Multicomponent Mendelian randomization and machine learning studies of potential drug targets for neurodegenerative diseases

Multicomponent Mendelian randomization and machine learning studies of potential drug targets for neurodegenerative diseases | 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 Multicomponent Mendelian randomization and machine learning studies of potential drug targets for neurodegenerative diseases Xun Li, Jing Cai, Jinyan Xia, Meiling Zheng This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5848172/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 Neurodegenerative diseases (NDDs) remain a global health challenge. Alzheimer's disease (AD) and Parkinson's disease (PD) are the main types of NDDs worldwide, and Mendelian Randomization (MR) analysis across multi-omics and the entire genome offers novel strategies for identifying potential drug targets. This study used MR and summary-based MR(SMR) analysis to explore the causal relationship between genes and NDDs. Colocalization analysis and machine learning further validated and reinforced the MR findings. The pharmacological activity of candidate drug targets was confirmed via molecular docking and Molecular dynamics. This study revealed 14 genes that were closely associated with both NDDs. Specifically, IQCE(AD), HDHD2(AD), COMMD10(AD), ALPP (AD), FXYD6 (AD), STK3 (PD), LHFPL2 (PD), and ENPP4 (PD) were identified as risk factors for NDDs (OR > 1), whereas HEXIM2 (AD), TSC22D4 (AD), CHRNB1 (PD), BAG4 (PD), SLC25A1 (PD), and IL15 (PD) were protective factors (OR < 1). Molecular docking results revealed strong binding activities for PREDNISOLONE(ALPP = -7.6 kcal/mol), PANCURONIUM BROMIDE(CHRNB1 = -8 kcal/mol), CHEMBL379975(STK3 =-10.7 kcal/mol) and SIROLIMUS(IL15 = -9 kcal/mol). Molecular dynamics simulations confirmed the stable binding of the IL15-Sirolimus, ALPP-Prednisolone, STK3-CHEMBL379975, and CHRNB1-Rocuronium bromide complexes. This multi-omics study revealed 14 promising therapeutic targets for NDDs, providing new insights for targeted therapies and clinical strategies for NDDs. Our results provide evidence for future studies aimed at developing appropriate therapeutic interventions. Health sciences/Neurology/Neurological disorders/Neurodegenerative diseases/Alzheimers disease Health sciences/Neurology/Neurological disorders/Neurodegenerative diseases/Parkinsons disease Biological sciences/Drug discovery/Target identification Health sciences/Medical research Health sciences/Risk factors Neurodegenerative diseases Mendelian Randomization Machine Learning Drug targets Genetics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction With continuous advancements in medical and healthcare services, the life expectancy of the global elderly population has increased. However, the prevalence of Neurodegenerative diseases (NDDs) is expected to rise, leading to substantial disease burdens and critical public health challenges 1 . Alzheimer's disease (AD) and Parkinson's disease (PD) are the most common NDDs. The estimated prevalence of AD and PD is 585.20–782.73 cases per 100,000 and 91.20 -122.21 cases per 100,000, respectively 2 , 3 . These diseases place significant physical, emotional, and financial burdens on individuals, families, and society at large. Previous studies have identified risk factors for AD and PD, including ethnicity, genetic predispositions 4 , physical activity 5 , obesity, alcohol consumption, and smoking 6 . NDDs are characterized by disrupted neuronal connectivity and communication, leading to progressive functional loss in the central or peripheral nervous systems and resulting in impaired motor and cognitive functions 7 . However, the etiological factors underlying most NDDs remain unclear. Given the irreversible and progressive nature of neuronal damage, identifying modifiable risk factors is critically important. Traditional GWAS on NDDs have concentrated on analyzing single links between individual exposures and specific diseases or on Mendelian Randomization(MR) studies confined to single omics data. These approaches are constrained by data limitations, measurement errors, confounding variables, and biases, hindering the establishment of true directional causation and lacking external validation or druggability analysis of targeted therapeutics 1 , 8 – 10 . Thus, there is a need to enhance the methodological rigor and comprehensiveness of these studies. Recent advances in large-scale GWAS data provide an opportunity to address these limitations. As a modern epidemiological tool, MR employs genetic variants as instrumental variables to assess potential causal relationships between exposures and outcomes 11 . Summary data-based Mendelian Randomization (SMR) extends this concept further, focusing on the intricate relationships among genotypes, gene expression, and phenotypes 12 . Furthermore, SMR analysis enables multi-omics integration, aiding researchers in exploring potential causal links between specific drug targets and diseases. In this study, we utilized SMR and MR analyses to integrate large-scale GWAS data with expression QTL (eQTL), DNA methylation QTL (mQTL), protein QTL (pQTL), and splicing QTL(sQTL) data from human blood and brain tissues aiming to elucidate potential links between gene/protein expression and AD or PD risk. Colocalization analysis seeks to determine whether specific gene expression and disease regions share identical causal variants 13 , identifying genetic evidence linked to exposures and outcomes to strengthen MR findings. The application of machine learning aims to utilize hypothesis-flexible models to compare the relative importance of identified genes and disease exposures, thereby further solidifying the identification of gene-exposure relationships. The use of molecular docking facilitates the prediction of possible binding modes and capacities between genes and target drugs, aiding in the evaluation of druggability. The study employed eQTL, mQTL, sQTL, and pQTL datasets from human plasma and brain tissues for performing MR and SMR analyses. Colocalization analysis and machine learning were utilized to validate and strengthen the MR findings, while molecular docking and Molecular dynamics evaluated the binding efficacy of potential targeted drugs. This research provides a foundation for developing novel therapeutic strategies for NDDs. Methods Overall analysis plan We integrated eQTL, mQTL, sQTL, and pQTL datasets with outcome data (PD or AD) to conduct two-sample MR and univariate SMR analyses. Colocalization analysis was then performed to identify significant associations, followed by machine learning , Molecular docking and Molecular dynamics studies. The overall workflow of the multi-omics Mendelian randomization analysis is illustrated in Figure 1. Source of exposure data The mQTL data were obtained from McRae et al 14 and Qi, T. et al. 15 , derived from 1,160 brain and 1,980 blood samples of European individuals. The eQTL dataset from Võsa, U. et al 16 , includes 31,684 European blood samples. Additionally, the brain eQTL dataset from Qi, T. et al 15 , comprises 2,865 brain tissue samples and 755 blood samples from Europeans in the GTEx Consortium 17 . The sQTL data were derived from 755 European blood samples in the GTEx Consortium 17 and 2,865 European brain tissue samples from Qi, T. et al 15 . Lastly, the pQTL datasets were derived from three extensive studies by Folkersen et al 18 , Pietzner et al 19 , and Sun et al 20 , encompassing 3,301, 10,708, and 30,931 European blood samples, respectively, for aggregated genetic association analysis (Supplementary Table S1). It should be noted that part of the data in this study comes from brain tissue, while some human data are sourced from human plasma, which presents certain limitations. Source of o utcomes data The International Parkinson’s Disease Genomics Consortium provided genetic variation data for PD, which includes genome analyses of 482,730 Europeans, with 33,674 cases and 449,056 controls 21 . Summary statistics for AD GWAS were obtained from the FinnGen R9 Alzheimer's Genomics Project, which includes 392,423 Europeans with 9,301 cases and 383,122 controls 22 (Supplementary Table S2). Machine Learning Data Sources The GEO database (https://www.ncbi.nlm.nih.gov/gds) provided genetic variation data for AD from the GSE110226 study, analyzing the brain tissues of 20 European individuals through whole-genome sequencing. The GSE110226 study also supplied genetic variation data for PD, conducting whole-genome analyses of brain tissues from 18 Europeans (Supplementary Table S3). Summary data-based Mendelian Randomization Analysis The Summary Mendelian Randomization (SMR) approach utilizes single nucleotide polymorphisms (SNPs) to assess potential causal links between exposures and outcomes. This study used SNPs as instrumental variables, considering mQTL, eQTL, pQTL, and sQTL as exposures and AD and PD as outcomes 23 . Chromosomal windows (±1000 kb) centered on target genes were defined, applying thresholds of P < 5.0 × 10⁻⁸, MAF 1.57 × 10⁻³ to identify relevant cis-QTLs. The heterogeneity-dependent instrument (HEIDI) test assessed linkage disequilibrium effects, with P > 0.05 indicating no significant heterogeneity. A HEIDI P -value < 0.05 suggested potential heterogeneity. A false discovery rate (FDR)-corrected P -SMR < 0.05 was considered evidence of a significant causal relationship. Mendelian Randomization Analysis MR analysis was conducted to assess bidirectional causal relationships between eQTL, mQTL, sQTL, pQTL, and NDDs (AD and PD), using genetic variations as proxies for risk factors 24 . Quantitative trait loci were treated as exposures, while AD and PD served as outcome variables. For genes with at least three SNPs (nSNP ≥ 3), MR-Inverse Variance Weighting (MR-IVW) and Generalized SMR (GSMR) were applied to validate bidirectional effects. MR-Egger regression was used to detect potential pleiotropy. MR-IVW served as the primary analysis, with other methods providing supplementary validation. MR-Egger results with P > 0.05 were deemed statistically significant. Causal estimates were reported as odds ratios (OR) with 95% confidence intervals (CI) for a one-standard-deviation increase in exposure. FDR correction was applied, with P -MR-IVW and P -GSMR < 0.05 considered significant. Colocalization Analysis Colocalization analysis enhances genetic studies by identifying shared genetic variants linked to specific exposures and outcomes 25 , ensuring that associations arise from causal relationships rather than linkage disequilibrium (LD) or confounders. This analysis considers five hypotheses: H₀ (no association), H₁ (association with one trait), H₂ (association with the other trait), H₃ (association with both traits but different causal variants), and H₄ (association with both traits via the same causal variant). A posterior probability 4 (pph4) threshold of ≥0.50 suggests a strong genetic correlation, indicating potential shared causal variants between populations. Machine Learning Analysis Machine learning utilizes flexible models to evaluate the significance of identified genes in disease exposure 26 . This study applied seven models-Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Neural Networks (NNET), K-Nearest Neighbors (KNN), Least Absolute Shrinkage and Selection Operator (LASSO), and Generalized Linear Model (GLM) to identify disease-associated genes. Gene data from the GEO database was integrated, and the optimal model was selected based on its predictive performance. To prevent overfitting in machine learning models, bootstrap cross validation was employed to evaluate model robustness. A testing set comprising 40% of the dataset was selected, and 5 repeated samplings were conducted. Receiver Operating Characteristic (ROC) curves were used to assess model accuracy, with the area under the curve (AUC) serving as the primary evaluation metric. Residuals, representing the difference between predicted and actual values, further informed model selection. The model with the highest AUC was chosen, with residual-based metrics considered as secondary criteria. Druggability Assessment Genes identified through MR analysis were evaluated for druggability using the DGIdb 4.0 database (https://dgidb.org/), documenting associated drugs and their development stages. SwissADME(http://www.swissadme.ch/) online prediction platform to assess the pharmacokinetics and toxicity of potentially targeted drugs. Semi-flexible docking was conducted to form stable complexes. Protein preprocessing in PyMOL 2.4 involved removing water molecules and excess ligands while adding hydrogen atoms. PDBQT files were prepared using AutoDock Tools 1.5.6, and molecular docking analyses were performed with AutoDock Vina 1.2.2. Binding energies below -5 kcal/mol indicated effective ligand-receptor interactions, with values below -7 kcal/mol suggesting strong binding affinity 27 . Molecular dynamics Molecular dynamics (MD) simulations provide a realistic representation of the biological environment by incorporating factors such as temperature, pressure, and charge to assess receptor-ligand binding stability. A 100 ns MD simulation of the complexes was conducted using Gromacs 2022, with the Charmm36 and Gaff2 force fields for proteins and ligands, respectively 28 . The TIP3P water model was used to solvate the system within a 1.2 nm periodic boundary 29 . Electrostatic interactions were handled using the particle mesh Ewald (PME) method and the Verlet algorithm. Equilibration involved 100,000 steps of isothermal-isochoric (NVT) and isothermal-isobaric (NPT) simulations with a 0.1 ps coupling constant over 100 ps. Van der Waals and Coulomb interactions were calculated with a 1.0 nm cutoff. The system was then simulated at 310 K and 1 bar for 100 ns. Root mean square fluctuation (RMSF) measured amino acid flexibility, while root mean square deviation (RMSD) assessed conformational stability, with lower RMSD values indicating higher stability. Statistical Analysis Statistical analysis and data visualization were primarily performed using R version 4.4.1 and an online analytical platform SMR Portal(https://yanglab.westlake.edu.cn/smr-portal/). The "geni.plots" and "Coloc" packages were specifically used for colocalization analysis and graphical representations. The "ComplexHeatmap" package was utilized for circular heatmap generation, and "forestploter" was used for forest plot creation. The packages "caret," "DALEX," "ggplot2," "randomForest," "kernlab," and "pROC" were employed for machine learning analyses and associated visualizations. The "Venn Diagram" and "pheatmap" packages were used to generate Venn diagrams and heatmaps. Results Summary-Mendelian Randomization Analysis This study analysed 14,026 eQTLs, 16,107 mQTLs, 9,794 sQTLs, and 1,621 pQTLs causally linked to AD. For PD, We analysed 14,008 eQTLs, 16,086 mQTLs, 9,755 sQTLs, and 1,620 pQTLs with causal associations. SMR results indicate significant regulatory roles for the APOC1 gene, modulated