{"paper_id":"ac333005-8774-4035-913e-a601a03bfa5d","body_text":"Kai et al. BMC Women’s Health          (2025) 25:161  \nhttps://doi.org/10.1186/s12905-025-03697-0\nRESEARCH Open Access\n© The Author(s) 2025. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 \nInternational License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long \nas you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if \nyou modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or \nparts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated \notherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not \npermitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To \nview a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.\nBMC Women’s Health\nIdentifying key palmitoylation-associated \ngenes in endometriosis through genomic data \nanalysis\nJinyan Kai1, Jiaqi Su1, Yinping You2, Xiaoliang Liang1, Haitao Huang3, Jie Fang4* and Qiong Chen5* \nAbstract \nBackground Palmitoylation, a post-translational lipid modification, has garnered increasing attention for its role \nin inflammatory processes and tumorigenesis. Emerging evidence suggests a potential association between pal-\nmitoylation and inflammatory responses in the pathogenesis of endometriosis. However, the precise mechanistic \ninterplay remains elusive, necessitating further investigation.\nMethods This study integrated transcriptomic analysis and Mendelian randomization (MR) to identify a causal gene \nset implicated in endometriosis. Differentially expressed genes (DEGs) were first identified in the training dataset using \nthe limma package in R. Weighted gene co-expression network analysis (WGCNA) was subsequently performed, lev-\neraging Single Sample Gene Set Enrichment Analysis (ssGSEA)-derived scores of palmitoylation-related genes (PRGs) \nas phenotypic traits to identify key modular genes. The intersection of these key modular genes with DEGs yielded \na refined gene set. Machine learning algorithms were then applied to further optimize gene selection, followed \nby external validation, immune infiltration analysis, RNA network construction, and exploration of potential targeted \ndrug candidates.\nResults Through a rigorous screening process, VRK1, GALNT12, and RMI1 emerged as key genes associated with pal-\nmitoylation, exhibiting significant downregulation in endometriosis samples (P < 0.05), indicative of a potential \nprotective role. Immune infiltration analysis further revealed strong correlations between these genes and M2 \nmacrophages as well as resting Natural Killer (NK) cells. Additionally, investigations into the targeted RNA network \nand drug association profiling provided novel insights, laying the groundwork for future high-quality validation \nstudies.\nConclusions This study employed a comprehensive analytical framework to identify palmitoylation-associated key \ngenes in endometriosis. The integration of immunoinfiltration analysis, RNA network construction, and drug associa-\ntion profiling offers valuable insights for advancing clinical diagnostics, disease monitoring, and therapeutic develop-\nment in endometriosis.\nKeywords Palmitoylation, Endometriosis, Transcriptomics, Mendelian Randomization\n*Correspondence:\nJie Fang\nfang_jie0831@163.com\nQiong Chen\nchenqiongjone@shsmu.edu.cn\nFull list of author information is available at the end of the article\n\nPage 2 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nIntroduction\nEndometriosis is an estrogen-dependent chronic inflam -\nmatory condition characterized by the ectopic pro -\nliferation of endometrial-like tissue [1, 2]. Despite its \nhistologically benign classification, the condition exhibits \ninvasive, metastatic, and recurrent behaviors reminiscent \nof malignancies, affecting approximately 10% of women \nof reproductive age [3, 4]. Common clinical manifesta -\ntions, including chronic pelvic pain, dysmenorrhea, dysu-\nria, and infertility, substantially diminish patients’ quality \nof life [4]. As the pathophysiology of the disease becomes \nincreasingly understood, attention has shifted toward the \nroles of hormonal dysregulation, inflammatory media -\ntors, and genetic susceptibility [5, 6]. However, defini -\ntive conclusions regarding their precise contributions \nremain elusive. Current therapeutic strategies primar -\nily rely on hormonal suppression therapy and minimally \ninvasive surgical interventions [7]. Nevertheless, these \napproaches are often associated with high recurrence \nrates and treatment-related complications [8, 9]. Given \nthese limitations, elucidating the underlying molecular \nmechanisms and identifying reliable diagnostic and prog-\nnostic biomarkers represent crucial avenues for advanc -\ning precision medicine, optimizing clinical management, \nand mitigating disease recurrence.\nProtein palmitoylation, a reversible post-translational \nlipid modification, is dynamically regulated by a cohort \nof palmitoyl S-acyltransferases characterized by the \nAsp-His-His-Cys (DHHC) motif [10, 11]. This process \nis counterbalanced by acylprotein thioesterases, which \nmodulate protein localization and function in a highly \ndynamic manner [10, 11]. Recent studies have highlighted \nthe intricate role of palmitoylation in inflammatory regu -\nlation, with ZDHHC12 promoting the degradation of \nNOD-like receptor family pyrin domain-containing \n3 (NLRP3) through chaperone-mediated autophagy \n[12]. In autoinflammatory disorders, the NOD2 variant \nNOD2 s-R444 C demonstrates an increased affinity for \nZDHHC5, leading to excessive palmitoylation and exac -\nerbated inflammatory responses [13]. Inflammation is \nrecognized as a central etiological factor in endometrio -\nsis. Recent findings suggest that NLRP3-mediated pyrop -\ntosis contributes to the pathogenesis of inflammatory \nendometriosis by driving ectopic endometrial cell prolif -\neration and angiogenesis [14]. Targeted anti-inflamma -\ntory interventions, such as long-acting anti-IL8 antibody \nadministration, have shown promise in mitigating disease \nprogression [15]. Despite extensive research on palmi -\ntoylation across various biological processes [16–18], its \nspecific role in endometriosis remains largely unexplored. \nNotably, ZDHHC12 has been implicated in modulating \nNLRP3 palmitoylation, thereby influencing its activation \nstatus and regulating myocardial inflammation, oxidative \nstress, and associated cellular damage [19]. Furthermore, \nloss of palmitoyl protein thioesterase 1 (Ppt1) impairs \ndepalmitoylation, leading to aberrant synaptic protein \ntrafficking and neuroinflammation through mechanisms \ninvolving A-kinase anchor protein 5 (Akap5) and nuclear \nfactor of activated T cells (NFAT) [20]. Building on these \ninsights, this study aims to elucidate the regulatory role \nof palmitoylation in the pathogenesis of endometriosis, \nassessing its potential as a critical modulatory mecha -\nnism. By uncovering previously unrecognized pathogenic \npathways and identifying novel therapeutic targets, these \nfindings are anticipated to advance the development of \nmore effective treatment strategies.\nBy integrating transcriptomic and genomic data from \nthe Gene Expression Omnibus (GEO) and Genome-Wide \nAssociation Studies (GWAS) databases, key palmitoyla -\ntion-associated genes in endometriosis were identified \nthrough differential expression analysis, Mendelian ran -\ndomization (MR), machine learning, and expression vali -\ndation. Further exploration of the interplay between \npalmitoylation and endometriosis was conducted via \nimmune infiltration analysis, chromosomal localization, \nand regulatory network reconstruction, providing a theo-\nretical foundation for the precise diagnosis, surveillance, \nand therapeutic intervention of endometriosis.