rsquo;s disease. \u003cem\u003eAgeing Res Rev\u003c/em\u003e \u003cstrong\u003e89\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Parkinson’s disease with mild cognitive impairment, Multi-omics, Machine learning, Mendelian Randomization, Drug Repurposing, Neural Connectopathy","lastPublishedDoi":"10.21203/rs.3.rs-7937441/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7937441/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eParkinson\u0026rsquo;s disease with mild cognitive impairment (PD-MCI) represents a devastating outcome for patients, yet its pathogenic drivers remain poorly understood, and predictive biomarkers are critically lacking. This study aims to establish an early diagnostic model for PD-MCI (from PD), delineate causal mechanisms and identify translational opportunities for PD-MCI.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eWe established an integrated analytical pipeline using the longitudinal Parkinson\u0026rsquo;s Progression Markers Initiative (PPMI) cohort.\u003csup\u003e1\u003c/sup\u003e Our framework combined plasma transcriptomics and cerebrospinal fluid proteomics with systems biology, machine learning, longitudinal proteomic trajectories analysis, clinical correlation analysis, single-nucleus RNA sequencing, Mendelian randomization, and computational drug repurposing.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eWe identified a molecular signature of PD-MCI progression, revealing profound dysregulation of pathways governing neural connectivity\u0026mdash;axon guidance, synaptogenesis, and neuron projection development\u0026mdash;suggesting a \"neural connectopathy\" mechanism. Longitudinal proteomic trajectories analysis revealed that the pattern of dysregulated proteins changes years before PD-MCI diagnosis, underscoring an early pathological window. Gene-interaction network and protein-interaction network highlighted key hubs. We then implemented LASSO analysis to select key features and developed an interpretable machine learning model (Logistic Regression, transcripts\u0026thinsp;+\u0026thinsp;proteomics\u0026thinsp;+\u0026thinsp;demographics, AUC\u0026thinsp;=\u0026thinsp;0.74) that accurately predicted cognitive decline, with SHAP analysis to interpret the order. Crucially, Mendelian randomization established putative causal drivers of PDD risk, and single-nucleus sequencing anchored these findings to the brain, showing altered expression of the causal candidate. specifically in excitatory neurons. Finally, computational drug repurposing nominated targeted candidates, including MEDRONIC ACID, ZILEUTON and ATENOLOL, for immediate therapeutic consideration.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eOur study delivers an integrated translational advance by: (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e) defining \"neural connectopathy\" as a core disease mechanism in PD-MCI; (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) establish an efficient machine learning model to early diagnose PD-MCI from PD; (3) providing a clinically actionable predictive tool, fingding causal genes and proposing a pathway to treatment via drug repurposing. This end-to-end framework establishes a foundation for early prediction, biological insight, and targeted intervention in PD-MCI.\u003c/p\u003e","manuscriptTitle":"From Neural Connectopathy to a Therapeutic Path: An Integrated Multi-omics Framework Identifies a Causal Gene and Drug Candidates for Parkinson's Disease with mild cognitive impairment","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-18 08:46:34","doi":"10.21203/rs.3.rs-7937441/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cd504329-f368-4b58-a18d-845f38e80529","owner":[],"postedDate":"November 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":58069180,"name":"Biological sciences/Computational biology and bioinformatics"},{"id":58069181,"name":"Health sciences/Diseases"},{"id":58069182,"name":"Health sciences/Neurology"},{"id":58069183,"name":"Biological sciences/Neuroscience"}],"tags":[],"updatedAt":"2025-12-15T20:53:37+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-18 08:46:34","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7937441","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7937441","identity":"rs-7937441","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}