by eQTLs, mQTLs, sQTLs, and pQTLs, and the APOE gene, influenced by mQTLs, sQTLs, and pQTLs in AD (-log10(pSMR) > 10) (Figure 2.A). For PD, significant regulatory impacts were observed for the SNCA gene, influenced by eQTLs, mQTLs, sQTLs, and pQTLs, as well as the KANSL1 and MAPT genes, modulated by eQTLs, mQTLs, and sQTLs (-log10(pSMR) > 8) (Figure 2. B). Mendelian Randomization Analysis A total of 6,596 genes for AD (nSNPs ≥ 3) and 6,202 genes for PD (nSNPs ≥ 3) were selected in this study. After FDR adjustment and SMR integration, 107 causally associated genes for AD were identified. (Supplementary Tables S3, S5; Figure 2.C). Of particular interest, the RASA4B gene in brain tissue, regulated by eQTLs and sQTLs, and the HDHD2 gene in blood, modulated by pQTLs and eQTLs (Figures 2.E). A total of 140 genes were found to have significant causal links to PD(Supplementary Tables S4, S6; Figure 2.D). Of particular interest, NT5DC2 in blood was regulated by mQTLs and eQTLs, while FCN1 was influenced by pQTLs and mQTLs, and MFGE8 by pQTLs and eQTLs (Figures 2.F). Seven genes were identified to co-regulate both AD and PD(Figures 2.G). These include VWDE and KCNA6, regulated by mQTLs in blood, INIP, TBK1, and THRA, regulated by eQTLs in blood, and CCDC25, regulated by eQTLs in brain tissue . The IQCE gene was eQTL-regulated in PD and mQTL-regulated in AD(Figures 2.H). These results reveal that the 107 AD-associated genes and 140 PD-associated genes identified in the initial analysis lacked reverse causal effects and were significantly linked to their respective diseases. Colocalization Analysis Colocalization analysis integrated results from SMR and MR to evaluate 107 AD-related genes and 140 PD-related genes. The analysis revealed that TSC22D4 (pph4=0.875) and FXYD6 (pph4=0.766) were among the genes with causal links to AD (Supplementary Table S7, Figure 3). For PD, colocalization analysis identified ENPP4 (pph4 = 0.561), CHRNB1 (pph4 = 0.588), SLC25A1 (pph4 = 0.518), and BAG4 (pph4 = 0.557) as genes with distinct shared genetic loci (Supplementary Table S8, Figure 3). Trajectory and colocalization plots for six genes are shown in Figure 4. Furthermore, the mQTL-TSC22D4(SMR_OR=0.9636, MR-IVW=0.9550) for AD and eQTL-BAG4(SMR_OR=0.9625, MR-OR=0.9672) for PD were linked to decreased disease risk. In contrast, increased disease risk was associated with mQTL-FXYD6 in AD (SMR_OR = 1.0964, MR-IVW = 1.0748) and eQTL-ENPP4 in PD (SMR_OR = 1.1286, MR_OR = 1.1205). Contrasting results were observed for eQTL-CHRNB1 (SMR_OR = 1.2100, MR_OR = 0.8255) and eQTL-SLC25A1 (SMR_OR = 1.2394, MR_OR = 0.8766) in PD, likely reflecting methodological differences between SMR and MR analyses. (Figure 3). Consequently, these six genes may be linked to the development of both NDDs in this region; however, existing evidence favors the presence of independent causal variants. Machine Learning Analysis Utilizing systematically curated data from the GEO database, this study applied seven machine learning models to 107 candidate genes associated with AD and 140 candidate genes linked to PD. The AD-associated GLM model demonstrated the largest area under the ROC curve (AUC=1.0) and minimal residual and cumulative residual area (Figure 4.A.B.C). Bootstrap Cross Validation show that the accuracy (ACU=0.833) of GLM models (Figure 4.I). Consequently, the GLM model identified five relatively significant genes for AD: ALPP, HEXIM2, IQCE, HDHD2, and COMMD10 (Supplementary Table S9, Figure 4.G). Analysis for PD indicated that the RF model achieved a superior ROC curve area (AUC=0.8) while maintaining low residuals and minimal cumulative residual area (Figure 4.D.E.F). Bootstrap Cross Validation show that the accuracy (ACU=0.607) of GLM models (Figure 4.J). The RF model pinpointed five relatively significant genes for PD: IL15, STK3, LHFPL2, CHRNB1, and BAG4 (Supplementary Table S9, Figure 4.H). The results underscore the precision of the GLM model for AD and the RF model for PD, reinforcing the reliability of the MR analyses. Druggability Analysis Querying the DGIdb database for potential target drugs with 14 key genes (Supplementary Table S10), of which 4 genes have been targeted for clinical drug development. CHRNB1 has multiple approved targeting drugs, including skeletal muscle relaxants, anti-rheumatic drugs, and anti-inflammatory agents. A drug targeting IL15 has been applied to PD treatment, while other IL15-targeting drugs have been utilized as immunosuppressants and for macular degeneration, especially in ophthalmic immunosuppression. Drugs targeting STK3 have been developed as anticancer agents for oncology treatments. Drugs targeting ALPP serve as corticosteroid anti-inflammatories, antihypertensives, and therapies for erectile dysfunction (Supplementary Table S10). Pharmacokinetic analysis revealed that 12 candidate drugs were capable of crossing the blood–brain barrier and exhibited gastrointestinal solubility, suggesting high drugability (Supplementary Table S12). Toxicity analysis indicated that 17 candidate drugs did not violate drug-likeness criteria, and 1 candidate drug no structural alerts, supporting a favorable safety profile (Supplementary Table S12). Molecular docking analysis indicated that PREDNISOLONE (ALPP = -7.6 kcal/mol), PANCURONIUM BROMIDE (CHRNB1 = -8 kcal/mol), CHEMBL379975 (STK3 = -10.7 kcal/mol), and SIROLIMUS (IL15 = -9 kcal/mol) exhibited the best binding energies, highlighting them as strong ligand-protein binding candidates (Supplementary Table S11), Figure 5.A.B.C.D). At present, no additional information is available for drugs targeting the other identified genes. Molecular dynamics Molecular dynamics analysis of RMSF, RMSD, and hydrogen bond counts validated the docking reliability of IL15–Sirolimus, ALPP–Prednisolone, STK3–CHEMBL379975, and CHRNB1–Rocuronium bromide complexes. As shown in Figure 5.H, RMSF values remained low (mostly below 4 Å), indicating limited flexibility and high structural stability. Figure 5.E reveals that the IL15–Sirolimus and STK3–CHEMBL379975 complexes reached equilibrium at 10 ns, fluctuating around 5 Å and 2.6 Å, respectively. The ALPP–Prednisolone complex stabilized at 90 ns (≈3.6 Å fluctuation), while CHRNB1–Rocuronium bromide equilibrated at 80 ns (≈11.7 Å fluctuation). The STK3–CHEMBL379975 complex exhibited the lowest RMSD, suggesting strong stability upon binding. Additionally, the radius of gyration (Rg) remained stable throughout the simulation, confirming that all four compounds maintained their conformations within their respective complexes without significant structural contraction or expansion (Figure 5.F). Hydrogen bonding is essential for ligand-protein binding. Figure 5.G illustrates the hydrogen bond interactions between small molecules and target proteins during the simulation. The IL15–Sirolimus complex forms 0 to 4 hydrogen bonds, with an average of ~2. The ALPP–Prednisolone complex forms 0 to 6 hydrogen bonds, averaging ~2. The STK3–CHEMBL379975 complex forms 0 to 5 hydrogen bonds, with an average of ~3. These results suggest favorable hydrogen bond interactions (Figure 5.G). In summary, the IL15–Sirolimus, ALPP–Prednisolone, STK3–CHEMBL379975, and CHRNB1–Rocuronium bromide complexes show stable binding. The STK3–CHEMBL379975 complex stands out with the lowest RMSD and strong hydrogen bonding, indicating that CHEMBL379975 binds more favorably to STK3 than the other compounds. Discussion NDDs significantly impact patients' quality of life and lifespan. Current pharmacological and surgical interventions are insufficient to meet clinical therapeutic needs. Therefore, the development of effective, targeted pharmacotherapies is urgently required. This study represents the combined use of QTL analysis, machine learning, molecular docking and molecular dynamics simulations to systematically identify molecular targets with potential neuroprotective effects against NDDs. The study identify two plasma eQTLs (HDHD2, COMMD10), three plasma pQTLs (ALPP, HDHD2, HEXIM2), and three plasma mQTLs (IQCE, TSC22D4, FXYD6) linked to AD risk. Additionally, the research establishes four plasma eQTLs (LHFPL2, CHRNB1, ENPP4, SLC25A1), one brain eQTL (IL15), one brain mQTL (BAG4), and one plasma mQTL (STK3) related to PD risk, offering new insights into genetic predispositions for NDDs (Supplementary Table S13). Molecular docking indicates that PREDNISOLONE (ALPP), PANCURONIUM BROMIDE (CHRNB1), CHEMBL379975 (STK3), and SIROLIMUS (IL15) are the most viable drug candidates for the treatment of both NDDs (Supplementary Table S13). Molecular dynamics simulations confirmed the stable binding of the IL15–Sirolimus, ALPP–Prednisolone, STK3–CHEMBL379975, and CHRNB1–Rocuronium bromide complexes. This study has elucidated several key characteristics of NDDs, ranging from genetic determinants to targeted drug pathways, providing a solid foundation for NDD research and guiding both basic and translational investigations into therapeutic drug development. Based on AUC and residual values, this study found that the GLM is suitable for screening target genes in AD, while RF is more appropriate for screening target genes in PD. GLM is an extension of traditional linear regression, capable of handling more complex situations. Unlike linear regression, GLM does not require a linear relationship between the dependent and independent variables, nor does it assume that the dependent variable follows a normal distribution 30 . In simple terms, GLM provides greater flexibility in handling a wider variety of data types, enabling the development of effective regression models for more complex scenarios. The core concept of RF is to construct multiple decision trees (each acting as a 'weak classifier') and combine their predictions to form a more accurate and stable 'strong classifier' This approach effectively reduces overfitting, thereby improving the model's ability to predict new data and enhancing its generalization capability 31 . In simple terms, while a single decision tree may overfit the training data, combining the results of multiple trees helps balance individual errors and yields a more robust prediction. Studies in PD patients have shown that Parkin and NIX support the formation of memory T cells by being upregulated in response to interleukin 15 (IL-15). IL-15, a cytokine involved in the survival and differentiation of T cells, stimulates the expression of Parkin and NIX, which are essential for maintaining mitochondrial integrity and regulating cellular energy. The upregulation of these proteins facilitates the metabolic and functional adaptations required for the generation and persistence of memory T cells, which are crucial for the adaptive immune response. Significant alterations in IL-15 have been observed in both the substantia nigra and striatum of patients with clinical PD 32-35 . The MR analysis indicates a protective role of IL15 in PD progression, consistent with foundational research outcomes. The IL-15-targeting drug LEVODOPA has been clinically approved for PD treatment. Molecular docking results suggest that SIROLIMUS may exhibit stronger ligand-protein binding compared to LEVODOPA (Supplementary Table S11, S13). The inhibition of mTOR activity by SIROLIMUS leads to autophagy activation. In a model of synucleinopathy, SIROLIMUS decreased α-synuclein accumulation, indicating that sirolimus treatment could prevent α-synuclein-induced neurodegeneration 36 . Through the Hippo pathway’s Mst1/2 (STK3/STK4), STK3 modulates autophagy under mitochondrial stress, maintaining mitochondrial stability and cellular integrity. In PD models, reduced Mst1 (STK3) expression helps mitigate the loss of TH-positive neurons, improving behavioral deficits and mitochondrial function 37 , further supporting its identification as a PD risk factor. CHEMBL379975 demonstrates the most optimal binding mode for targeting STK3 (Supplementary Table S11,S13). ALPP functions as a positive regulator of placental growth and is involved in essential cellular processes, including protein phosphorylation, cell growth, apoptosis, and migration during embryonic development. While no studies have yet established a direct association between ALPP and AD, it has been linked to increased disease risk. PREDNISOLONE treatment improves amyloid-beta (Aβ)-induced cognitive deficits in AD mice and inhibits microglial activation in the cortex and hippocampus. RNA sequencing analysis revealed that PREDNISOLONE ultimately salvages cognitive dysfunction by improving synaptic function and inhibiting immune and inflammatory processes 38 . PREDNISOLONE, which exhibits the strongest molecular affinity for ALPP, demonstrates significant therapeutic potential but requires further validation through fundamental research. CHRNB1 encodes the β subunit of the acetylcholine receptor at the neuromuscular junction, with mutations in this gene linked to congenital myasthenic syndrome (CMS) 39,40 . A potential relationship exists between CHRNB1 and tremor symptoms in PD, and our findings indicate a negative correlation with PD risk. Despite strong colocalization evidence, the precise mechanisms through which CHRNB1 mitigates PD risk remain to be determined. Among potential drugs, PANCURONIUM BROMIDE shows the highest promise for targeting CHRNB1 (Supplementary Table S11,S13). The research found that neuronal expression of HEXIM1 mRNA is closely associated with the pathology of AD in humans. Furthermore, HEXIM1's regulation of P-TEFb significantly influences the rapid induction of neuronal gene transcription, particularly in response to repeated depolarization. The HEXIM1/P-TEFb complex plays a crucial role in balancing the robust activation of genes necessary for setting and resetting synaptic plasticity 41,42 . TSC22D4, located within the NYAP1 locus in cortical samples from AD patients, is regulated by the APOE transcription factors (THRA and JUN) during cellular resting states. It participates in the formation of complexes with BRI2 and BRI3, which inhibit Aβ production and aggregation, thereby contributing to the regulation of Aβ-related processes in AD 43 . IQCE is involved in promoting Hedgehog signaling, which plays a crucial role in neuronal differentiation and regeneration. It is upregulated in association with favorable neuroblastoma event-free survival. IQCE protects post-mitotic neurons from amyloid-beta peptide-induced re-entry into the cell cycle and subsequent apoptosis, thereby supporting neuronal survival and mitigating neurodegeneration 44,45 . The protein encoded by the gene of unknown function, HDHD2, exhibits differential expression in the brains of Pon1-null mice fed a high-homocysteine, high-methionine diet. HDHD2 expression becomes dependent on the Pon1 genotype, with an increase of 1.28 to 1.55 times ( P < 0.05) in Pon1-null animals, suggesting a potential involvement in neurodegenerative processes in response to diet-induced metabolic changes 46 . COMMD10, a gene containing a COMM domain, significantly decreases with age in African green monkeys. It