\nMaterials and methods\nData collection and extraction\nEndometriosis-related datasets (GSE51981 and \nGSE25628) were obtained from the GEO database for \ntranscriptomic analysis. The GSE51981 dataset, desig -\nnated as the training set, comprised 77 pelvic endometri -\nosis (PE) samples and 34 control samples, with genomic \nprofiling conducted on the GPL570 platform. To enhance \ndata specificity, 37 samples with uterine or pelvic pathol -\nogy were excluded from the analysis. The GSE25628 data-\nset, serving as the independent validation set, included \n16 endometriosis and 6 control endometrial tissue sam -\nples, sequenced on the GPL571 platform. Additionally, \nMendelian randomization (MR) data on endometriosis \nwere retrieved from the publicly available Integrative \nEpidemiology Unit (IEU) Open GWAS database. The \nselected dataset (ukb-b- 9668) comprised genomic data \nfrom 463,010 European individuals, including 1,121 cases \nand 461,889 controls, encompassing a total of 9,851,867 \nsingle nucleotide polymorphisms (SNPs). A curated list \nof 23 palmitoylation-related genes (PRGs) was extracted \nfrom relevant literature [21].\nPRGs score and weighted gene co‑expression network \nanalysis\nIn the GSE51981 dataset, PRG scores were computed \nusing single-sample gene set enrichment analysis \n\nPage 3 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \n(ssGSEA) from the GSVA package (v1.46.0; data of use: \n2024.11.20) [22], based on the differential expression of \nPRGs in PE and control samples. Statistical comparisons \nof PRG scores between PE and control groups were per -\nformed using the Wilcoxon test, with significance set at \nP < 0.05.\nWeighted gene co-expression network analysis \n(WGCNA) was subsequently applied to identify key \nmodule genes in GSE51981, utilizing ssGSEA-derived \nPRG scores as trait variables via the WGCNA package \n(v1.7.1; data of use: 2024.11.20) [23]. Initial sample clus -\ntering was conducted to detect and eliminate outliers. \nThe optimal soft threshold power was determined by \nachieving an  R2 exceeding 0.8 while maintaining near-\nzero mean connectivity. A co-expression matrix was \nthen constructed using the selected soft threshold, with \na minimum module size of 30 genes, a dynamic tree cut \nparameter of 2, and a module merging threshold of 0.25. \nDistinct gene modules were assigned unique color labels. \nCorrelation coefficients between endometriosis samples, \ncontrol samples, and PRG scores were computed for each \nmodule, and the associations were visualized in a heat -\nmap. Modules demonstrating a significant correlation \nwith PRG scores (|r|> 0.5, P < 0.05) were designated as \nkey modules, with their constituent genes identified as \nkey module genes.\nDifferential expression analysis\nDifferentially expressed genes (DEGs) between PE and \ncontrol samples in GSE51981 were identified using the \nlimma package (v3.54.0; data of use: 2024.11.20) [24], \napplying selection criteria of |log 2 fold change (FC)|> 1 \nand P < 0.05. The distribution of DEGs was illustrated via \na volcano plot and heatmap, generated using the ggplot2 \npackage (v3.4.3; data of use: 2024.11.20) [25] and Com -\nplexHeatmap package (v2.14.0; data of use: 2024.11.20) \n[26], respectively.\nFunction analysis\nThe intersecting genes were identified by overlapping \nDEGs with key module genes. To elucidate their func -\ntional significance, Gene Ontology (GO) and Kyoto \nEncyclopedia of Genes and Genomes (KEGG) pathway \nenrichment analyses were performed using the cluster -\nProfiler package (v4.7.1.003; data of use: 2024.11.21) [27], \nwith a significance threshold of P < 0.05. GO enrichment \nanalysis categorized functional annotations into biologi -\ncal processes (BP), cellular components (CC), and molec-\nular functions (MF).\nMR study\nBased on these intersecting genes, an MR analysis was \nconducted using the TwoSampleMR package (v0.5.6; data \nof use: 2024.11.21) [28], treating these genes as expo -\nsure factors and endometriosis as the outcome variable. \nStringent adherence to classical MR assumptions was \nmaintained throughout the analysis: (i) the independence \nassumption ensured that instrumental variables (IVs) \nwere not confounded by external factors, (ii) the associa -\ntion assumption confirmed a direct influence of IVs on \nthe exposure, and (iii) the exclusivity assumption verified \nthat IVs affected the outcome solely through the expo -\nsure, without alternative causal pathways.\nGWAS data for the intersecting genes (expression \nQuantitative Trait Loci, eQTL) and endometriosis (ukb-\nb- 9668) were retrieved from the IEU Open GWAS \ndatabase. Initial IV screening was performed using the \nVariantAnnotation (v1.44.0; data of use: 2024.11.21) [29] \nand ieugwasr (v1.0.1; data of use: 2024.11.21) [30] pack -\nages, with a significance threshold of P < 5 ×  10–6 . Linkage \ndisequilibrium (LD) filtering was applied (clump = TRUE, \n R2 = 0.001, kb = 10), and genes with at least three SNPs \n(nSNP ≥ 3) were retained, ensuring harmonization of \neffect alleles and effect sizes. Weak IVs were identified \nbased on the F-statistic, with IVs excluded when F < 10. \nMR analysis was conducted using five complementary \nmethods: MR Egger [31], Weighted Median [32], Inverse \nVariance Weighted (IVW) [33], Simple Mode [34], and \nWeighted Mode [35], with IVW serving as the primary \nstatistical approach (P < 0.05). Results were visualized \nthrough scatter plots, forest plots, and funnel plots. \nTo assess the robustness of the MR findings, sensitivity \nanalyses were performed, including heterogeneity test -\ning (Cochran’s Q test, P > 0.05), horizontal pleiotropy \nevaluation (P > 0.05), and leave-one-out (LOO) analysis \nusing the mr heterogeneity [36], mr pleiotropy test [37], \nand mr leaveoneout [38] functions, respectively. The \ncausal direction was further validated using the Steiger \ntest, with a correct causal direction indicated by Steiger \nP < 0.05. Following these analyses, genes demonstrating \nsignificant causal relationships with endometriosis were \nidentified as candidate genes for further investigation.\nMachine learning and gene expression analysis\nTo further refine the selection of feature genes, four \nmachine learning algorithms—Random Forest (RF), \nSupport Vector Machine (SVM), Generalized Linear \nModel (GLM), and k-Nearest Neighbor (KNN)—were \nemployed to construct predictive models based on the \nexpression profiles of candidate genes in the GSE51981 \ndataset. The core reason for selecting these machine \nlearning methods was that their advantages comple -\nmented each other, providing more comprehensive and \naccurate feature gene selection results. Each method \nhad its strengths in data processing, feature impor -\ntance evaluation, high-dimensional data handling, and \n\nPage 4 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nmodel interpretability. Therefore, using multiple meth -\nods for comparison improved the reliability and accu -\nracy of the results. Model training and validation were \nconducted using the caret package (v6.0–93; data of \nuse: 2024.11.21) [39]. To evaluate model performance, \nreceiver operating characteristic (ROC) curves and \nresidual box plots were generated using the DALEX \npackage (v1.1.0; data of use: 2024.11.21) [40]. Addition -\nally, gene importance scores derived from each machine \nlearning model were visualized using bar plots. The top \n10 most important genes from each model were inter -\nsected to identify a refined set of feature genes.\nExpression validation of the identified feature genes \nwas subsequently performed in both the GSE51981 \nand GSE25628 datasets. Wilcoxon tests were applied to \ncompare gene expression between endometriosis and \ncontrol samples, with significance set at P < 0.05. Genes \nexhibiting significant differential expression in both \ndatasets, with a consistent expression trend, were des -\nignated as key genes.\nImmune infiltration analysis\nTo assess immune cell infiltration in endometriosis and \ncontrol samples from GSE51981, the CIBERSORT algo -\nrithm (v1.0.3; data of use: 2024.11.22) [41] was applied \nto estimate immune scores for 22 immune cell types. \nSamples with P  > 0.05 were excluded to ensure reliable \ndeconvolution results. Wilcoxon tests were then used \nto compare immune cell composition between endo -\nmetriosis and control samples, and immune cell types \nexhibiting significant differential infiltration (P  < 0.05) \nwere selected for further analysis.\nSpearman correlation analysis was subsequently con -\nducted to explore relationships among the 22 immune \ncell types and to assess associations between key genes \nand differentially infiltrated immune cells, with correla -\ntion thresholds set at |r|> 0.3 and P  < 0.05.\nChromosomal localization and functional similarity \nanalyses\nTo determine the genomic distribution of key genes \nacross the 23 pairs of human chromosomes, the Univer -\nsity of California Santa Cruz (UCSC) Genome Browser \n(http:// genome. ucsc. edu/) was utilized to retrieve their \nchromosomal start and stop positions. The RCircos \npackage (v1.2.2; data of use: 2024.11.22) [42] was then \nemployed to generate a genome-wide visualization of key \ngene loci. Additionally, functional relationships among \nthe key genes were further explored using the GoSem -\nSim package (v6.5–0; data of use: 2024.11.22) [43].\nRegulation network analysis\nThe miRwalk (http:// mirwa lk. umm. uni- heide lberg. de/) \nand miRDB (https:// mirdb. org/) databases were utilized \nto predict MicroRNAs (miRNAs) targeting the identi -\nfied key genes. The intersection of miRNAs derived from \nboth databases was considered the final set of key miR -\nNAs. An mRNA-miRNA regulatory network was sub -\nsequently constructed and visualized using Cytoscape \n(v3.10.2; data of use: 2024.11.22) [44].\nSimilarly, transcription factors (TFs) associated with \nkey genes were predicted using the hTFtarget (https://  \nguolab. wchscu. cn/ hTFta rget/# !/) and miRNet (https:// \nwww. mirnet. ca/) databases. Key TFs were identified \nby overlapping the predictions from both sources, and \nan mRNA-TF regulatory network was constructed and \nvisualized in Cytoscape.\nTo further explore potential therapeutic targets, \ndrug-gene interactions were analyzed using the Com -\nparative Toxicogenomics Database (CTD) (http:// ctdba  \nse. org/) and the Enrichr database (https:// maaya nlab.  \ncloud/ Enric hr/). Drugs targeting endometriosis-asso -\nciated key genes were extracted from both databases, \nwith duplicate entries removed. An mRNA-drug inter -\naction network was then established and visualized in \nCytoscape.\nImmunohistochemistry\nFor experimental validation, three paraffin-embedded \nsections of ectopic endometrial and normal endometrial \ntissue were collected from the Pathology Department \nof Shanghai General Hospital. Immunohistochemistry \n(IHC) was performed using primary antibodies against \nGALNT12 (Solarbio, K108365P , 1:100), VRK1 (Pro -\nteintech, 28,018–1-AP , 1:100), and RMI1 (Proteintech, \n14630–1-AP , 1:100), diluted in Phosphate-Buffered \nSaline (PBS). Five-micrometer-thick paraffin sections \nwere deparaffinized and rehydrated, followed by incuba -\ntion with 0.3%  H2O2 in methanol to inhibit endogenous \nperoxidase activity. After antigen retrieval and cooling, \nsections were blocked with 1% Bovine Serum Albumin \n(BSA) and incubated with primary antibodies overnight \nat 4  °C. The following day, sections were treated with \nHRP-conjugated secondary antibodies (Shanghai Long \nIsland Biotech, Shanghai, China) for 1  h at room tem -\nperature, followed by diaminobenzidine (DAB) staining \nand hematoxylin counterstaining. Slides were examined \nand imaged under a Leica SP5 light microscope (Leica, \nChina) at 100 × and 200 × magnification.\nStatistical analysis\nStatistical analyses were conducted using R (v4.2.2), \nwith inter-group differences assessed via the Wilcoxon \n\nPage 5 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \ntest (P < 0.05). Regulatory networks were generated and \nvisualized using Cytoscape (v3.10.2).\nResults\nScreening of palmitoacylation related gene modules \nin endometriosis\nPRG score analysis in the GSE51981 dataset revealed \nsignificantly elevated scores in PE samples compared to \ncontrols (Fig. 1A). To further explore gene co-expression \npatterns, WGCNA was performed on the GSE51981 \ndataset. Clustering analysis confirmed the absence of \noutlier samples (Fig.  1B). An optimal soft-thresholding \npower of 19 was determined based on scale-free topol -\nogy criteria  (R2 = 0.8) while maintaining mean connec -\ntivity near zero (Fig.  1C). Subsequently, a co-expression \nmatrix was constructed, identifying 18 distinct modules, \neach represented by a unique color, with the Grey mod -\nule excluded as it contained unassigned genes (Fig.  1D). \nPearson correlation analysis revealed significant associa -\ntions between PRG scores and two key modules: MEgree-\nnyellow (r = 0.68, P < 0.001) and MEbrown (r = − 0.55, P < \n0.001) (Fig.  1E). These modules were designated as key \nmodules, collectively encompassing 307 genes, referred \nto as key module genes.\nIdentification and functional exploration \nof the intersection genes\nDifferential expression analysis in the GSE51981 dataset \nidentified 3,376 DEGs between endometriosis and con -\ntrol samples, with 1,267 genes exhibiting upregulation \nand 2,109 showing downregulation (Fig.  2A and B). By \nintersecting the 3,376 DEGs with the 307 key module \ngenes, 204 intersection genes were identified (Fig.  2C). \nFunctional enrichment analysis of these 204 genes \nrevealed significant enrichment in 368 GO terms and \n31 KEGG pathways. GO enrichment analysis, catego -\nrized into BP , CC, and MF, identified key terms such \nas\"nuclear division,\"\"chromosomal region,\"and\"tubulin \nbinding\"(Fig.  2D). KEGG pathway enrichment analy -\nsis highlighted pathways including\"cell cycle,\"\"mineral \nabsorption,\"and\"progesterone-mediated oocyte \nmaturation\"(Fig. 2E).