inhibits nuclear factor kappa B, a central transcription factor that regulates the expression of inflammatory genes. COMMD10 is also associated with the synthesis of TYW1B, which plays a role in wobble uridine synthesis. Additionally, GWAS have linked COMMD10 to tau, suggesting its potential involvement in neurodegeneration and dysregulation of cell death during aging 47 . FXYD6 mRNA is essential for dendritic localization, and the loss of this localization is associated with impaired Na+/K+-ATPase (NKA) function in dendrites, while NKA function in somatic cells remains unaffected. Additionally, FXYD6 expression is decreased in the brains of Tg2576 mice and human hippocampal tissues, suggesting that reduced FXYD6 expression may be detrimental to neurons. The dysfunction of Na+/K+-ATPase due to low FXYD6 expression may lead to disrupted calcium balance, contributing to neurodegeneration by disturbing calcium homeostasis in NDDs 48,49 . LHFPL2 protein is abundantly expressed in malignant brain tissues and may play a role in linking cancer and PD genetically, potentially through interactions with TPM1. Short-term LPS treatment results in the downregulation of several genes associated with immune cell differentiation, including LHFPL2, suggesting a potential role for LHFPL2 in immune regulation and its involvement in the pathophysiology of NDDs like PD 50 . BAG4 functions as a negative regulator of Parkin translocation and works in concert with HSPA1L to regulate Parkin's localization following mitochondrial damage in HeLa cells. Knockdown of HSPA1L significantly reduces Parkin translocation ( P < 0.01), while knockdown of BAG4 enhances Parkin translocation, specifically in a PINK1-dependent manner, promoting Parkin's relocation to damaged mitochondria 51 . Currently, research on the relationship between ENPP4 and PD is limited. ENPP4 and the PD-related gene PARK2 are located in the same chromosomal region, sparking interest in their potential association. However, existing studies have primarily focused on changes in gene expression and metabolic pathways, without investigating the specific role of ENPP4 in the pathogenesis of PD 52 . SLC25A1, a mitochondrial membrane transporter, is crucial for mitochondrial function regulation. Its expression is influenced by SUCLG1, which enhances mitochondrial mass and may increase SLC25A1 levels. Metabolite assays show a correlation between SLC25A1 expression and increased CA expression. In cells overexpressing SUCLG1, the addition of an SLC25A1 inhibitor partially restores PNF cell function, indicating that SUCLG1 affects PNF cell development through SLC25A1. In mice, abnormal SLC25A1 expression disrupts citrate/acetyl-CoA homeostasis, damages white matter integrity, and alters synaptic plasticity and morphology, potentially contributing to ASD-like phenotypes and proteomic changes 53-55 . Our study identified IQCE, HDHD2, COMMD10, and FXYD6 as risk factors for AD, while HEXIM1 and TSC22D4 were associated with a reduced risk of AD. BAG4 and SLC25A1 were identified as protective factors for PD, while ENPP4 and LHFPL2 were linked to increased PD risk. However, the mechanisms through which these QTLs influence the two NDDs remain unclear and require further investigation. This study's primary strength is the integration of eQTL, mQTL, sQTL, and pQTL data from plasma and brain tissues, facilitating a comprehensive MR analysis linking NDDs to gene expression, DNA methylation, protein levels, and RNA splicing. Additionally, SMR and MR analyses were employed for cross-validation. The advantage of this approach is its large sample size and extensive coverage, which reduce the risk of reverse causation and confounding biases. Furthermore, colocalization analysis and seven machine learning models were applied for result validation, mitigating biases from linkage disequilibrium and horizontal pleiotropy. Molecular docking of targeted drugs from the DGIdb 4.0 database was also conducted to evaluate their optimal binding modes, facilitating the prioritization of drug targets. Finally, all GWAS data used in this study were derived from populations of European ancestry, minimizing biases arising from different genetic backgrounds. While the study provides valuable insights, several limitations should be considered. First, the study sample is restricted to European populations, which may limit the generalizability of the findings to other ethnic groups. Future studies should aim to validate these results across diverse ethnic populations. Second, although molecular docking predicted potential interactions between drugs and their targets, the feasibility of these interactions requires further validation through in vitro and in vivo experiments 56 . Third, some eQTL, mQTL, sQTL, and pQTL data were derived from plasma, which may not fully capture brain-specific changes. Fourth, the biological significance of most identified drug targets for NDDs remains unclear, with their roles in the pathological progression of the two NDDs being inferred from their properties. Conclusion In this study, we performed large-scale MR analyses and utilized GWAS genetic data to uncover the causal relationships between QTL-regulated genes and NDDs, ultimately identifying 68 potential NDD-targeting drugs and 14 actionable NDDs targets derived from blood and/or brain tissues. We evaluated the effects of these 14 targets on two types of NDDs and provided preliminary insights into their potential mechanisms. Among these, PREDNISOLONE was identified as the most promising targeted drug for treating AD via ALPP, while PANCURONIUM BROMIDE, CHEMBL379975, and SIROLIMUS showed the highest potential for treating PD via CHRNB1, STK3, and IL15, respectively. Additionally, LEVODOPA, targeting the IL15 gene, has been approved for PD treatment. In conclusion, this research provides genetic support for causal links between QTL-regulated genes and NDDs, highlighting novel therapeutic targets that warrant further large-scale studies to confirm their practicality and safety. Overall, our research lays a strong groundwork for future studies on NDD pathogenesis, proposes new therapeutic strategies, and provides fresh insights into the treatment and investigation of NDDs. Declarations Competing interests The authors declare no competing interests. Funding This study was funded by The National Natural Science Foundation of China (NSFC; No. 82474605); Natural Science Foundation of Fujian(No. 2023J0202, XJG20230163). Author Contribution X.L. and J.C. conceptualized and designed the study. X.L. performed the statistical analyses. X.L. J.C. wrote the initial draft of the manuscript. X.L., M.L.Z., and J.Y.X. were involved in data curation, analysis, and visualization. J.C. provided supervision, funding acquisition, critical review, and editing of the manuscript. Acknowledgement The authors thank the participants and investigators of the studies used in this research. Data Availability The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author. References Li, D. et al. Associations of environmental factors with neurodegeneration: An exposome-wide Mendelian randomization investigation. Ageing Res. Rev. 95 , 102254. 10.1016/j.arr.2024.102254 (2024). Fu, Y. X. et al. Fluorescence imaging opens a new window for the diagnosis of early-stage Alzheimer's disease. Biosens. Bioelectron. 271 , 117051–117051 (2025). Zhang, J., Fan, Y., Liang, H. & Zhang, Y. Global, regional and national temporal trends in Parkinson’s disease incidence, disability-adjusted life year rates in middle-aged and older adults: a cross-national inequality analysis and Bayesian age-period-cohort analysis based on the global burden of disease 2021. Neurol. Sci. , 1–14 (2024). Shlomo, Y. B. et al. The epidemiology of Parkinson's disease. Lancet 403 , 283–292 (2024). Xinyuan, Z., Alberto, A. M. S. A. S. M., Xiang, G. & A. & Association of Diet and Physical Activity With All-Cause Mortality Among Adults With Parkinson Disease. JAMA Netw. open. 5 , e2227738–e2227738 (2022). Thomas, A. G., Mohan, M. & Thomas, R. A study on possible risk factors for progressive supranuclear palsy in southern part of India. Curr. J. Neurol. 23 , 66–73 (2024). Keonhee, K., Eunyee, J., Seungsoo, C. & Ah, S. D. 125 Contribution of Sensory Neuron-associated Macrophages to Neuropathic Pain via Up-regulation of Calcium Channel in C Fiber Sensory Neurons in Dorsal Root Ganglion: An Animal Model Study. Neurosurgery 70 , 26–26 (2024). Wu, P. F. et al. Assessment of causal effects of physical activity on neurodegenerative diseases: A Mendelian randomization study. J. Sport Health Sci. 10 , 454–461. 10.1016/j.jshs.2021.01.008 (2021). Tang, C. et al. Causal relationship between immune cells and neurodegenerative diseases: a two-sample Mendelian randomisation study. Front. Immunol. 15 , 1339649. 10.3389/fimmu.2024.1339649 (2024). Yusufujiang, A., Zeng, S. & Li, H. Cathepsins and Parkinson's disease: insights from Mendelian randomization analyses. Front. Aging Neurosci. 16 , 1380483. 10.3389/fnagi.2024.1380483 (2024). Lin, C., Chen, J. & Zhao, X. [Genetic Causation Analysis of Hyperandrogenemia Testing Indicators and Preeclampsia]. Sichuan da xue xue bao Yi xue ban = J. Sichuan Univ. Med. Sci. Ed. 55 , 566–573 (2024). Eilis, H. et al. An integrated epigenetic-genetic study of neuropathology in the Brains for Dementia Research cohort. Alzheimer's Dement. 16 (2020). Li, Y. et al. Mitochondrial dysfunction and onset of type 2 diabetes along with its complications: a multi-omics Mendelian randomization and colocalization study. Front. Endocrinol. 15 , 1401531–1401531 (2024). McRae, A. F. et al. Identification of 55,000 Replicated DNA Methylation QTL. Sci. Rep. 8 , 17605. 10.1038/s41598-018-35871-w (2018). Qi, T. et al. Genetic control of RNA splicing and its distinct role in complex trait variation. Nat. Genet. 54 , 1355–1363. 10.1038/s41588-022-01154-4 (2022). Võsa, U. et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nat. Genet. 53 , 1300–1310. 10.1038/s41588-021-00913-z (2021). The GTEx Consortium atlas of genetic regulatory effects across human tissues. Sci. (New York N Y) 369 , 1318–1330, doi: 10.1126/science.aaz1776 (2020). Folkersen, L. et al. Genomic and drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. Nat. metabolism . 2 , 1135–1148. 10.1038/s42255-020-00287-2 (2020). Pietzner, M. et al. Mapping the proteo-genomic convergence of human diseases. Sci. (New York N Y) . 374 , eabj1541. 10.1126/science.abj1541 (2021). Sun, B. B. et al. Genomic atlas of the human plasma proteome. Nature 558 , 73–79. 10.1038/s41586-018-0175-2 (2018). Nalls, M. A. et al. 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. 10.1016/s1474-4422(19)30320-5 (2019). Kurki, M. I. et al. Author Correction: FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 615 , E19. 10.1038/s41586-023-05837-8 (2023). Dai, L. et al. Identifying prioritization of therapeutic targets for ankylosing spondylitis: a multi-omics Mendelian randomization study. J. translational Med. 22 10.1186/s12967-024-05925-x (2024). Gao, Z. et al. Genetic prediction of blood metabolites mediating the relationship between gut microbiota and postpartum depression: A mendelian randomization study. J. Psychiatr. Res. 181 , 614–622. 10.1016/j.jpsychires.2024.12.025 (2024). Lin, Q. et al. Proteome-Wide Mendelian Randomization Analysis to Identify Potential Plasma Biomarkers and Therapeutic Targets for Epithelial Ovarian Cancer Subtypes. Int. J. women's health . 16 , 2263–2279. 10.2147/ijwh.S491414 (2024). Liu, Y., Wang, W., Cui, X., Lyu, J. & Xie, Y. Exploring Genetic Associations of 3 Types of Risk Factors With Ischemic Stroke: An Integrated Bioinformatics Study. Stroke 55 , 1619–1628. 10.1161/strokeaha.123.044424 (2024). Bai, Y. et al. Identification of drug targets for Sjogren's syndrome: multi-omics Mendelian randomization and colocalization analyses. Front. Immunol. 15 , 1419363. 10.3389/fimmu.2024.1419363 (2024). Jo, S., Kim, T., Iyer, V. G. & Im, W. CHARMM-GUI: a web-based graphical user interface for CHARMM. J. Comput. Chem. 29 , 1859–1865. 10.1002/jcc.20945 (2008). Linse, J. B. & Hub, J. S. Three- and four-site models for heavy water: SPC/E-HW, TIP3P-HW, and TIP4P/2005-HW. J. Chem. Phys. 154 , 194501. 10.1063/5.0050841 (2021). Liu, Q. et al. Development and Validation of Prediction Models for Incident Reversible Cognitive Frailty Based on Social-Ecological Predictors Using Generalized Linear Mixed Model and Machine Learning Algorithms: A Prospective Cohort Study. J. Appl. gerontology: official J. South. Gerontological Soc. 44 , 255–266. 10.1177/07334648241270052 (2025). Gupta, M. et al. Deep transfer learning hybrid techniques for precision in breast cancer tumor histopathology classification. Health Inform. Sci. Syst. 13 10.1007/s13755-025-00337-7 (2025). Franco, F. et al. Regulatory circuits of mitophagy restrict distinct modes of cell death during memory CD8(+) T cell formation. Sci. Immunol. 8 , eadf7579. 10.1126/sciimmunol.adf7579 (2023). Walker, D. G. et al. Altered Expression Patterns of Inflammation-Associated and Trophic Molecules in Substantia Nigra and Striatum Brain Samples from Parkinson's Disease, Incidental Lewy Body Disease and Normal Control Cases. Front. NeuroSci. 9 , 507. 10.3389/fnins.2015.00507 (2015). Rentzos, M. et al. Circulating interleukin-15 and RANTES chemokine in Parkinson's disease. Acta Neurol. Scand. 116 , 374–379. 10.1111/j.1600-0404.2007.00894.x (2007). Gangemi, S. et al. Effect of levodopa on interleukin-15 and RANTES circulating levels in patients affected by Parkinson's disease. Mediat. Inflamm. 12 , 251–253. 10.1080/09629350310001599701 (2003). Pucha, K. A. et al. Neuron-derived extracellular vesicles to examine brain mTOR target engagement with sirolimus in patients with multiple system atrophy. Parkinsonism Relat. Disord. 115 , 105821. 10.1016/j.parkreldis.2023.105821 (2023). Jeong, D. J. et al. The Mst1/2-BNIP3 axis is required for mitophagy induction and neuronal viability under mitochondrial stress. Exp. Mol. Med. 56 , 674–685. 10.1038/s12276-024-01198-y (2024). Sun, Y. et al. Methylprednisolone alleviates cognitive functions through the regulation of neuroinflammation in Alzheimer's disease. Front. Immunol. 14 , 1192940. 10.3389/fimmu.2023.1192940 (2023). Bakr, M. N., Takahashi, H. & Kikuchi, Y. CHRNA1 and its correlated-myogenesis/cell cycle genes are prognosis-related markers of metastatic melanoma. Biochem. Biophys. Rep. 33 , 101425. 10.1016/j.bbrep.2023.101425 (2023). Turk, S. et al. NK-cell dysfunction of acute myeloid leukemia in relation to the renin-angiotensin system and neurotransmitter genes. Open. Med. (Warsaw Poland) . 17 , 1495–1506. 10.1515/med-2022-0551 (2022). Htet, M. et al. HEXIM1 is correlated with Alzheimer's disease pathology and regulates immediate early gene dynamics in neurons. bioRxiv: preprint Serv. biology . 10.1101/2024.09.27.615234 (2024). Yang, Y. et al. HEXIM1 homodimer binds two sites on 7SK RNA to release autoinhibition for P-TEFb inactivation. bioRxiv: preprint Serv. biology . 