\nCandidate genes with a significant causal relationship \nwith endometriosis\nThe causal association between the 204 intersect -\ning genes and endometriosis was further examined. \nFollowing the IV screening, 126 genes remained as \nexposure factors for further investigation. The MR \nanalysis identified 17 genes with a statistically signifi -\ncant causal relationship with endometriosis (P  < 0.05) \n(Table 1). Among them, seven genes (e.g., CFD, ECT2, \nHMMR) were classified as risk factors (Odds Ratio \n[OR] > 1), whereas ten genes (e.g., GALNT12, RMI1, \nVRK1) exhibited a protective effect (OR < 1). To visu -\nalize these associations, scatter plots, forest plots, and \nfunnel plots were generated. Specifically, scatter plots \nfor GALNT12, RMI1, and VRK1 (Fig.  3A) displayed a \nnegative slope in their fitted regression lines, consist -\nent with a protective association. Forest plots (Fig.  3B) \nFig. 1 Results of Screening palmitoacylation related gene modules. A The PRGs score of PE and control samples. B The result of cluster analysis. \nC The scale-free fit index for soft threshold power and mean connectivity. D Gene and trait clustering dendrograms. Each branch represents \nan expression module of a highly interconnected groups of genes; each color indicates a corresponding co-expression module. E Heatmap of 18 \ngene co-expression modules. The numbers in each cell means the correlation coefficient and p value\n\nPage 6 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nfurther illustrated the MR effect sizes, all of which were \nnegative under the IVW method, reinforcing their \nprotective role. Funnel plots (Fig.  3C) demonstrated a \nsymmetrical distribution of IVs around the IVW line, \nindicating adherence to Mendel’s second law. Scatter \nplots, forest plots, and funnel plots for the remaining \nFig. 2 Identification and functional exploration of the intersection genes. A Volcano plot. We set the criteria of |log2fold-change (FC)|> 1 and P < \n0.05 as the difference genes. Red dots are upregulated genes, and blue dots are downregulated genes. B Heatmap plot. The heatmap reflects \nthe distribution of gene expression density and gene expression differences in each sample. C Venn diagram. The key module genes obtained \nfrom WGCNA were intersected with DEGS genes. D GO enrichment analysis results. E KEGG enrichment analysis results\nTable 1 Mendelian randomization analysis unveils 17 causal genes in endometriosis\nAbbreviation: PE pelvic endometriosis\nNO exposure outcome method nsnp pval or\n1 CENPE PE Inverse variance weighted 25 0.000458026 0.99892761\n2 CFD PE Inverse variance weighted 7 0.013505093 1.000750339\n3 ECT2 PE Inverse variance weighted 11 0.049775503 1.000690854\n4 FBXO5 PE Inverse variance weighted 8 0.01256748 0.999008793\n5 GALNT12 PE Inverse variance weighted 9 0.011019426 0.998513982\n6 HMMR PE Inverse variance weighted 5 0.008649861 1.002027803\n7 IER3 PE Inverse variance weighted 17 0.012138092 0.999591358\n8 MKI67 PE Inverse variance weighted 3 0.020648678 0.997693863\n9 NDC80 PE Inverse variance weighted 15 0.014084781 0.999332329\n10 PARPBP PE Inverse variance weighted 4 0.003391087 0.997696177\n11 PRIM1 PE Inverse variance weighted 9 0.008881558 1.000353618\n12 RLN2 PE Inverse variance weighted 4 0.045332653 0.998816069\n13 RMI1 PE Inverse variance weighted 11 0.048692518 0.999587813\n14 STIL PE Inverse variance weighted 16 0.009312601 1.000757596\n15 STMN1 PE Inverse variance weighted 19 0.00385043 1.000811607\n16 TYMS PE Inverse variance weighted 12 0.002906875 1.00065104\n17 VRK1 PE Inverse variance weighted 10 0.002029756 0.999426071\n\nPage 7 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \ngenes are provided in Figures S1–S3. Additionally, het -\nerogeneity and horizontal pleiotropy tests across all \n17 genes yielded P  values exceeding 0.05 (Tables  2 and \n3), suggesting the absence of significant heterogene -\nity or confounding influences in the MR study. LOO \nanalysis (Fig.  3D and Fig. S4) further corroborated the \nrobustness of the MR results, as no substantial devia -\ntions were observed upon sequential exclusion of indi -\nvidual IVs. Finally, Steiger directionality tests (Table  4) \nconfirmed the correct causal direction for all 17 genes, \nwith P values below 0.05, reinforcing the validity of the \nfindings. Collectively, these 17 genes emerge as poten -\ntial causal candidates implicated in endometriosis.\nVRK1, GALNT12, and RMI1 were deemed as key genes \nfor endometriosis\nBuilding on the 17 candidate genes identified through the \nMR study, machine learning algorithms were employed \nto further refine the selection of feature genes. Four \ndistinct models were constructed, with their predic -\ntive performance assessed via ROC curves. All models \nFig. 3 Identification of candidate genes through MR study. A Scatter plots for GALNT12, RMI1, and VRK1. B Forest plots for GALNT12, RMI1, \nand VRK1. C Funnel plots for GALNT12, RMI1, and VRK1. D LOO analysis for GALNT12, RMI1, and VRK1\nTable 2 Results of Mendelian randomization heterogeneity test\nAbbreviation: PE pelvic endometriosis\nNO exposure outcome heterogeneity_pval\n1 CENPE PE 0.994236723\n2 CFD PE 0.904994304\n3 ECT2 PE 0.974875149\n4 FBXO5 PE 0.856902477\n5 GALNT12 PE 0.999813501\n6 HMMR PE 0.987296141\n7 IER3 PE 0.414274668\n8 MKI67 PE 0.935968918\n9 NDC80 PE 0.999792808\n10 PARPBP PE 0.988250764\n11 PRIM1 PE 0.946071659\n12 RLN2 PE 0.925913532\n13 RMI1 PE 0.659333301\n14 STIL PE 0.999978701\n15 STMN1 PE 0.99963283\n16 TYMS PE 0.999579046\n17 VRK1 PE 0.99775246\nTable 3 Results of Mendelian randomization level pleiotropy \ntest\nAbbreviation: PE pelvic endometriosis\nNO exposure outcome pleiotropy_pval\n1 CENPE PE 0.159937324\n2 CFD PE 0.353849828\n3 ECT2 PE 0.170419632\n4 FBXO5 PE 0.612553883\n5 GALNT12 PE 0.831829309\n6 HMMR PE 0.873546239\n7 IER3 PE 0.053648145\n8 MKI67 PE 0.791493743\n9 NDC80 PE 0.975018594\n10 PARPBP PE 0.773482392\n11 PRIM1 PE 0.362264287\n12 RLN2 PE 0.808950553\n13 RMI1 PE 0.070378789\n14 STIL PE 0.220005997\n15 STMN1 PE 0.781020923\n16 TYMS PE 0.894010125\n17 VRK1 PE 0.947229403\n\nPage 8 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nachieved an area under the curve (AUC) exceeding 0.7, \nindicative of high classification accuracy (Fig.  4A). Addi-\ntionally, residual box plots compared true observed val -\nues with model-predicted outcomes, further validating \nmodel reliability (Fig.  4B). To prioritize genes with the \ngreatest potential relevance to endometriosis treatment, \ngene importance scores were derived from each model \n(Fig. 4C). By selecting the top 10 genes from each model \nand determining their intersection, six feature genes were \nidentified: TYMS, VRK1, MK167, GALNT12, CFD, and \nRMI1 (Fig. 4D).\nSubsequent gene expression analysis in the GSE51981 \nand GSE25628 datasets revealed significantly lower \nexpression levels of VRK1, GALNT12, and RMI1 in both \ndatasets ( P < 0.05) (Fig.  4E and F). Consequently, these \nthree genes were designated as key genes implicated in \nendometriosis.\nImmune cell infiltration analysis\nImmune infiltration analysis (Fig.  5A) characterized the \ndistribution of 22 immune cell types in endometriosis \nand control samples from GSE51981. The Wilcoxon test \nidentified 11 differentially abundant immune cells. Nota -\nbly, M2 macrophages and resting mast cells exhibited sig-\nnificantly higher proportions in control samples, whereas \nmonocytes and resting natural killer (NK) cells were sig -\nnificantly enriched in endometriosis samples (Fig.  5B). \nCorrelation analysis among immune cell populations \ndemonstrated a strong positive association between rest -\ning mast cells and M0 macrophages (r = 0.51, P < 0.05), \nwhile regulatory T cells (Tregs) displayed the strong -\nest negative correlation with activated memory CD4 T \ncells (r = − 0.53, P < 0.05) (Fig.  5C). Further correlation \nTable 4 Mendelian randomization Steiger directivity analysis\nAbbreviation: PE pelvic endometriosis\nNO exposure outcome correct_causal_\ndirection\nsteiger_pval\n1 CENPE PE TRUE 1.8803E- 163\n2 CFD PE TRUE  < 0.001\n3 ECT2 PE TRUE 6.3846E- 150\n4 FBXO5 PE TRUE 8.09018E- 92\n5 GALNT12 PE TRUE 2.53764E- 52\n6 HMMR PE TRUE 2.45247E- 33\n7 IER3 PE TRUE  < 0.001\n8 MKI67 PE TRUE 7.05518E- 18\n9 NDC80 PE TRUE 8.5987E- 214\n10 PARPBP PE TRUE 4.02091E- 22\n11 PRIM1 PE TRUE  < 0.001\n12 RLN2 PE TRUE 3.27508E- 44\n13 RMI1 PE TRUE  < 0.001\n14 STIL PE TRUE 2.468E- 171\n15 STMN1 PE TRUE 1.9595E- 242\n16 TYMS PE TRUE  < 0.001\n17 VRK1 PE TRUE  < 0.001\nFig. 4 Obtained key genes via machine learning and external validation. A ROC curves constructed by four machine learning models. B Residual \nbox diagram of four machine learning models. C Feature importance of four machine learning models. D Venn diagram of the top 10 feature \nimportance genes across four machine learning models. E Expression of feature genes in GSE51981. F Expression of feature genes in GSE25628\n\nPage 9 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \nanalysis between key genes and differentially abundant \nimmune cells revealed a consistent positive association \nbetween all key genes and M2 macrophages, alongside a \nstrong negative correlation with resting NK cells (|r|> 0.3, \nP < 0.001) (Fig. 5D).