10.1101/2024.10.10.617642 (2024). Brase, L., Yu, Y., McDade, E., Harari, O. & Benitez, B. A. Comparative gene regulatory networks modulating APOE expression in microglia and astrocytes. medRxiv: preprint Serv. health Sci. 10.1101/2024.04.19.24306098 (2024). Cai, Z., Jia, X., Liu, M., Yang, X. & Cui, L. Epigenome-wide DNA methylation study of whole blood in patients with sporadic amyotrophic lateral sclerosis. Chin. Med. J. 135 , 1466–1473. 10.1097/cm9.0000000000002090 (2022). Currie, D. et al. A Potential Prognostic Gene Signature Associated with p53-Dependent NTRK1 Activation and Increased Survival of Neuroblastoma Patients. Cancers 16 10.3390/cancers16040722 (2024). Suszyńska-Zajczyk, J., Luczak, M., Marczak, L. & Jakubowski, H. Inactivation of the paraoxonase 1 gene affects the expression of mouse brain proteins involved in neurodegeneration. J. Alzheimer's disease: JAD . 42 , 247–260. 10.3233/jad-132714 (2014). Negrey, J. D. et al. Transcriptional profiles in olfactory pathway-associated brain regions of African green monkeys: Associations with age and Alzheimer's disease neuropathology. Alzheimer's & dementia (New York, N. Y.) 8, e12358, (2022). 10.1002/trc2.12358 Shiina, N., Yamaguchi, K. & Tokunaga, M. RNG105 deficiency impairs the dendritic localization of mRNAs for Na+/K + ATPase subunit isoforms and leads to the degeneration of neuronal networks. J. neuroscience: official J. Soc. Neurosci. 30 , 12816–12830. 10.1523/jneurosci.6386-09.2010 (2010). George, A. J. et al. A serial analysis of gene expression profile of the Alzheimer's disease Tg2576 mouse model. Neurotox. Res. 17 , 360–379. 10.1007/s12640-009-9112-3 (2010). Hill-Burns, E. M. et al. Identification of genetic modifiers of age-at-onset for familial Parkinson's disease. Hum. Mol. Genet. 25 , 3849–3862. 10.1093/hmg/ddw206 (2016). Hasson, S. A. et al. High-content genome-wide RNAi screens identify regulators of parkin upstream of mitophagy. Nature 504 , 291–295. 10.1038/nature12748 (2013). Shen, L. & Dettmer, U. Alpha-Synuclein Effects on Mitochondrial Quality Control in Parkinson's Disease. Biomolecules 14 10.3390/biom14121649 (2024). Zhou, Z., Li, Q. & Huo, R. SUCLG1 promotes aerobic respiration and progression in plexiform neurofibroma. Int. J. Oncol. 66 10.3892/ijo.2024.5716 (2025). Cimadamore-Werthein, C. et al. Human mitochondrial carriers of the SLC25 family function as monomers exchanging substrates with a ping-pong kinetic mechanism. EMBO J. 43 , 3450–3465. 10.1038/s44318-024-00150-0 (2024). Rigby, M. J. et al. Increased expression of SLC25A1/CIC causes an autistic-like phenotype with altered neuron morphology. Brain: J. Neurol. 145 , 500–516. 10.1093/brain/awab295 (2022). Gui, J. et al. Identification of Brain Cell Type-Specific Therapeutic Targets for Glioma From Genetics. CNS Neurosci. Ther. 30 , e70185. 10.1111/cns.70185 (2024). Additional Declarations No competing interests reported. Supplementary Files SupplementaryFile.xlsx 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-5848172","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":444557288,"identity":"dfbe6d6e-c7eb-4aa2-b57c-01f095dd92c8","order_by":0,"name":"Xun Li","email":"","orcid":"","institution":"Institute of Integrative Chinese and Western Medicine of Fujian University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xun","middleName":"","lastName":"Li","suffix":""},{"id":444557289,"identity":"cd73e382-c6c2-474c-9e7f-e8c045b32f6c","order_by":1,"name":"Jing Cai","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAoklEQVRIiWNgGAWjYFACxgaDhAoJOXnStDw4Y2Fs2ECSPQ/bKhIZDhCrnL/9cENB4jyJBMYG5oePbhCjReJMYoNB4jaJPHYGNmPjHGK0GDBAtBQzNvCwSROnhf8hUMscicSGA0RrkQDZ0kCKFokbQFsSjkkYGzYT6xf+/vRnhj9q6uTk2ZsfPiZKCxCwGYApZiKVg9U+IEHxKBgFo2AUjEQAAMgtLvDe400aAAAAAElFTkSuQmCC","orcid":"","institution":"The Third People's Hospital of Fujian University of Traditional Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Jing","middleName":"","lastName":"Cai","suffix":""},{"id":444557290,"identity":"6ae8ed3d-71bb-40ae-ba74-43334d8b997d","order_by":2,"name":"Jinyan Xia","email":"","orcid":"","institution":"Institute of Integrative Chinese and Western Medicine of Fujian University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jinyan","middleName":"","lastName":"Xia","suffix":""},{"id":444557291,"identity":"c655990a-b5f7-47ce-babe-4215009d9941","order_by":3,"name":"Meiling Zheng","email":"","orcid":"","institution":"Institute of Integrative Chinese and Western Medicine of Fujian University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Meiling","middleName":"","lastName":"Zheng","suffix":""}],"badges":[],"createdAt":"2025-01-17 10:08:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5848172/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5848172/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80886764,"identity":"104be5bd-0a37-44b2-b686-72b1146ea831","added_by":"auto","created_at":"2025-04-18 09:06:50","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":784511,"visible":true,"origin":"","legend":"\u003cp\u003eMulti-omics Mendelian randomization study design. AD: Alzheimer's disease; PD: Parkinson's disease; PPH4: posterior probability of H4; nSNPs: number of SNPs; eQTL: expression quantitative trait loci; mQTL: DNA methylation quantitative trait loci; pQTL: protein quantitative trait loci; sQTL: splicing quantitative trait loci; MAF: minor allele frequency.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/9e362ca43a39f1388f109b37.png"},{"id":80886765,"identity":"afcd227c-217b-4fba-92da-f10ea3093ec3","added_by":"auto","created_at":"2025-04-18 09:06:50","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1355740,"visible":true,"origin":"","legend":"\u003cp\u003eThis figure shows the distribution of SNPs for QTLs in GWAS at -log10(SMR-Paval). ( A ) Representing the distribution for AD. ( B ) Representing the distribution for PD. QTL: quantitative trait loci. ( C ) Heatmap of Paval and OR values for 104 genes in Alzheimer's disease(AD), including Paval and OR values from SMR, GMSR, MR-IVW, and Pval values from MR-Egger. ( D ) Heatmap of Paval and OR values for 140 genes in Parkinson's disease(PD), including Paval and OR values from SMR, GMSR, MR-IVW, and Pval values from MR-Egger.Venn diagram of druggable target genes. ( E ) shows genes co-regulated by quantitative trait loci(QTL) in Alzheimer's disease(AD). ( F ) shows genes co-regulated by QTLs in Parkinson's disease(PD); ( G ) shows genes co-regulated in both PD and AD. ( H ) Heatmap of druggable target genes. QTL: quantitative trait loci.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/66825aaa70347a96c3f87033.png"},{"id":80886766,"identity":"355c59e1-488d-4e5f-a06e-70d918211c79","added_by":"auto","created_at":"2025-04-18 09:06:50","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":584532,"visible":true,"origin":"","legend":"\u003cp\u003eVisualization of colocalization analysis results for NDDs drug targets and QTL. Indicates shared genetic variant loci associated with specific exposure genes and AD or PD. (A, B, C, F) compare the loci co-localized with BAG4 and CHRNB1, ENPP4, SLC25A1 in PD susceptibility. (D, E) compare the loci co-localized with TSC22D4 and FXYD5 in AD susceptibility.QTL: quantitative trait loci, (G) Forest plot showing genes with significant co-localization with Parkinson's disease and Alzheimer's disease. AD: Alzheimer's disease, PD: Parkinson's disease, r2: correlation between SNP and top SNP.G. Forest plot of colocalization of druggable target exploration results. PPH4: posterior probability of H4, nSNPs: number of SNPs, OR: odds ratio, CI: confidence interval, FDR: false discovery rate, Chr: chromosome position, HEIDI: heterogeneity test p-valuee, QTL: quantitative trait loci.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/a97c9bd4045d9cb0bd77f722.png"},{"id":80886769,"identity":"54fd76b6-a541-4dcf-b6e2-cb744ed22434","added_by":"auto","created_at":"2025-04-18 09:06:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":539434,"visible":true,"origin":"","legend":"\u003cp\u003eMachine learning results for druggable targets. ( A, D)The Roc Curve Chart of AD and PD. ( B, E )The boxplots of residuals for 7 models in AD and PD. ( C, F )The cumulative area under the residual curves for 7 models in AD and PD. ( G, H). The gene importance for 7 models in AD and PD. ( I ) show the cross-validated ROC curve for AD. ( J ) show the cross-validated ROC curve for PD. SVM: Support Vector Machine, DT: Decision Tree, RF: Random Forest, NNET: Neural Networks, KNN: K-Nearest Neighbors, LASSO: Least Absolute Shrinkage and Selection Operator, GLM: Generalized Linear Model.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/38327a68e45d7bdfa697e2ab.png"},{"id":80886771,"identity":"1e8aa224-024e-456f-97f4-e2d09660ed3a","added_by":"auto","created_at":"2025-04-18 09:06:51","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":744926,"visible":true,"origin":"","legend":"\u003cp\u003eAvailable docking and validation results for proteins and small molecules. ( A ). ALPP docking PREDNISOLONE, ( B. ) CHRNB1 docking PANCURONIUM BROMIDE, ( C ). IL15 docking SIROLIMUS, ( D ). STK3 docking CHEMBL379975. ( E ) RMSD values of IL15-Sirolimus, ALPP-Prednisolone, STK3-Chembl379975 and CHRNB1-Rrocuronium_Bromid complex systems. ( F ) Rg values for IL15-Sirolimus, ALPP-Prednisolone, STK3-Chembl379975 and CHRNB1-Rrocuronium_Bromid complex systems. ( G ) Number of hydrogen bonds between small molecules and target proteins during the dynamics of the IL15-Sirolimus, ALPP-Prednisolone, STK3-Chembl379975 and CHRNB1-Rrocuronium_Bromid complex systems. ( H ) RMSF values of IL15-Sirolimus, ALPP-Prednisolone, STK3-Chembl379975 and CHRNB1-Rrocuronium_Bromid complex systems. RMSF: Root mean square fuctuation, RMSD: Root Mean Square Deviation, Rg:Radius of Gyration.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/e53d92f8035c2ea762992ea8.png"},{"id":88419112,"identity":"3929d0b8-64db-49e6-b693-fb808874fdc8","added_by":"auto","created_at":"2025-08-06 09:17:13","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4473684,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/79aca9b1-9e85-438d-8730-0672d6942bb0.pdf"},{"id":80886768,"identity":"1349dfd4-dbd5-4def-9279-f752fe47b7b0","added_by":"auto","created_at":"2025-04-18 09:06:50","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":159461,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFile.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-5848172/v1/0c2aa55cd68e94276efaec8e.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multicomponent Mendelian randomization and machine learning studies of potential drug targets for neurodegenerative diseases","fulltext":[{"header":"Introduction","content":"\u003cp\u003eWith continuous advancements in medical and healthcare services, the life expectancy of the global elderly population has increased. However, the prevalence of Neurodegenerative diseases (NDDs) is expected to rise, leading to substantial disease burdens and critical public health challenges \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Alzheimer's disease (AD) and Parkinson's disease (PD) are the most common NDDs. The estimated prevalence of AD and PD is 585.20\u0026ndash;782.73 cases per 100,000 and 91.20 -122.21 cases per 100,000, respectively \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e,\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. These diseases place significant physical, emotional, and financial burdens on individuals, families, and society at large.\u003c/p\u003e \u003cp\u003ePrevious studies have identified risk factors for AD and PD, including ethnicity, genetic predispositions \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e, physical activity \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e, obesity, alcohol consumption, and smoking \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. NDDs are characterized by disrupted neuronal connectivity and communication, leading to progressive functional loss in the central or peripheral nervous systems and resulting in impaired motor and cognitive functions \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. However, the etiological factors underlying most NDDs remain unclear. Given the irreversible and progressive nature of neuronal damage, identifying modifiable risk factors is critically important. Traditional GWAS on NDDs have concentrated on analyzing single links between individual exposures and specific diseases or on Mendelian Randomization(MR) studies confined to single omics data. These approaches are constrained by data limitations, measurement errors, confounding variables, and biases, hindering the establishment of true directional causation and lacking external validation or druggability analysis of targeted therapeutics \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Thus, there is a need to enhance the methodological rigor and comprehensiveness of these studies. Recent advances in large-scale GWAS data provide an opportunity to address these limitations. As a modern epidemiological tool, MR employs genetic variants as instrumental variables to assess potential causal relationships between exposures and outcomes \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Summary data-based Mendelian Randomization (SMR) extends this concept further, focusing on the intricate relationships among genotypes, gene expression, and phenotypes \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Furthermore, SMR analysis enables multi-omics integration, aiding researchers in exploring potential causal links between specific drug targets and diseases. In this study, we utilized SMR and MR analyses to integrate large-scale GWAS data with expression QTL (eQTL), DNA methylation QTL (mQTL), protein QTL (pQTL), and splicing QTL(sQTL) data from human blood and brain tissues aiming to elucidate potential links between gene/protein expression and AD or PD risk. Colocalization analysis seeks to determine whether specific gene expression and disease regions share identical causal variants \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, identifying genetic evidence linked to exposures and outcomes to strengthen MR findings. The application of machine learning aims to utilize hypothesis-flexible models to compare the relative importance of identified genes and disease exposures, thereby further solidifying the identification of gene-exposure relationships. The use of molecular docking facilitates the prediction of possible binding modes and capacities between genes and target drugs, aiding in the evaluation of druggability.\u003c/p\u003e \u003cp\u003eThe study employed eQTL, mQTL, sQTL, and pQTL datasets from human plasma and brain tissues for performing MR and SMR analyses. Colocalization analysis and machine learning were utilized to validate and strengthen the MR findings, while molecular docking and Molecular dynamics evaluated the binding efficacy of potential targeted drugs. This research provides a foundation for developing novel therapeutic strategies for NDDs.