\nChromosome localization and functional similarity analysis \nof key genes\nChromosomal localization analysis provided further \ninsights into the genomic context of the key genes. Spe -\ncifically, GALNT12 and RMI1 were mapped to chro -\nmosome 9, whereas VRK1 was located on chromosome \n14 (Fig.  6A). Functional similarity analysis revealed that \nVRK1 exhibited the highest similarity with the other key \ngenes, suggesting its potential central role in the patho -\ngenesis of endometriosis (Fig. 6B).\nAnalysis of regulatory networks associated with key genes\nPrediction of miRNA interactions with key genes iden -\ntified nine key miRNAs through overlapping results \nfrom the miRWalk and miRDB databases, enabling the \nconstruction of an mRNA-miRNA regulatory network \ncomprising 12 nodes and 9 edges. Notable interactions \nincluded VRK1- ‘hsa-mir- 4428’ , GALNT12- ‘hsa-mir- \n202 - 3p’ , and RMI1- ‘hsa-mir- 3190 - 3p’ (Fig. 7A). Fur-\nthermore, 61 TFs targeting the three key genes were \nidentified through overlapping predictions from the hTF-\ntarget and miRNet databases. These interactions were \nvisualized in an mRNA-TF network consisting of 64 \nnodes (3 key genes and 61 TFs) and 76 edges, with SPI1 \nidentified as a common regulator of all three key genes \n(Fig.  7B). Additionally, drug-gene interaction analysis \nidentified 195 drugs targeting the three key genes, leading \nto the construction of a key gene-drug network (Fig.  7C). \nNotably, enterolactone was found to co-target RMI1 \nand VRK1, while retinoic acid co-targeted GALNT12 \nand VRK1. These regulatory networks provide valu -\nable insights into the molecular mechanisms underlying \nendometriosis and potential therapeutic targets.\nValidation of key genes by immunohistochemistry\nTo validate the expression patterns of the key genes, \nthree cases of ectopic endometrial tissues and three cases \nof normal endometrial tissues were collected from the \npathology department. Immunohistochemical staining \nwas performed using antibodies against VRK1, RMI1, \nFig. 5 Immune cell infiltration analysis. A Proportions of 22 immune cell types in PE and controls. B Expression differences of 22 immune cell types \nin PE and controls. C Relationships among immune cells. D Associations between immune cells and key genes\n\nPage 10 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nand GALNT12. The results demonstrated significantly \nhigher positive staining rates for all three proteins in nor-\nmal endometrial tissues compared to endometriotic tis -\nsues, further corroborating their potential involvement in \nendometriosis pathophysiology (Fig. 8).\nDiscussion\nEndometriosis is an inflammatory disease character -\nized by invasiveness and recurrence, and currently \nlacks reliable diagnostic and monitoring indicators \n[2, 4]. Palmitoylation stands as a pivotal mechanism \nof protein post-translational modification, exerting a \nsignificant influence on inflammatory responses, lipid \nmetabolism, and the genesis of tumors [12, 45]. \nResearch indicates that palmitoylation plays a sig -\nnificant role in the migration and adhesion of neutro -\nphils by regulating the function of CRACR2 A protein, \nthereby affecting inflammatory responses and associ -\nated tissue damage [46]. Additionally, palmitoylation \nplays an important role in inflammatory responses by \nmodulating the functions of immune proteins and the \nmetabolism of gut microbiota [47]. Although the spe -\ncific role of palmitoylation in endometriosis remains \nunclear, its close association with inflammation sug -\ngests that it may play a key role in the inflammatory \nprocess of this disease. This study employed bioinfor -\nmatics approaches to identify DEGs associated with \npalmitoylation in endometriosis and further elucidated \ntheir functional relevance. Using the IEU OpenGWAS \ndatabase, 17 genes were identified with statistically sig -\nnificant associations, establishing a causal relationship \nbetween these genes and endometriosis. Subsequently, \nmachine learning algorithms, combined with external \ndataset validation, refined this selection to three key \ngenes—VRK1, GALNT12, and RMI1—each exhibit -\ning reduced expression in endometriotic tissues and \ndemonstrating a negative correlation with disease \noccurrence.\nThe VRK1 (vaccinia-related kinase 1) gene, which \nencodes a serine/threonine protein kinase, is localized \non chromosome 14 and exhibits broad expression across \nhuman tissues, with predominant nuclear localization \n[48]. The VRK1-encoded protein regulates cell cycle pro -\ngression and genomic stability through phosphorylation \nand is implicated in apoptosis, thus contributing to cel -\nlular proliferation and tissue regeneration [49]. Previous \nstudies have demonstrated that VRK1 modulates p53 \nstability and activity via phosphorylation, thereby influ -\nencing lung cancer cell proliferation [50]. Additionally, \nVRK1 promotes cell cycle progression by phosphorylat -\ning VREB, thereby enhancing cAMP-responsive element-\nbinding protein activity at the CCND1 promoter, leading \nto CCND1 upregulation [51]. Furthermore, VRK1 plays a \npivotal role in DNA damage repair by stabilizing histone \nH2 AX-H3 interactions, neutralizing ionizing radiation-\ninduced H2 AX phosphorylation, and participating in \nearly DNA repair mechanisms [52].\nThe GALNT12 (N-Acetylgalactosaminyltransferase \n12) gene, located on chromosome 9, belongs to the poly -\npeptide N-acetylgalactosaminyltransferase family and is \nprimarily involved in protein post-translational modi -\nfication. It catalyzes the transfer of N-acetylgalactosa -\nmine to serine or threonine residues of target proteins, \nthereby influencing protein conformation, functional \nFig. 6 Chromosome localization and functional similarity analysis of key genes. A The chromosome localization of key genes. B The functional \nsimilarity analysis of key genes\n\nPage 11 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \nproperties, and genomic stability [53]. Aberrant expres -\nsion or dysregulation of GALNT12 has been implicated \nin various pathological conditions. For instance, muta -\ntions in GALNT12 leading to abnormal glycosylation \nplay a critical role in the pathogenesis of colorectal can -\ncer [54]. Moreover, elevated GALNT12 expression is sig -\nnificantly associated with poor prognosis in patients with \nglioblastoma, where it enhances tumor cell prolifera -\ntion and invasiveness via modulation of the PI3 K/AKT/\nmTOR signaling pathway [55]. Additionally, GALNT12 \nhas been closely linked to IgA1 galactose deficiency, with \nsignificantly lower mRNA expression levels observed in \naffected individuals compared to healthy controls [56].\nThe RMI1 (RecQ Mediated Genome Instability 1) \ngene, also localized on chromosome 9, encodes a key \nprotein involved in DNA repair and recombination. As \nan integral component of the BLM/RMI1/Top3α com -\nplex, RMI1 plays a pivotal role in maintaining genomic \nstability and facilitating DNA damage repair [57]. Loss \nof RMI1 function leads to increased DNA damage \naccumulation, cell cycle arrest, and impaired homolo -\ngous recombination repair, particularly following ioniz -\ning radiation exposure [58]. Beyond its role in genomic \nmaintenance, RMI1 is involved in metabolic regulation, \nwith its expression in adipocytes being modulated by \nglucose through the E2 F pathway [59]. RMI1-deficient \nFig. 7 The regulatory networks associated with key genes. A The mRNA-miRNA network of key genes. B The mRNA-TF network of key genes. C The \nkey genes-drugs network\n\nPage 12 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nmice exhibit resistance to diet- and genetically induced \nobesity, highlighting its involvement in metabolic home -\nostasis [60]. Furthermore, mutations in RMI1 contribute \nto the pathogenesis of Bloom syndrome, a genetic disor -\nder characterized by primary microcephaly, intrauterine \ngrowth restriction, and short stature [61].\nAlthough direct evidence linking GALNT12, RMI1, \nand VRK1 to endometriosis remains limited, their well-\ndocumented roles in gene expression regulation, cell \nsignaling, DNA repair, cell cycle progression, and apop -\ntosis suggest potential involvement in the disease’s patho-\ngenesis. For instance, mutations in GALNT12 or RMI1 \nleading to aberrant protein function may compromise the \nstability and proliferative capacity of endometrial cells. \nSimultaneously, dysregulated VRK1 activity could dis -\nrupt normal cell cycle control, potentially contributing to \nthe onset and progression of endometriosis.\nImmunofiltration analysis identified 11 distinct \nimmune cell types exhibiting differential infiltration pat -\nterns in endometriosis. Notably, M2 macrophages dem -\nonstrated reduced abundance in endometriotic tissues, \nwhereas resting NK cells were significantly enriched. \nM2 macrophages are recognized for their role in tissue \nrepair, angiogenesis, and tumor progression [62]. Prior \nstudies have reported a marked decline in M2 mac -\nrophage proportions across all stages of endometriosis \nin affected individuals [63]. Consistent with these find -\nings, the key genes identified in this study were down -\nregulated in ectopic endometrial tissues and exhibited \na positive correlation with M2 macrophage infiltration. \nNK cells, as critical components of the innate immune \nsystem, contribute to immune surveillance and tissue \nhomeostasis. Within the endometrium, a specialized \nsubset known as uterine NK (uNK) cells has been identi -\nfied [64]. Research indicates that  CD16+ uNK cells pro -\nduce cytotoxic factors capable of affecting trophoblast \nfunction, potentially leading to infertility, miscarriage, or \nplacental abnormalities [65]. While this study observed a \nnegative correlation between key genes and resting NK \ncells, alongside increased NK cell infiltration in ectopic \nendometrial tissues, the precise role of NK cells in endo -\nmetriosis remains inconclusive [66]. Further investiga -\ntion with larger sample cohorts and additional functional \nvalidation is required.\nMicroRNAs (miRNAs), a class of short non-coding \nRNAs, regulate gene expression post-transcriptionally \nby modulating mRNA stability and translation efficiency. \nAberrant miRNA expression has been extensively docu -\nmented in endometriosis. This study predicted miRNA \ninteractions with the three key genes, highlighting \nmiR- 202, which has been reported to be upregulated \nin ectopic endometrial tissue. Notably, miR- 202 sup -\npresses SOX6 expression, thereby enhancing the invasive \ncapacity of ectopic endometrial cells [67]. Although the \nmiRNAs identified in this study have not been directly \ninvestigated in endometriosis, their involvement in other \npathological conditions has been documented. In cervi -\ncal cancer, RGMB-AS1 promotes tumor proliferation and \ninvasiveness via the miR- 4428/PBX1 axis [68], while in \novarian cancer, miR- 6086 suppresses angiogenesis by \ndownregulating the OC2/VEGFA/EGFL6 signaling path -\nway [69]. Within the mRNA-TF regulatory network, SPI1 \nFig. 8 Immunohistochemical Validation of GALNT12, RMI1 and VRK1 in Normal and Endometriosis\n\nPage 13 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \nwas identified as a shared transcriptional regulator of the \nthree key genes. Notably, SPI1 is upregulated in ectopic \nendometrial tissues, contributing to the aggressive phe -\nnotype of endometriotic lesions [70]. Furthermore, drug \nrepurposing analysis using the CTD and Enrichr data -\nbases identified 195 drug candidates targeting the key \ngenes. This network encompasses a diverse range of \ntherapeutic agents, including retinoic acid, which has \ndemonstrated potential for endometriosis treatment by \ninhibiting estradiol secretion in ovarian endometriotic \ncysts and attenuating disease progression [71]. Addi -\ntionally, while enterolactones have not been studied in \nthe context of endometriosis, their therapeutic potential \nin other malignancies has been explored. Specifically, \nenterolactones have been shown to enhance radiotherapy \nefficacy in breast cancer by inhibiting DNA repair mech -\nanisms and promoting apoptotic pathways [72].\nThe mechanisms underlying targeted drug actions are \nhighly intricate, with potential impacts on disease pro -\ngression mediated through diverse pathways. While most \nstudies suggest that targeted therapies exert their effects \nprimarily by downregulating the expression of target \ngenes [73, 74], their functional scope extends beyond \nmere gene suppression. For example, TP53 serves as a \npivotal tumor suppressor gene, and its functional loss is \nimplicated in the pathogenesis of numerous malignan -\ncies. Restoring TP53 activity via targeted therapies can \nreestablish its antitumor function, thereby inhibiting \ntumor progression [75]. Similarly, FoxP3, a key transcrip-\ntion factor essential for the development and function \nof Tregs, plays a critical role in immune modulation. \nUpregulation of FoxP3 enhances the immunosuppressive \ncapacity of Tregs, influencing the onset and progression \nof esophageal cancer [76]. Given the multifaceted mecha-\nnisms of targeted drugs, identifying effective therapeutic \ntargets is imperative for advancing treatment strategies, \nimproving clinical outcomes, and improving patient \nprognosis.\nThis study has inherent limitations stemming from its \nreliance on data sourced from multiple public databases, \nwhich may introduce potential biases. Furthermore, the \nanalysis is predominantly bioinformatics-driven, lacking \nextensive experimental validation. Although preliminary \nimmunohistochemical analysis corroborated the compu -\ntational findings regarding the expression of key genes in \ntissue samples, additional validation is required. Future \nwork will focus on quantifying gene expression using \nWestern blot and quantitative polymerase chain reac -\ntion (qPCR) methodologies. Moreover, functional assays \nwill be conducted at the cellular level, including gene \noverexpression experiments to assess the impact of these \ngenetic alterations on cell proliferation, migration, inva -\nsion, and apoptosis.