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eOverall analysis plan\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe integrated eQTL, mQTL, sQTL, and pQTL datasets with outcome data (PD or AD) to conduct two-sample MR and univariate SMR analyses. Colocalization analysis was then performed to identify significant associations, followed by machine learning , Molecular docking and Molecular dynamics studies. The overall workflow of the multi-omics Mendelian randomization analysis is illustrated in Figure 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSource of exposure data\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe mQTL data were obtained from McRae et al\u003csup\u003e14\u003c/sup\u003e and Qi, T. et al.\u003csup\u003e15\u003c/sup\u003e\u0026nbsp; , derived from 1,160 brain and 1,980 blood samples of European individuals. The eQTL dataset from V\u0026otilde;sa, U. et al\u003csup\u003e16\u003c/sup\u003e, includes 31,684 European blood samples. Additionally, the brain eQTL dataset from Qi, T. et al\u003csup\u003e15\u003c/sup\u003e, comprises 2,865 brain tissue samples and 755 blood samples from Europeans in the GTEx Consortium \u003csup\u003e17\u003c/sup\u003e. The sQTL data were derived from 755 European blood samples in the GTEx Consortium \u003csup\u003e17\u003c/sup\u003e and 2,865 European brain tissue samples from Qi, T. et al\u003csup\u003e15\u003c/sup\u003e. Lastly, the pQTL datasets were derived from three extensive studies by Folkersen et al\u003csup\u003e18\u003c/sup\u003e, Pietzner et al\u003csup\u003e19\u003c/sup\u003e, and Sun et al\u003csup\u003e20\u003c/sup\u003e, encompassing 3,301, 10,708, and 30,931 European blood samples, respectively, for aggregated genetic association analysis (Supplementary Table S1). It should be noted that part of the data in this study comes from brain tissue, while some human data are sourced from human plasma, which presents certain limitations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSource of\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;o\u003c/strong\u003e\u003cstrong\u003eutcomes data\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe International Parkinson\u0026rsquo;s Disease Genomics Consortium provided genetic variation data for PD, which includes genome analyses of 482,730 Europeans, with 33,674 cases and 449,056 controls \u003csup\u003e21\u003c/sup\u003e. Summary statistics for AD GWAS were obtained from the FinnGen R9 Alzheimer\u0026apos;s Genomics Project, which includes 392,423 Europeans with 9,301 cases and 383,122 controls \u003csup\u003e22\u003c/sup\u003e(Supplementary Table S2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning Data Sources\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe GEO database (https://www.ncbi.nlm.nih.gov/gds) provided genetic variation data for AD from the GSE110226 study, analyzing the brain tissues of 20 European individuals through whole-genome sequencing. The GSE110226 study also supplied genetic variation data for PD, conducting whole-genome analyses of brain tissues from 18 Europeans (Supplementary Table S3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSummary data-based Mendelian Randomization Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Summary Mendelian Randomization (SMR) approach utilizes single nucleotide polymorphisms (SNPs) to assess potential causal links between exposures and outcomes. This study used SNPs as instrumental variables, considering mQTL, eQTL, pQTL, and sQTL as exposures and AD and PD as outcomes\u003csup\u003e23\u003c/sup\u003e. Chromosomal windows (\u0026plusmn;1000 kb) centered on target genes were defined, applying thresholds of \u003cem\u003eP\u003c/em\u003e \u0026lt; 5.0 \u0026times; 10⁻⁸, MAF \u0026lt; 0.01, and \u003cem\u003eP\u003c/em\u003e-HEIDI \u0026gt; 1.57 \u0026times; 10⁻\u0026sup3; to identify relevant cis-QTLs. The heterogeneity-dependent instrument (HEIDI) test assessed linkage disequilibrium effects, with \u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05 indicating no significant heterogeneity. A HEIDI \u003cem\u003eP\u003c/em\u003e-value \u0026lt; 0.05 suggested potential heterogeneity. A false discovery rate (FDR)-corrected \u003cem\u003eP\u003c/em\u003e-SMR \u0026lt; 0.05 was considered evidence of a significant causal relationship.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMendelian Randomization Analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMR analysis was conducted to assess bidirectional causal relationships between eQTL, mQTL, sQTL, pQTL, and NDDs (AD and PD), using genetic variations as proxies for risk factors\u003csup\u003e24\u003c/sup\u003e. Quantitative trait loci were treated as exposures, while AD and PD served as outcome variables. For genes with at least three SNPs (nSNP \u0026ge; 3), MR-Inverse Variance Weighting (MR-IVW) and Generalized SMR (GSMR) were applied to validate bidirectional effects. MR-Egger regression was used to detect potential pleiotropy. MR-IVW served as the primary analysis, with other methods providing supplementary validation. MR-Egger results with \u003cem\u003eP\u003c/em\u003e \u0026gt; 0.05 were deemed statistically significant. Causal estimates were reported as odds ratios (OR) with 95% confidence intervals (CI) for a one-standard-deviation increase in exposure. FDR correction was applied, with \u003cem\u003eP\u003c/em\u003e-MR-IVW and \u003cem\u003eP\u003c/em\u003e-GSMR \u0026lt; 0.05 considered significant.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eColocalization Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eColocalization analysis enhances genetic studies by identifying shared genetic variants linked to specific exposures and outcomes\u003csup\u003e25\u003c/sup\u003e, ensuring that associations arise from causal relationships rather than linkage disequilibrium (LD) or confounders. This analysis considers five hypotheses: H₀ (no association), H₁ (association with one trait), H₂ (association with the other trait), H₃ (association with both traits but different causal variants), and H₄ (association with both traits via the same causal variant). A posterior probability 4 (pph4) threshold of \u0026ge;0.50 suggests a strong genetic correlation, indicating potential shared causal variants between populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine Learning Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMachine learning utilizes flexible models to evaluate the significance of identified genes in disease exposure\u003csup\u003e26\u003c/sup\u003e. This study applied seven models-Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), Neural Networks (NNET), K-Nearest Neighbors (KNN), Least Absolute Shrinkage and Selection Operator (LASSO), and Generalized Linear Model (GLM) to identify disease-associated genes. Gene data from the GEO database was integrated, and the optimal model was selected based on its predictive performance.\u0026nbsp;To prevent overfitting in machine learning models, bootstrap cross validation was employed to evaluate model robustness. A testing set comprising 40% of the dataset was selected, and 5 repeated samplings were conducted.\u0026nbsp;Receiver Operating Characteristic (ROC) curves were used to assess model accuracy, with the area under the curve (AUC) serving as the primary evaluation metric. Residuals, representing the difference between predicted and actual values, further informed model selection. The model with the highest AUC was chosen, with residual-based metrics considered as secondary criteria.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDruggability Assessment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGenes identified through MR analysis were evaluated for druggability using the DGIdb 4.0 database (https://dgidb.org/), documenting associated drugs and their development stages.\u0026nbsp;SwissADME(http://www.swissadme.ch/) online prediction platform to assess the pharmacokinetics and toxicity of potentially targeted drugs.\u0026nbsp;Semi-flexible docking was conducted to form stable complexes. Protein preprocessing in PyMOL 2.4 involved removing water molecules and excess ligands while adding hydrogen atoms. PDBQT files were prepared using AutoDock Tools 1.5.6, and molecular docking analyses were performed with AutoDock Vina 1.2.2. Binding energies below -5 kcal/mol indicated effective ligand-receptor interactions, with values below -7 kcal/mol suggesting strong binding affinity\u003csup\u003e27\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMolecular dynamics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMolecular dynamics (MD) simulations provide a realistic representation of the biological environment by incorporating factors such as temperature, pressure, and charge to assess receptor-ligand binding stability. A 100 ns MD simulation of the complexes was conducted using Gromacs 2022, with the Charmm36 and Gaff2 force fields for proteins and ligands, respectively\u003csup\u003e28\u003c/sup\u003e. The TIP3P water model was used to solvate the system within a 1.2 nm periodic boundary\u003csup\u003e29\u003c/sup\u003e. Electrostatic interactions were handled using the particle mesh Ewald (PME) method and the Verlet algorithm. Equilibration involved 100,000 steps of isothermal-isochoric (NVT) and isothermal-isobaric (NPT) simulations with a 0.1 ps coupling constant over 100 ps. Van der Waals and Coulomb interactions were calculated with a 1.0 nm cutoff. The system was then simulated at 310 K and 1 bar for 100 ns. Root mean square fluctuation (RMSF) measured amino acid flexibility, while root mean square deviation (RMSD) assessed conformational stability, with lower RMSD values indicating higher stability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analysis and data visualization were primarily performed using R version 4.4.1 and an online analytical platform SMR Portal(https://yanglab.westlake.edu.cn/smr-portal/). The \u0026quot;geni.plots\u0026quot; and \u0026quot;Coloc\u0026quot; packages were specifically used for colocalization analysis and graphical representations. The \u0026quot;ComplexHeatmap\u0026quot; package was utilized for circular heatmap generation, and \u0026quot;forestploter\u0026quot; was used for forest plot creation. The packages \u0026quot;caret,\u0026quot; \u0026quot;DALEX,\u0026quot; \u0026quot;ggplot2,\u0026quot; \u0026quot;randomForest,\u0026quot; \u0026quot;kernlab,\u0026quot; and \u0026quot;pROC\u0026quot; were employed for machine learning analyses and associated visualizations. The \u0026quot;Venn Diagram\u0026quot; and \u0026quot;pheatmap\u0026quot; packages were used to generate Venn diagrams and heatmaps.\u003c/p\u003e"},{"header":"Results","content":"\u003ch3\u003eSummary-Mendelian Randomization Analysis\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eThis study analysed 14,026 eQTLs, 16,107 mQTLs, 9,794 sQTLs, and 1,621 pQTLs causally linked to AD. For PD, We analysed 14,008 eQTLs, 16,086 mQTLs, 9,755 sQTLs, and 1,620 pQTLs with causal associations. SMR results indicate significant regulatory roles for the APOC1 gene, modulated by eQTLs, mQTLs, sQTLs, and pQTLs, and the APOE gene, influenced by mQTLs, sQTLs, and pQTLs in AD (-log10(pSMR) \u0026gt; 10) (Figure 2.A). For PD, significant regulatory impacts were observed for the SNCA gene, influenced by eQTLs, mQTLs, sQTLs, and pQTLs, as well as the KANSL1 and MAPT genes, modulated by eQTLs, mQTLs, and sQTLs (-log10(pSMR) \u0026gt; 8) (Figure 2. B).\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eMendelian Randomization Analysis\u003c/h3\u003e\n\u003cp\u003eA total of 6,596 genes for AD (nSNPs \u0026ge; 3) and 6,202 genes for PD (nSNPs \u0026ge; 3) were selected in this study. After FDR adjustment and SMR integration, 107 causally associated genes for AD were identified. (Supplementary Tables S3, S5; Figure 2.C). Of particular interest, the RASA4B gene in brain tissue, regulated by eQTLs and sQTLs, and the HDHD2 gene in blood, modulated by pQTLs and eQTLs (Figures 2.E). A total of 140 genes were found to have significant causal links to PD(Supplementary Tables S4, S6; Figure 2.D). Of particular interest, NT5DC2 in blood was regulated by mQTLs and eQTLs, while FCN1 was influenced by pQTLs and mQTLs, and MFGE8 by pQTLs and eQTLs (Figures 2.F). Seven genes were identified to co-regulate both AD and PD(Figures 2.G). These include VWDE and KCNA6, regulated by mQTLs in blood, INIP, TBK1, and THRA, regulated by eQTLs in blood, and CCDC25, regulated by eQTLs in brain tissue . The IQCE gene was eQTL-regulated in PD and mQTL-regulated in AD(Figures 2.H). These results reveal that the 107 AD-associated genes and 140 PD-associated genes identified in the initial analysis lacked reverse causal effects and were significantly linked to their respective diseases.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eColocalization Analysis\u003c/h3\u003e\n\u003cp\u003eColocalization analysis integrated results from SMR and MR to evaluate 107 AD-related genes and 140 PD-related genes. The analysis revealed that TSC22D4 (pph4=0.875) and FXYD6 (pph4=0.766) were among the genes with causal links to AD (Supplementary Table S7, Figure 3). For PD, colocalization analysis identified ENPP4 (pph4 = 0.561), CHRNB1 (pph4 = 0.588), SLC25A1 (pph4 = 0.518), and BAG4 (pph4 = 0.557) as genes with distinct shared genetic loci (Supplementary Table S8, Figure 3). Trajectory and colocalization plots for six genes are shown in Figure 4. Furthermore, the mQTL-TSC22D4(SMR_OR=0.9636, MR-IVW=0.9550) for AD and eQTL-BAG4(SMR_OR=0.9625, MR-OR=0.9672) for PD were linked to decreased disease risk. In contrast, increased disease risk was associated with mQTL-FXYD6 in AD (SMR_OR = 1.0964, MR-IVW = 1.0748) and eQTL-ENPP4 in PD (SMR_OR = 1.1286, MR_OR = 1.1205). Contrasting results were observed for eQTL-CHRNB1 (SMR_OR = 1.2100, MR_OR = 0.8255) and eQTL-SLC25A1 (SMR_OR = 1.2394, MR_OR = 0.8766) in PD, likely reflecting methodological differences between SMR and MR analyses. (Figure 3). Consequently, these six genes may be linked to the development of both NDDs in this region; however, existing evidence favors the presence of independent causal variants.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eMachine Learning Analysis\u003c/h3\u003e\n\u003cp\u003eUtilizing systematically curated data from the GEO database, this study applied seven machine learning models to 107 candidate genes associated with AD and 140 candidate genes linked to PD. The AD-associated GLM model demonstrated the largest area under the ROC curve (AUC=1.0) and minimal residual and cumulative residual area (Figure 4.A.B.C).\u0026nbsp;Bootstrap Cross Validation show that the accuracy (ACU=0.833) of GLM models (Figure 4.I).\u0026nbsp;Consequently, the GLM model identified five relatively significant genes for AD: ALPP, HEXIM2, IQCE, HDHD2, and COMMD10 (Supplementary Table S9, Figure 4.G). Analysis for PD indicated that the RF model achieved a superior ROC curve area (AUC=0.8) while maintaining low residuals and minimal cumulative residual area (Figure 4.D.E.F).