\nConclusions\nOverall, this study employed an integrative approach \ncombining differential gene expression analysis, \nWGCNA, MR analysis, and machine learning to iden -\ntify three key genes associated with palmitoylation in \nendometriosis. Subsequent analyses explored immune \ninfiltration dynamics, gene functional similarity, and \npharmacological correlations. These findings provide \nnovel insights that may inform clinical diagnostics, \ndisease surveillance, and therapeutic development for \nendometriosis.\nAbbreviations\nMR  Mendelian Randomization\nGEO  Gene Expression Omnibus\nGWAS  Genome-Wide Association Studies\nPE  Pelvic endometriosis\nIEU  Integrative Epidemiology Unit\nSNPs  Single nucleotide polymorphisms\nPRGs  Palmitoylation relacated genes\nWGCNA  Weighted Gene Co-expression Network Analysis\nssGSEA  Single-sample gene set enrichment analysis\nDEGs  Differentially expressed genes\nGO  Gene Ontology\nKEGG  Kyoto Encyclopedia of Genes and Genomes\nBP  Biological process\nCC  Cellular component\nMF  Molecular function\nIVs  Instrumental variables\nLD  Linkage disequilibrium\nIVW  Inverse Variance Weighted\nLOO  Leave-one-out\nRF  Random Forest\nSVM  Support Vector Machine\nGLM  Generalized Linear Model\nKNN  K-Nearest Neighbor\nROC  Receiver operating characteristic\nAUC   Area under the curve\nCTD  Comparative toxicogenomics\nNK  Nature killer\nTregs  Regulatory T cells\nUCSC  University of California Santa Cruz\nTFs  Transcription factors\nuNK  Uterine natural killer\nIHC  Immunohistochemistry.\nFC  Fold Change\nqPCR  Quantitative Polymerase Chain Reaction\nSupplementary Information\nThe online version contains supplementary material available at https:// doi. \norg/ 10. 1186/ s12905- 025- 03697-0.\nSupplementary Material 1.\nAcknowledgements\nThe authors would like to express their gratitude to the generous contributors \nof the GEO and GWAS databases for sharing their valuable data.\nAuthors’ contributions\nQ. C. and J. F. designed the thesis. J. K. carried out the study data analysis and \ncontributed to the writing-original draft, review & editing. J. S. contributed \nin the calibration of the data and the figures. Y. Y. contributed in the software \nand hardware maintenance. X. L. contributed to assist in writing-original draft, \nreview & editing. H. H. managed the typesetting of the manuscript. All authors \nhave read and approved the final version of the manuscript.\n\nPage 14 of 15Kai et al. BMC Women’s Health          (2025) 25:161 \nFunding\nThis work was supported by the National Natural Science Foundation of China \n(No. 82104908).\nData availability\nAll data generated or analysed during this study are included in this published \narticle and its supplementary information files. The datasets used for analysis \nin this paper are derived from GEO (https:// www. ncbi. nlm. nih. gov/ gds) and \nIEU OpenGWAS (https:// gwas. mrcieu. ac. uk/) databases.\nDeclarations\nEthics approval and consent to participate\nThe study was approved by Shanghai General Hospital Institutional Review \nBoard. The approval number is 20240711101424110.\nConsent for publication\nNot applicable.\nCompeting interests\nThe authors declare no competing interests.\nAuthor details\n1 Department of Clinical Medical Laboratory, The Affiliated Second Hospital \nof Xiamen Medical College, Xiamen, Fujian, China. 2 Department of Pathol-\nogy, The Affiliated Second Hospital of Xiamen Medical College, Xiamen, \nFujian, China. 3 Department of Microbiology, Guilin Medical University, Guilin, \nGuangxi, China. 4 Department of Laboratory Medicine, Shanghai General \nHospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. \n5 Department of Traditional Chinese Medicine, Shanghai General Hospital, \nShanghai Jiao Tong University School of Medicine (Originally Named “Shang-\nhai First People’s Hospital”), Shanghai, China. \nReceived: 31 August 2024   Accepted: 28 March 2025\nReferences\n 1. Allaire C, Bedaiwy MA, Yong PJ. Diagnosis and management of endome-\ntriosis. Can Med ASsoc J. 2023;195(10):E363–71.\n 2. Symons LK, Miller JE, Kay VR, Marks RM, Liblik K, Koti M, Tayade C. \nThe Immunopathophysiology of Endometriosis. Trends Mol Med. \n2018;24(9):748–62.\n 3. Nnoaham KE, Hummelshoj L, Webster P , D’Hooghe T, de Cicco NF, de \nCicco NC, Jenkinson C, Kennedy SH, Zondervan KT. Impact of endome-\ntriosis on quality of life and work productivity: a multicenter study across \nten countries. Fertil Steril. 2011;96(2):366–73.\n 4. Burney RO, Giudice LC. Pathogenesis and pathophysiology of endome-\ntriosis. Fertil Steril. 2012;98(3):511–9.\n 5. Sanchez AM, Viganò P , Somigliana E, Cioffi R, Panina-Bordignon P , Candi-\nani M. The endometriotic tissue lining the internal surface of endome-\ntrioma: hormonal, genetic, epigenetic status, and gene expression profile. \nReprod Sci. 2015;22(4):391–401.\n 6. Reis FM, Petraglia F, Taylor RN. Endometriosis: hormone regulation \nand clinical consequences of chemotaxis and apoptosis. Hum Reprod \nUpdate. 2013;19(4):406–18.\n 7. Leyland N, Casper R, Laberge P , Singh SS. Endometriosis: diagnosis and \nmanagement. J Obstet Gynaecol CA. 2010;32(7 Suppl 2):S1–32.\n 8. Bedaiwy MA, Allaire C, Alfaraj S. Long-term medical management of \nendometriosis with dienogest and with a gonadotropin-releasing \nhormone agonist and add-back hormone therapy. Fertil Steril. \n2017;107(3):537–48.\n 9. Donnez J, Squifflet J. Complications, pregnancy and recurrence in \na prospective series of 500 patients operated on by the shaving \ntechnique for deep rectovaginal endometriotic nodules. Hum Reprod. \n2010;25(8):1949–58.\n 10. Gu M, Jiang H, Tan M, Yu L, Xu N, Li Y, Wu H, Hou Q, Dai C. Palmitoyl-\ntransferase DHHC9 and acyl protein thioesterase APT1 modulate renal \nfibrosis through regulating β-catenin palmitoylation. Nat Commun. \n2023;14(1):6682.\n 11. Chamberlain LH, Shipston MJ. The physiology of protein S-acylation. \nPhysiol Rev. 2015;95(2):341–76.\n 12. Wang L, Cui J. Palmitoylation promotes chaperone-mediated \nautophagic degradation of NLRP3 to modulate inflammation. \nAutophagy. 2023;19(10):2821–3.\n 13. Zhou L, He X, Wang L, Wei P , Cai Z, Zhang S, Jin S, Zeng H, Cui J. Palmi-\ntoylation restricts SQSTM1/p62-mediated autophagic degradation of \nNOD2 to modulate inflammation. Cell Death Differ. 2022;29(8):1541–51.\n 14. Zhang M, Shi Z, Peng X, Cai D, Peng R, Lin Y, Dai L, Li J, Chen Y, Xiao \nJ, et al. NLRP3 inflammasome-mediated Pyroptosis induce Notch \nsignal activation in endometriosis angiogenesis. Mol Cell Endocrinol. \n2023;574: 111952.\n 15. Nishimoto-Kakiuchi A, Sato I, Nakano K, Ohmori H, Kayukawa Y, \nTanimura H, Yamamoto S, Sakamoto Y, Nakamura G, Maeda A et al: A \nlong-acting anti-IL-8 antibody improves inflammation and fibrosis in \nendometriosis. Sci Transl Med 2023, 15(684):eabq5858.\n 16. Huang Y, Huang Y, Cai Z, Ferrari MW, Li C, Zhang T, Lyu G, Wang Z. \nMiR-21-3p inhibitor exerts myocardial protective effects by altering \nmacrophage polarization state and reducing excessive mitophagy. \nCommun Biol. 2024;7(1):1371.\n 17. Yang S, Li M, Lian G, Wu Y, Cui J, Wang L: ABHD8 antagonizes inflam-\nmation by facilitating chaperone-mediated autophagy-mediated \ndegradation of NLRP3. Autophagy 2024:1–14.\n 18. Nie L, Fei C, Fan Y, Dang F, Zhao Z, Zhu T, Wu X, Dai T, Balasubramanian \nA, Pan J, et al. Consecutive palmitoylation and phosphorylation orches-\ntrates NLRP3 membrane trafficking and inflammasome activation. Mol \nCell. 