\u0026nbsp;Bootstrap Cross Validation show that the accuracy (ACU=0.607) of GLM models (Figure 4.J).\u0026nbsp;The RF model pinpointed five relatively significant genes for PD: IL15, STK3, LHFPL2, CHRNB1, and BAG4 (Supplementary Table S9, Figure 4.H). The results underscore the precision of the GLM model for AD and the RF model for PD, reinforcing the reliability of the MR analyses.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eDruggability Analysis\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eQuerying the DGIdb database for potential target drugs with 14 key genes (Supplementary Table S10), of which 4 genes have been targeted for clinical drug development. CHRNB1 has multiple approved targeting drugs, including skeletal muscle relaxants, anti-rheumatic drugs, and anti-inflammatory agents. A drug targeting IL15 has been applied to PD treatment, while other IL15-targeting drugs have been utilized as immunosuppressants and for macular degeneration, especially in ophthalmic immunosuppression. Drugs targeting STK3 have been developed as anticancer agents for oncology treatments. Drugs targeting ALPP serve as corticosteroid anti-inflammatories, antihypertensives, and therapies for erectile dysfunction (Supplementary Table S10).\u0026nbsp;Pharmacokinetic analysis revealed that 12 candidate drugs were capable of crossing the blood\u0026ndash;brain barrier and exhibited gastrointestinal solubility, suggesting high drugability (Supplementary Table S12). Toxicity analysis indicated that 17 candidate drugs did not violate drug-likeness criteria, and 1 candidate drug no structural alerts, supporting a favorable safety profile (Supplementary Table S12).\u0026nbsp;Molecular docking analysis indicated that PREDNISOLONE (ALPP = -7.6 kcal/mol), PANCURONIUM BROMIDE (CHRNB1 = -8 kcal/mol), CHEMBL379975 (STK3 = -10.7 kcal/mol), and SIROLIMUS (IL15 = -9 kcal/mol) exhibited the best binding energies, highlighting them as strong ligand-protein binding candidates (Supplementary Table S11), Figure 5.A.B.C.D). At present, no additional information is available for drugs targeting the other identified genes.\u003c/p\u003e\n\u003ch3\u003eMolecular dynamics\u003c/h3\u003e\n\u003cp\u003eMolecular dynamics analysis of RMSF, RMSD, and hydrogen bond counts validated the docking reliability of IL15\u0026ndash;Sirolimus, ALPP\u0026ndash;Prednisolone, STK3\u0026ndash;CHEMBL379975, and CHRNB1\u0026ndash;Rocuronium bromide complexes. As shown in Figure 5.H, RMSF values remained low (mostly below 4 \u0026Aring;), indicating limited flexibility and high structural stability. Figure 5.E reveals that the IL15\u0026ndash;Sirolimus and STK3\u0026ndash;CHEMBL379975 complexes reached equilibrium at 10 ns, fluctuating around 5 \u0026Aring; and 2.6 \u0026Aring;, respectively. The ALPP\u0026ndash;Prednisolone complex stabilized at 90 ns (\u0026asymp;3.6 \u0026Aring; fluctuation), while CHRNB1\u0026ndash;Rocuronium bromide equilibrated at 80 ns (\u0026asymp;11.7 \u0026Aring; fluctuation). The STK3\u0026ndash;CHEMBL379975 complex exhibited the lowest RMSD, suggesting strong stability upon binding. Additionally, the radius of gyration (Rg) remained stable throughout the simulation, confirming that all four compounds maintained their conformations within their respective complexes without significant structural contraction or expansion (Figure 5.F). Hydrogen bonding is essential for ligand-protein binding. Figure 5.G illustrates the hydrogen bond interactions between small molecules and target proteins during the simulation. The IL15\u0026ndash;Sirolimus complex forms 0 to 4 hydrogen bonds, with an average of ~2. The ALPP\u0026ndash;Prednisolone complex forms 0 to 6 hydrogen bonds, averaging ~2. The STK3\u0026ndash;CHEMBL379975 complex forms 0 to 5 hydrogen bonds, with an average of ~3. These results suggest favorable hydrogen bond interactions (Figure 5.G). In summary, the IL15\u0026ndash;Sirolimus, ALPP\u0026ndash;Prednisolone, STK3\u0026ndash;CHEMBL379975, and CHRNB1\u0026ndash;Rocuronium bromide complexes show stable binding. The STK3\u0026ndash;CHEMBL379975 complex stands out with the lowest RMSD and strong hydrogen bonding, indicating that CHEMBL379975 binds more favorably to STK3 than the other compounds.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eNDDs significantly impact patients\u0026apos; quality of life and lifespan. Current pharmacological and surgical interventions are insufficient to meet clinical therapeutic needs. Therefore, the development of effective, targeted pharmacotherapies is urgently required. This study represents the combined use of QTL analysis, machine learning, molecular docking and molecular dynamics simulations to systematically identify molecular targets with potential neuroprotective effects against NDDs. The study identify two plasma eQTLs (HDHD2, COMMD10), three plasma pQTLs (ALPP, HDHD2, HEXIM2), and three plasma mQTLs (IQCE, TSC22D4, FXYD6) linked to AD risk. Additionally, the research establishes four plasma eQTLs (LHFPL2, CHRNB1, ENPP4, SLC25A1), one brain eQTL (IL15), one brain mQTL (BAG4), and one plasma mQTL (STK3) related to PD risk, offering new insights into genetic predispositions for NDDs (Supplementary Table S13). Molecular docking indicates that PREDNISOLONE (ALPP), PANCURONIUM BROMIDE (CHRNB1), CHEMBL379975 (STK3), and SIROLIMUS (IL15) are the most viable drug candidates for the treatment of both NDDs (Supplementary Table S13). Molecular dynamics simulations confirmed the stable binding of the IL15\u0026ndash;Sirolimus, ALPP\u0026ndash;Prednisolone, STK3\u0026ndash;CHEMBL379975, and CHRNB1\u0026ndash;Rocuronium bromide complexes. This study has elucidated several key characteristics of NDDs, ranging from genetic determinants to targeted drug pathways, providing a solid foundation for NDD research and guiding both basic and translational investigations into therapeutic drug development.\u003c/p\u003e\n\u003cp\u003eBased on AUC and residual values, this study found that the GLM is suitable for screening target genes in AD, while RF is more appropriate for screening target genes in PD. GLM is an extension of traditional linear regression, capable of handling more complex situations. Unlike linear regression, GLM does not require a linear relationship between the dependent and independent variables, nor does it assume that the dependent variable follows a normal distribution\u003csup\u003e30\u003c/sup\u003e. In simple terms, GLM provides greater flexibility in handling a wider variety of data types, enabling the development of effective regression models for more complex scenarios. The core concept of RF is to construct multiple decision trees (each acting as a \u0026apos;weak classifier\u0026apos;) and combine their predictions to form a more accurate and stable \u0026apos;strong classifier\u0026apos; This approach effectively reduces overfitting, thereby improving the model\u0026apos;s ability to predict new data and enhancing its generalization capability\u003csup\u003e31\u003c/sup\u003e. In simple terms, while a single decision tree may overfit the training data, combining the results of multiple trees helps balance individual errors and yields a more robust prediction.\u003c/p\u003e\n\u003cp\u003eStudies in PD patients have shown that Parkin and NIX support the formation of memory T cells by being upregulated in response to interleukin 15 (IL-15). IL-15, a cytokine involved in the survival and differentiation of T cells, stimulates the expression of Parkin and NIX, which are essential for maintaining mitochondrial integrity and regulating cellular energy. The upregulation of these proteins facilitates the metabolic and functional adaptations required for the generation and persistence of memory T cells, which are crucial for the adaptive immune response. Significant alterations in IL-15 have been observed in both the substantia nigra and striatum of patients with clinical PD\u003csup\u003e32-35\u003c/sup\u003e. The MR analysis indicates a protective role of IL15 in PD progression, consistent with foundational research outcomes. The IL-15-targeting drug LEVODOPA has been clinically approved for PD treatment. Molecular docking results suggest that SIROLIMUS may exhibit stronger ligand-protein binding compared to LEVODOPA (Supplementary Table S11, S13). The inhibition of mTOR activity by SIROLIMUS leads to autophagy activation. In a model of synucleinopathy, SIROLIMUS decreased \u0026alpha;-synuclein accumulation, indicating that sirolimus treatment could prevent \u0026alpha;-synuclein-induced neurodegeneration\u003csup\u003e36\u003c/sup\u003e.\u0026nbsp;Through the Hippo pathway\u0026rsquo;s Mst1/2 (STK3/STK4), STK3 modulates autophagy under mitochondrial stress, maintaining mitochondrial stability and cellular integrity. In PD models, reduced Mst1 (STK3) expression helps mitigate the loss of TH-positive neurons, improving behavioral deficits and mitochondrial function\u003csup\u003e37\u003c/sup\u003e, further supporting its identification as a PD risk factor. CHEMBL379975 demonstrates the most optimal binding mode for targeting STK3 (Supplementary Table S11,S13). ALPP functions as a positive regulator of placental growth and is involved in essential cellular processes, including protein phosphorylation, cell growth, apoptosis, and migration during embryonic development. While no studies have yet established a direct association between ALPP and AD, it has been linked to increased disease risk. PREDNISOLONE treatment improves amyloid-beta (A\u0026beta;)-induced cognitive deficits in AD mice and inhibits microglial activation in the cortex and hippocampus. RNA sequencing analysis revealed that PREDNISOLONE ultimately salvages cognitive dysfunction by improving synaptic function and inhibiting immune and inflammatory processes\u003csup\u003e38\u003c/sup\u003e. PREDNISOLONE, which exhibits the strongest molecular affinity for ALPP, demonstrates significant therapeutic potential but requires further validation through fundamental research. CHRNB1 encodes the \u0026beta; subunit of the acetylcholine receptor at the neuromuscular junction, with mutations in this gene linked to congenital myasthenic syndrome (CMS)\u003csup\u003e39,40\u003c/sup\u003e. A potential relationship exists between CHRNB1 and tremor symptoms in PD, and our findings indicate a negative correlation with PD risk. Despite strong colocalization evidence, the precise mechanisms through which CHRNB1 mitigates PD risk remain to be determined. Among potential drugs, PANCURONIUM BROMIDE shows the highest promise for targeting CHRNB1 (Supplementary Table S11,S13).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe research found that neuronal expression of HEXIM1 mRNA is closely associated with the pathology of AD in humans. Furthermore, HEXIM1\u0026apos;s regulation of P-TEFb significantly influences the rapid induction of neuronal gene transcription, particularly in response to repeated depolarization. The HEXIM1/P-TEFb complex plays a crucial role in balancing the robust activation of genes necessary for setting and resetting synaptic plasticity\u003csup\u003e41,42\u003c/sup\u003e. TSC22D4, located within the NYAP1 locus in cortical samples from AD patients, is regulated by the APOE transcription factors (THRA and JUN) during cellular resting states. It participates in the formation of complexes with BRI2 and BRI3, which inhibit A\u0026beta; production and aggregation, thereby contributing to the regulation of A\u0026beta;-related processes in AD\u003csup\u003e43\u003c/sup\u003e. IQCE is involved in promoting Hedgehog signaling, which plays a crucial role in neuronal differentiation and regeneration. It is upregulated in association with favorable neuroblastoma event-free survival. IQCE protects post-mitotic neurons from amyloid-beta peptide-induced re-entry into the cell cycle and subsequent apoptosis, thereby supporting neuronal survival and mitigating neurodegeneration\u003csup\u003e44,45\u003c/sup\u003e. The protein encoded by the gene of unknown function, HDHD2, exhibits differential expression in the brains of Pon1-null mice fed a high-homocysteine, high-methionine diet. HDHD2 expression becomes dependent on the Pon1 genotype, with an increase of 1.28 to 1.55 times (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05) in Pon1-null animals, suggesting a potential involvement in neurodegenerative processes in response to diet-induced metabolic changes\u003csup\u003e46\u003c/sup\u003e. COMMD10, a gene containing a COMM domain, significantly decreases with age in African green monkeys. It inhibits nuclear factor kappa B, a central transcription factor that regulates the expression of inflammatory genes. COMMD10 is also associated with the synthesis of TYW1B, which plays a role in wobble uridine synthesis. Additionally, GWAS have linked COMMD10 to tau, suggesting its potential involvement in neurodegeneration and dysregulation of cell death during aging\u003csup\u003e47\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFXYD6 mRNA is essential for dendritic localization, and the loss of this localization is associated with impaired Na+/K+-ATPase (NKA) function in dendrites, while NKA function in somatic cells remains unaffected. Additionally, FXYD6 expression is decreased in the brains of Tg2576 mice and human hippocampal tissues, suggesting that reduced FXYD6 expression may be detrimental to neurons. The dysfunction of Na+/K+-ATPase due to low FXYD6 expression may lead to disrupted calcium balance, contributing to neurodegeneration by disturbing calcium homeostasis in NDDs\u003csup\u003e48,49\u003c/sup\u003e. LHFPL2 protein is abundantly expressed in malignant brain tissues and may play a role in linking cancer and PD genetically, potentially through interactions with TPM1. Short-term LPS treatment results in the downregulation of several genes associated with immune cell differentiation, including LHFPL2, suggesting a potential role for LHFPL2 in immune regulation and its involvement in the pathophysiology of NDDs like PD\u003csup\u003e50\u003c/sup\u003e. BAG4 functions as a negative regulator of Parkin translocation and works in concert with HSPA1L to regulate Parkin\u0026apos;s localization following mitochondrial damage in HeLa cells. Knockdown of HSPA1L significantly reduces Parkin translocation (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.01), while knockdown of BAG4 enhances Parkin translocation, specifically in a PINK1-dependent manner, promoting Parkin\u0026apos;s relocation to damaged mitochondria\u003csup\u003e51\u003c/sup\u003e. Currently, research on the relationship between ENPP4 and PD is limited. ENPP4 and the PD-related gene PARK2 are located in the same chromosomal region, sparking interest in their potential association. However, existing studies have primarily focused on changes in gene expression and metabolic pathways, without investigating the specific role of ENPP4 in the pathogenesis of PD\u003csup\u003e52\u003c/sup\u003e. SLC25A1, a mitochondrial membrane transporter, is crucial for mitochondrial function regulation. Its expression is influenced by SUCLG1, which enhances mitochondrial mass and may increase SLC25A1 levels. Metabolite assays show a correlation between SLC25A1 expression and increased CA expression. In cells overexpressing SUCLG1, the addition of an SLC25A1 inhibitor partially restores PNF cell function, indicating that SUCLG1 affects PNF cell development through SLC25A1. In mice, abnormal SLC25A1 expression disrupts citrate/acetyl-CoA homeostasis, damages white matter integrity, and alters synaptic plasticity and morphology, potentially contributing to ASD-like phenotypes and proteomic changes\u003csup\u003e53-55\u003c/sup\u003e. Our study identified IQCE, HDHD2, COMMD10, and FXYD6 as risk factors for AD, while HEXIM1 and TSC22D4 were associated with a reduced risk of AD. BAG4 and SLC25A1 were identified as protective factors for PD, while ENPP4 and LHFPL2 were linked to increased PD risk. However, the mechanisms through which these QTLs influence the two NDDs remain unclear and require further investigation.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis study\u0026apos;s primary strength is the integration of eQTL, mQTL, sQTL, and pQTL data from plasma and brain tissues, facilitating a comprehensive MR analysis linking NDDs to gene expression, DNA methylation, protein levels, and RNA splicing. Additionally, SMR and MR analyses were employed for cross-validation. The advantage of this approach is its large sample size and extensive coverage, which reduce the risk of reverse causation and confounding biases. Furthermore, colocalization analysis and seven machine learning models were applied for result validation, mitigating biases from linkage disequilibrium and horizontal pleiotropy. Molecular docking of targeted drugs from the DGIdb 4.0 database was also conducted to evaluate their optimal binding modes, facilitating the prioritization of drug targets. Finally, all GWAS data used in this study were derived from populations of European ancestry, minimizing biases arising from different genetic backgrounds.\u003c/p\u003e\n\u003cp\u003eWhile the study provides valuable insights, several limitations should be considered. First, the study sample is restricted to European populations, which may limit the generalizability of the findings to other ethnic groups. Future studies should aim to validate these results across diverse ethnic populations. Second, although molecular docking predicted potential interactions between drugs and their targets, the feasibility of these interactions requires further validation through in vitro and in vivo experiments\u003csup\u003e56\u003c/sup\u003e. Third, some eQTL, mQTL, sQTL, and pQTL data were derived from plasma, which may not fully capture brain-specific changes. Fourth, the biological significance of most identified drug targets for NDDs remains unclear, with their roles in the pathological progression of the two NDDs being inferred from their properties.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, we performed large-scale MR analyses and utilized GWAS genetic data to uncover the causal relationships between QTL-regulated genes and NDDs, ultimately identifying 68 potential NDD-targeting drugs and 14 actionable NDDs targets derived from blood and/or brain tissues. We evaluated the effects of these 14 targets on two types of NDDs and provided preliminary insights into their potential mechanisms. Among these, PREDNISOLONE was identified as the most promising targeted drug for treating AD via ALPP, while PANCURONIUM BROMIDE, CHEMBL379975, and SIROLIMUS showed the highest potential for treating PD via CHRNB1, STK3, and IL15, respectively. Additionally, LEVODOPA, targeting the IL15 gene, has been approved for PD treatment. In conclusion, this research provides genetic support for causal links between QTL-regulated genes and NDDs, highlighting novel therapeutic targets that warrant further large-scale studies to confirm their practicality and safety. Overall, our research lays a strong groundwork for future studies on NDD pathogenesis, proposes new therapeutic strategies, and provides fresh insights into the treatment and investigation of NDDs.\u003c/p\u003e "},{"header":"Declarations","content":"\u003ch2\u003eCompeting interests\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis study was funded by The National Natural Science Foundation of China (NSFC; No. 82474605); Natural Science Foundation of Fujian(No. 2023J0202, XJG20230163).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eX.L. and J.C. conceptualized and designed the study. X.L. performed the statistical analyses. X.L. J.C. wrote the initial draft of the manuscript. X.L., M.L.Z., and J.Y.X. were involved in data curation, analysis, and visualization. J.C. provided supervision, funding acquisition, critical review, and editing of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors thank the participants and investigators of the studies used in this research.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLi, D. et al. Associations of environmental factors with neurodegeneration: An exposome-wide Mendelian randomization investigation. \u003cem\u003eAgeing Res. Rev.\u003c/em\u003e \u003cb\u003e95\u003c/b\u003e, 102254. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.arr.2024.102254\u003c/span\u003e\u003cspan address=\"10.1016/j.arr.2024.102254\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFu, Y. X. et al. Fluorescence imaging opens a new window for the diagnosis of early-stage Alzheimer's disease. \u003cem\u003eBiosens. Bioelectron.\u003c/em\u003e \u003cb\u003e271\u003c/b\u003e, 117051\u0026ndash;117051 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, J., Fan, Y., Liang, H. \u0026amp; Zhang, Y. Global, regional and national temporal trends in Parkinson\u0026rsquo;s disease incidence, disability-adjusted life year rates in middle-aged and older adults: a cross-national inequality analysis and Bayesian age-period-cohort analysis based on the global burden of disease 2021. \u003cem\u003eNeurol. Sci.\u003c/em\u003e, 1\u0026ndash;14 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShlomo, Y. B. et al. The epidemiology of Parkinson's disease. \u003cem\u003eLancet\u003c/em\u003e \u003cb\u003e403\u003c/b\u003e, 283\u0026ndash;292 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXinyuan, Z., Alberto, A. M. S. A. S. M., Xiang, G. \u0026amp; A. \u0026amp; Association of Diet and Physical Activity With All-Cause Mortality Among Adults With Parkinson Disease. \u003cem\u003eJAMA Netw. open.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, e2227738\u0026ndash;e2227738 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThomas, A. G., Mohan, M. \u0026amp; Thomas, R. A study on possible risk factors for progressive supranuclear palsy in southern part of India. \u003cem\u003eCurr. J. Neurol.\u003c/em\u003e \u003cb\u003e23\u003c/b\u003e, 66\u0026ndash;73 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeonhee, K., Eunyee, J., Seungsoo, C. \u0026amp; Ah, S. D. 125\u0026emsp;Contribution of Sensory Neuron-associated Macrophages to Neuropathic Pain via Up-regulation of Calcium Channel in C Fiber Sensory Neurons in Dorsal Root Ganglion: An Animal Model Study. \u003cem\u003eNeurosurgery\u003c/em\u003e \u003cb\u003e70\u003c/b\u003e, 26\u0026ndash;26 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, P. F. et al. Assessment of causal effects of physical activity on neurodegenerative diseases: A Mendelian randomization study. \u003cem\u003eJ. Sport Health Sci.\u003c/em\u003e \u003cb\u003e10\u003c/b\u003e, 454\u0026ndash;461. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jshs.2021.01.008\u003c/span\u003e\u003cspan address=\"10.1016/j.jshs.2021.01.008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, C. et al. Causal relationship between immune cells and neurodegenerative diseases: a two-sample Mendelian randomisation study. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1339649. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2024.1339649\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2024.1339649\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYusufujiang, A., Zeng, S. \u0026amp; Li, H. Cathepsins and Parkinson's disease: insights from Mendelian randomization analyses. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e, 1380483. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnagi.2024.1380483\u003c/span\u003e\u003cspan address=\"10.3389/fnagi.2024.1380483\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin, C., Chen, J. \u0026amp; Zhao, X. [Genetic Causation Analysis of Hyperandrogenemia Testing Indicators and Preeclampsia]. \u003cem\u003eSichuan da xue xue bao Yi xue ban = J. Sichuan Univ. Med. Sci. Ed.\u003c/em\u003e \u003cb\u003e55\u003c/b\u003e, 566\u0026ndash;573 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEilis, H. et al. An integrated epigenetic-genetic study of neuropathology in the Brains for Dementia Research cohort. \u003cem\u003eAlzheimer's Dement.\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y. et al. Mitochondrial dysfunction and onset of type 2 diabetes along with its complications: a multi-omics Mendelian randomization and colocalization study. \u003cem\u003eFront. Endocrinol.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1401531\u0026ndash;1401531 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcRae, A. F. et al. Identification of 55,000 Replicated DNA Methylation QTL. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 17605. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41598-018-35871-w\u003c/span\u003e\u003cspan address=\"10.1038/s41598-018-35871-w\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eQi, T. et al. Genetic control of RNA splicing and its distinct role in complex trait variation. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cb\u003e54\u003c/b\u003e, 1355\u0026ndash;1363. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41588-022-01154-4\u003c/span\u003e\u003cspan address=\"10.1038/s41588-022-01154-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eV\u0026otilde;sa, U. et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. \u003cem\u003eNat. Genet.\u003c/em\u003e \u003cb\u003e53\u003c/b\u003e, 1300\u0026ndash;1310. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41588-021-00913-z\u003c/span\u003e\u003cspan address=\"10.1038/s41588-021-00913-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThe GTEx Consortium atlas of genetic regulatory effects across human tissues. \u003cem\u003eSci. (New York N Y)\u003c/em\u003e \u003cb\u003e369\u003c/b\u003e, 1318\u0026ndash;1330, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.aaz1776\u003c/span\u003e\u003cspan address=\"10.1126/science.aaz1776\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFolkersen, L. et al. Genomic and drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. \u003cem\u003eNat. metabolism\u003c/em\u003e. \u003cb\u003e2\u003c/b\u003e, 1135\u0026ndash;1148. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s42255-020-00287-2\u003c/span\u003e\u003cspan address=\"10.1038/s42255-020-00287-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePietzner, M. et al. Mapping the proteo-genomic convergence of human diseases. \u003cem\u003eSci. (New York N Y)\u003c/em\u003e. \u003cb\u003e374\u003c/b\u003e, eabj1541. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/science.abj1541\u003c/span\u003e\u003cspan address=\"10.1126/science.abj1541\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun, B. B. et al. Genomic atlas of the human plasma proteome. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e558\u003c/b\u003e, 73\u0026ndash;79. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41586-018-0175-2\u003c/span\u003e\u003cspan address=\"10.1038/s41586-018-0175-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNalls, M. A. et al. Identification of novel risk loci, causal insights, and heritable risk for Parkinson's disease: a meta-analysis of genome-wide association studies. \u003cem\u003eLancet Neurol.\u003c/em\u003e \u003cb\u003e18\u003c/b\u003e, 1091\u0026ndash;1102. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/s1474-4422(19)30320-5\u003c/span\u003e\u003cspan address=\"10.1016/s1474-4422(19)30320-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKurki, M. I. et al. Author Correction: FinnGen provides genetic insights from a well-phenotyped isolated population. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e615\u003c/b\u003e, E19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41586-023-05837-8\u003c/span\u003e\u003cspan address=\"10.1038/s41586-023-05837-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDai, L. et al. Identifying prioritization of therapeutic targets for ankylosing spondylitis: a multi-omics Mendelian randomization study. \u003cem\u003eJ. translational Med.\u003c/em\u003e \u003cb\u003e22\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12967-024-05925-x\u003c/span\u003e\u003cspan address=\"10.1186/s12967-024-05925-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao, Z. et al. Genetic prediction of blood metabolites mediating the relationship between gut microbiota and postpartum depression: A mendelian randomization study. \u003cem\u003eJ. Psychiatr. Res.\u003c/em\u003e \u003cb\u003e181\u003c/b\u003e, 614\u0026ndash;622. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jpsychires.2024.12.025\u003c/span\u003e\u003cspan address=\"10.1016/j.jpsychires.2024.12.025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin, Q. et al. Proteome-Wide Mendelian Randomization Analysis to Identify Potential Plasma Biomarkers and Therapeutic Targets for Epithelial Ovarian Cancer Subtypes. \u003cem\u003eInt. J. women's health\u003c/em\u003e. \u003cb\u003e16\u003c/b\u003e, 2263\u0026ndash;2279. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2147/ijwh.S491414\u003c/span\u003e\u003cspan address=\"10.2147/ijwh.S491414\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, Y., Wang, W., Cui, X., Lyu, J. \u0026amp; Xie, Y. Exploring Genetic Associations of 3 Types of Risk Factors With Ischemic Stroke: An Integrated Bioinformatics Study. \u003cem\u003eStroke\u003c/em\u003e \u003cb\u003e55\u003c/b\u003e, 1619\u0026ndash;1628. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/strokeaha.123.044424\u003c/span\u003e\u003cspan address=\"10.1161/strokeaha.123.044424\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBai, Y. et al. Identification of drug targets for Sjogren's syndrome: multi-omics Mendelian randomization and colocalization analyses. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e, 1419363. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2024.1419363\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2024.1419363\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJo, S., Kim, T., Iyer, V. G. \u0026amp; Im, W. CHARMM-GUI: a web-based graphical user interface for CHARMM. \u003cem\u003eJ. Comput. Chem.\u003c/em\u003e \u003cb\u003e29\u003c/b\u003e, 1859\u0026ndash;1865. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/jcc.20945\u003c/span\u003e\u003cspan address=\"10.1002/jcc.20945\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2008).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLinse, J. B. \u0026amp; Hub, J. S. Three- and four-site models for heavy water: SPC/E-HW, TIP3P-HW, and TIP4P/2005-HW. \u003cem\u003eJ. Chem. Phys.\u003c/em\u003e \u003cb\u003e154\u003c/b\u003e, 194501. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1063/5.0050841\u003c/span\u003e\u003cspan address=\"10.1063/5.0050841\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu, Q. et al. Development and Validation of Prediction Models for Incident Reversible Cognitive Frailty Based on Social-Ecological Predictors Using Generalized Linear Mixed Model and Machine Learning Algorithms: A Prospective Cohort Study. \u003cem\u003eJ. Appl. gerontology: official J. South. Gerontological Soc.\u003c/em\u003e \u003cb\u003e44\u003c/b\u003e, 255\u0026ndash;266. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/07334648241270052\u003c/span\u003e\u003cspan address=\"10.1177/07334648241270052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGupta, M. et al. Deep transfer learning hybrid techniques for precision in breast cancer tumor histopathology classification. \u003cem\u003eHealth Inform. Sci. Syst.\u003c/em\u003e \u003cb\u003e13\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s13755-025-00337-7\u003c/span\u003e\u003cspan address=\"10.1007/s13755-025-00337-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranco, F. et al. Regulatory circuits of mitophagy restrict distinct modes of cell death during memory CD8(+) T cell formation. \u003cem\u003eSci. Immunol.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, eadf7579. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1126/sciimmunol.adf7579\u003c/span\u003e\u003cspan address=\"10.1126/sciimmunol.adf7579\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalker, D. G. et al. Altered Expression Patterns of Inflammation-Associated and Trophic Molecules in Substantia Nigra and Striatum Brain Samples from Parkinson's Disease, Incidental Lewy Body Disease and Normal Control Cases. \u003cem\u003eFront. NeuroSci.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e, 507. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnins.2015.00507\u003c/span\u003e\u003cspan address=\"10.3389/fnins.2015.00507\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRentzos, M. et al. Circulating interleukin-15 and RANTES chemokine in Parkinson's disease. \u003cem\u003eActa Neurol. Scand.\u003c/em\u003e \u003cb\u003e116\u003c/b\u003e, 374\u0026ndash;379. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1600-0404.2007.00894.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1600-0404.2007.00894.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGangemi, S. et al. Effect of levodopa on interleukin-15 and RANTES circulating levels in patients affected by Parkinson's disease. \u003cem\u003eMediat. Inflamm.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e, 251\u0026ndash;253. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/09629350310001599701\u003c/span\u003e\u003cspan address=\"10.1080/09629350310001599701\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2003).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePucha, K. A. et al. Neuron-derived extracellular vesicles to examine brain mTOR target engagement with sirolimus in patients with multiple system atrophy. \u003cem\u003eParkinsonism Relat. Disord.\u003c/em\u003e \u003cb\u003e115\u003c/b\u003e, 105821. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.parkreldis.2023.105821\u003c/span\u003e\u003cspan address=\"10.1016/j.parkreldis.2023.105821\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJeong, D. J. et al. The Mst1/2-BNIP3 axis is required for mitophagy induction and neuronal viability under mitochondrial stress. \u003cem\u003eExp. Mol. Med.\u003c/em\u003e \u003cb\u003e56\u003c/b\u003e, 674\u0026ndash;685. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s12276-024-01198-y\u003c/span\u003e\u003cspan address=\"10.1038/s12276-024-01198-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun, Y. et al. Methylprednisolone alleviates cognitive functions through the regulation of neuroinflammation in Alzheimer's disease. \u003cem\u003eFront. Immunol.\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e, 1192940. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fimmu.2023.1192940\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2023.1192940\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBakr, M. N., Takahashi, H. \u0026amp; Kikuchi, Y. CHRNA1 and its correlated-myogenesis/cell cycle genes are prognosis-related markers of metastatic melanoma. \u003cem\u003eBiochem. Biophys. Rep.\u003c/em\u003e \u003cb\u003e33\u003c/b\u003e, 101425. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.bbrep.2023.101425\u003c/span\u003e\u003cspan address=\"10.1016/j.bbrep.2023.101425\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTurk, S. et al. NK-cell dysfunction of acute myeloid leukemia in relation to the renin-angiotensin system and neurotransmitter genes. \u003cem\u003eOpen. Med. (Warsaw Poland)\u003c/em\u003e. \u003cb\u003e17\u003c/b\u003e, 1495\u0026ndash;1506. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1515/med-2022-0551\u003c/span\u003e\u003cspan address=\"10.1515/med-2022-0551\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHtet, M. et al. HEXIM1 is correlated with Alzheimer's disease pathology and regulates immediate early gene dynamics in neurons. \u003cem\u003ebioRxiv: preprint Serv. biology\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2024.09.27.615234\u003c/span\u003e\u003cspan address=\"10.1101/2024.09.27.615234\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang, Y. et al. HEXIM1 homodimer binds two sites on 7SK RNA to release autoinhibition for P-TEFb inactivation. \u003cem\u003ebioRxiv: preprint Serv. biology\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2024.10.10.617642\u003c/span\u003e\u003cspan address=\"10.1101/2024.10.10.617642\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrase, L., Yu, Y., McDade, E., Harari, O. \u0026amp; Benitez, B. A. Comparative gene regulatory networks modulating APOE expression in microglia and astrocytes. \u003cem\u003emedRxiv: preprint Serv. health Sci.\u003c/em\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2024.04.19.24306098\u003c/span\u003e\u003cspan address=\"10.1101/2024.04.19.24306098\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai, Z., Jia, X., Liu, M., Yang, X. \u0026amp; Cui, L. Epigenome-wide DNA methylation study of whole blood in patients with sporadic amyotrophic lateral sclerosis. \u003cem\u003eChin. Med. J.\u003c/em\u003e \u003cb\u003e135\u003c/b\u003e, 1466\u0026ndash;1473. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1097/cm9.0000000000002090\u003c/span\u003e\u003cspan address=\"10.1097/cm9.0000000000002090\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurrie, D. et al. A Potential Prognostic Gene Signature Associated with p53-Dependent NTRK1 Activation and Increased Survival of Neuroblastoma Patients. \u003cem\u003eCancers\u003c/em\u003e \u003cb\u003e16\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/cancers16040722\u003c/span\u003e\u003cspan address=\"10.3390/cancers16040722\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSuszyńska-Zajczyk, J., Luczak, M., Marczak, L. \u0026amp; Jakubowski, H. Inactivation of the paraoxonase 1 gene affects the expression of mouse brain proteins involved in neurodegeneration. \u003cem\u003eJ. Alzheimer's disease: JAD\u003c/em\u003e. \u003cb\u003e42\u003c/b\u003e, 247\u0026ndash;260. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3233/jad-132714\u003c/span\u003e\u003cspan address=\"10.3233/jad-132714\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNegrey, J. D. et al. Transcriptional profiles in olfactory pathway-associated brain regions of African green monkeys: Associations with age and Alzheimer's disease neuropathology. \u003cem\u003eAlzheimer's \u0026amp; dementia (New York, N. Y.)\u003c/em\u003e 8, e12358, (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/trc2.12358\u003c/span\u003e\u003cspan address=\"10.1002/trc2.12358\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShiina, N., Yamaguchi, K. \u0026amp; Tokunaga, M. RNG105 deficiency impairs the dendritic localization of mRNAs for Na+/K\u0026thinsp;+\u0026thinsp;ATPase subunit isoforms and leads to the degeneration of neuronal networks. \u003cem\u003eJ. neuroscience: official J. Soc. Neurosci.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, 12816\u0026ndash;12830. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1523/jneurosci.6386-09.2010\u003c/span\u003e\u003cspan address=\"10.1523/jneurosci.6386-09.2010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGeorge, A. J. et al. A serial analysis of gene expression profile of the Alzheimer's disease Tg2576 mouse model. \u003cem\u003eNeurotox. Res.\u003c/em\u003e \u003cb\u003e17\u003c/b\u003e, 360\u0026ndash;379. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s12640-009-9112-3\u003c/span\u003e\u003cspan address=\"10.1007/s12640-009-9112-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHill-Burns, E. M. et al. Identification of genetic modifiers of age-at-onset for familial Parkinson's disease. \u003cem\u003eHum. Mol. Genet.\u003c/em\u003e \u003cb\u003e25\u003c/b\u003e, 3849\u0026ndash;3862. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/hmg/ddw206\u003c/span\u003e\u003cspan address=\"10.1093/hmg/ddw206\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHasson, S. A. et al. High-content genome-wide RNAi screens identify regulators of parkin upstream of mitophagy. \u003cem\u003eNature\u003c/em\u003e \u003cb\u003e504\u003c/b\u003e, 291\u0026ndash;295. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nature12748\u003c/span\u003e\u003cspan address=\"10.1038/nature12748\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen, L. \u0026amp; Dettmer, U. Alpha-Synuclein Effects on Mitochondrial Quality Control in Parkinson's Disease. \u003cem\u003eBiomolecules\u003c/em\u003e \u003cb\u003e14\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3390/biom14121649\u003c/span\u003e\u003cspan address=\"10.3390/biom14121649\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, Z., Li, Q. \u0026amp; Huo, R. SUCLG1 promotes aerobic respiration and progression in plexiform neurofibroma. \u003cem\u003eInt. J. Oncol.\u003c/em\u003e \u003cb\u003e66\u003c/b\u003e \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3892/ijo.2024.5716\u003c/span\u003e\u003cspan address=\"10.3892/ijo.2024.5716\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCimadamore-Werthein, C. et al. Human mitochondrial carriers of the SLC25 family function as monomers exchanging substrates with a ping-pong kinetic mechanism. \u003cem\u003eEMBO J.\u003c/em\u003e \u003cb\u003e43\u003c/b\u003e, 3450\u0026ndash;3465. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s44318-024-00150-0\u003c/span\u003e\u003cspan address=\"10.1038/s44318-024-00150-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRigby, M. J. et al. Increased expression of SLC25A1/CIC causes an autistic-like phenotype with altered neuron morphology. \u003cem\u003eBrain: J. Neurol.\u003c/em\u003e \u003cb\u003e145\u003c/b\u003e, 500\u0026ndash;516. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/brain/awab295\u003c/span\u003e\u003cspan address=\"10.1093/brain/awab295\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGui, J. et al. Identification of Brain Cell Type-Specific Therapeutic Targets for Glioma From Genetics. \u003cem\u003eCNS Neurosci. Ther.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e, e70185. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/cns.70185\u003c/span\u003e\u003cspan address=\"10.1111/cns.70185\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e\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":"Neurodegenerative diseases, Mendelian Randomization, Machine Learning, Drug targets, Genetics","lastPublishedDoi":"10.21203/rs.3.rs-5848172/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5848172/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNeurodegenerative diseases (NDDs) remain a global health challenge. Alzheimer's disease (AD) and Parkinson's disease (PD) are the main types of NDDs worldwide, and Mendelian Randomization (MR) analysis across multi-omics and the entire genome offers novel strategies for identifying potential drug targets. This study used MR and summary-based MR(SMR) analysis to explore the causal relationship between genes and NDDs. Colocalization analysis and machine learning further validated and reinforced the MR findings. The pharmacological activity of candidate drug targets was confirmed via molecular docking and Molecular dynamics. This study revealed 14 genes that were closely associated with both NDDs. Specifically, IQCE(AD), HDHD2(AD), COMMD10(AD), ALPP (AD), FXYD6 (AD), STK3 (PD), LHFPL2 (PD), and ENPP4 (PD) were identified as risk factors for NDDs (OR\u0026thinsp;\u0026gt;\u0026thinsp;1), whereas HEXIM2 (AD), TSC22D4 (AD), CHRNB1 (PD), BAG4 (PD), SLC25A1 (PD), and IL15 (PD) were protective factors (OR\u0026thinsp;\u0026lt;\u0026thinsp;1). Molecular docking results revealed strong binding activities for PREDNISOLONE(ALPP = -7.6 kcal/mol), PANCURONIUM BROMIDE(CHRNB1 = -8 kcal/mol), CHEMBL379975(STK3 =-10.7 kcal/mol) and SIROLIMUS(IL15 = -9 kcal/mol). Molecular dynamics simulations confirmed the stable binding of the IL15-Sirolimus, ALPP-Prednisolone, STK3-CHEMBL379975, and CHRNB1-Rocuronium bromide complexes. This multi-omics study revealed 14 promising therapeutic targets for NDDs, providing new insights for targeted therapies and clinical strategies for NDDs. Our results provide evidence for future studies aimed at developing appropriate therapeutic interventions.\u003c/p\u003e","manuscriptTitle":"Multicomponent Mendelian randomization and machine learning studies of potential drug targets for neurodegenerative diseases","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-18 09:06:45","doi":"10.21203/rs.3.rs-5848172/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":"27691cd6-338f-46a3-a44b-daeeeb35ecb2","owner":[],"postedDate":"April 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":47332224,"name":"Health sciences/Neurology/Neurological disorders/Neurodegenerative diseases/Alzheimers disease"},{"id":47332225,"name":"Health sciences/Neurology/Neurological disorders/Neurodegenerative diseases/Parkinsons disease"},{"id":47332226,"name":"Biological sciences/Drug discovery/Target identification"},{"id":47332227,"name":"Health sciences/Medical research"},{"id":47332228,"name":"Health sciences/Risk factors"}],"tags":[],"updatedAt":"2025-08-06T09:09:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-18 09:06:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5848172","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5848172","identity":"rs-5848172","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}