2024;84(17):3336–53.\n 19. Zhu XX, Meng XY, Zhang AY, Zhao CY, Chang C, Chen TX, Huang YB, \nXu JP , Fu X, Cai WW, et al. Vaccarin alleviates septic cardiomyopathy by \npotentiating NLRP3 palmitoylation and inactivation. Phytomedicine. \n2024;131: 155771.\n 20. Koster KP , Fyke Z, Nguyen T, Niqula A, Noriega-González LY, Woolfrey \nKM, Dell’Acqua ML, Cologna SM, Yoshii A. Akap5 links synaptic dysfunc-\ntion to neuroinflammatory signaling in a mouse model of infantile \nneuronal ceroid lipofuscinosis. Front Synaptic Neuro. 2024;16:1384625.\n 21. Zhou B, Hao Q, Liang Y, Kong E. Protein palmitoylation in can-\ncer: molecular functions and therapeutic potential. Mol Oncol. \n2023;17(1):3–26.\n 22. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for \nmicroarray and RNA-seq data. BMC Bioinformatics. 2013;14:7.\n 23. Langfelder P , Horvath S. WGCNA: an R package for weighted correlation \nnetwork analysis. BMC Bioinformatics. 2008;9:559.\n 24. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. limma pow-\ners differential expression analyses for RNA-sequencing and microarray \nstudies. Nucleic Acids Res. 2015;43(7): e47.\n 25. Gustavsson EK, Zhang D, Reynolds RH, Garcia-Ruiz S, Ryten M. ggtran-\nscript: an R package for the visualization and interpretation of transcript \nisoforms using ggplot2. Bioinformatics. 2022;38(15):3844–6.\n 26. Gu Z. Complex heatmap visualization Imeta. 2022;1(3): e43.\n 27. Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R package for comparing \nbiological themes among gene clusters. OMICS. 2012;16(5):284–7.\n 28. Zhou H, Zhang Y, Liu J, Yang Y, Fang W, Hong S, Chen G, Zhao S, Zhang \nZ, Shen J, et al. Education and lung cancer: a Mendelian randomization \nstudy. Int J Epidemiol. 2019;48(3):743–50.\n 29. Obenchain V, Lawrence M, Carey V, Gogarten S, Shannon P , Morgan M. \nVariantAnnotation: a Bioconductor package for exploration and annota-\ntion of genetic variants. Bioinformatics. 2014;30(14):2076–8.\n 30. Hemani GEBP: ieugwasr: Interface to the “OpenGWAS” Database API. R \npackage version 1.0.1 edition.; 2024.\n 31. Bowden J, Davey SG, Burgess S. Mendelian randomization with invalid \ninstruments: effect estimation and bias detection through Egger regres-\nsion. Int J Epidemiol. 2015;44(2):512–25.\n 32. Bowden J, Davey SG, Haycock PC, Burgess S. Consistent Estimation \nin Mendelian Randomization with Some Invalid Instruments Using a \nWeighted Median Estimator. Genet Epidemiol. 2016;40(4):304–14.\n 33. Burgess S, Butterworth A, Thompson SG. Mendelian randomization \nanalysis with multiple genetic variants using summarized data. Genet \nEpidemiol. 2013;37(7):658–65.\n\nPage 15 of 15\nKai et al. BMC Women’s Health          (2025) 25:161 \n \n 34. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D, Laurin C, \nBurgess S, Bowden J, Langdon R et al: The MR-Base platform supports \nsystematic causal inference across the human phenome. Elife 2018, 7.\n 35. Hartwig FP , Davey SG, Bowden J. Robust inference in summary data \nMendelian randomization via the zero modal pleiotropy assumption. Int J \nEpidemiol. 2017;46(6):1985–98.\n 36. Lu L, Wan B, Li L, Sun M. Hypothyroidism has a protective causal associa-\ntion with hepatocellular carcinoma: A two-sample Mendelian randomiza-\ntion study. Front Endocrinol. 2022;13: 987401.\n 37. Verbanck M, Chen CY, Neale B, Do R. Detection of widespread horizontal \npleiotropy in causal relationships inferred from Mendelian randomization \nbetween complex traits and diseases. Nat Genet. 2018;50(5):693–8.\n 38. Jin T, Huang W, Cao F, Yu X, Guo S, Ying Z, Xu C. Causal association \nbetween systemic lupus erythematosus and the risk of dementia: A \nMendelian randomization study. Front Immunol. 2022;13:1063110.\n 39. Van Essen DC. Cortical cartography and Caret software. Neuroimage. \n2012;62(2):757–64.\n 40. Guan C, Ma F, Chang S, Zhang J. Interpretable machine learning models \nfor predicting venous thromboembolism in the intensive care unit: an \nanalysis based on data from 207 centers. Crit Care. 2023;27(1):406.\n 41. Chen B, Khodadoust MS, Liu CL, Newman AM, Alizadeh AA. Profiling \nTumor Infiltrating Immune Cells with CIBERSORT. Methods Mol Biol. \n2018;1711:243–59.\n 42. Zhang H, Meltzer P , Davis S. RCircos: an R package for Circos 2D track \nplots. BMC Bioinformatics. 2013;14:244.\n 43. Yang L, Yu X, Liu M, Cao Y. A comprehensive analysis of biomarkers associ-\nated with synovitis and chondrocyte apoptosis in osteoarthritis. Front \nImmunol. 2023;14:1149686.\n 44. Shannon P , Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin \nN, Schwikowski B, Ideker T. Cytoscape: a software environment for \nintegrated models of biomolecular interaction networks. Genome Res. \n2003;13(11):2498–504.\n 45. Zhao L, Zhang C, Luo X, Wang P , Zhou W, Zhong S, Xie Y, Jiang Y, Yang P , \nTang R, et al. CD36 palmitoylation disrupts free fatty acid metabolism and \npromotes tissue inflammation in non-alcoholic steatohepatitis. J Hepatol. \n2018;69(3):705–17.\n 46. Lee J, Balzraine B, Schweizer A, Kuzmanova V, Gwack Y, Razani B, Lee JM, \nMosher DF, Cho J: Neutrophil CRACR2A Promotes Neutrophil Recruit-\nment in Sterile Inflammation and Ischemic Stroke. Circulation 2024.\n 47. Ma Y, Chen Y, Li L, Wu Z, Cao H, Zhu C, Liu Q, Wang Y, Chen S, Liu Y et al: \n2-Bromopalmitate-Induced Intestinal Flora Changes and Testicular Dys-\nfunction in Mice. Int J Mol Sci 2024, 25(21).\n 48. Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S. The \nprotein kinase complement of the human genome. Science. \n2002;298(5600):1912–34.\n 49. Budziszewski GR, Zhao Y, Spangler CJ, Kedziora KM, Williams MR, Azzam \nDN, Skrajna A, Koyama Y, Cesmat AP , Simmons HC, et al. Multivalent DNA \nand nucleosome acidic patch interactions specify VRK1 mitotic localiza-\ntion and activity. Nucleic Acids Res. 2022;50(8):4355–71.\n 50. Vega FM, Sevilla A, Lazo PA. p53 Stabilization and accumulation induced \nby human vaccinia-related kinase 1. Mol Cell Biol. 2004;24(23):10366–80.\n 51. Kang TH, Park DY, Kim W, Kim KT. VRK1 phosphorylates CREB and medi-\nates CCND1 expression. J Cell SCI. 2008;121(Pt 18):3035–41.\n 52. Salzano M, Sanz-García M, Monsalve DM, Moura DS, Lazo PA. VRK1 \nchromatin kinase phosphorylates H2AX and is required for foci formation \ninduced by DNA damage. Epigenetics-US. 2015;10(5):373–83.\n 53. Ping X, Stark JM. O-GlcNAc transferase is important for homology-\ndirected repair. DNA Repair. 2022;119: 103394.\n 54. Guda K, Moinova H, He J, Jamison O, Ravi L, Natale L, Lutterbaugh J, \nLawrence E, Lewis S, Willson JK, et al. Inactivating germ-line and somatic \nmutations in polypeptide N-acetylgalactosaminyltransferase 12 in human \ncolon cancers. P Natl Acad Sci USA. 2009;106(31):12921–5.\n 55. Zheng Y, Liang M, Wang B, Kang L, Yuan Y, Mao Y, Wang S. GALNT12 is \nassociated with the malignancy of glioma and promotes glioblastoma \nmultiforme in vitro by activating Akt signaling. Biochem Bioph res CO. \n2022;610:99–106.\n 56. Wang YN, Zhou XJ, Chen P , Yu GZ, Zhang X, Hou P , Liu LJ, Shi SF, Lv JC, \nZhang H. Interaction between GALNT12 and C1GALT1 Associates with \nGalactose-Deficient IgA1 and IgA Nephropathy. J Am Soc Nephrol. \n2021;32(3):545–52.\n 57. Yang J, O’Donnell L, Durocher D, Brown GW. RMI1 promotes DNA replica-\ntion fork progression and recovery from replication fork stress. Mol Cell \nBiol. 2012;32(15):3054–64.\n 58. Fang L, Sun Y, Dong M, Yang M, Hao J, Li J, Zhang H, He N, Du L, Xu C. \nRMI1 facilitates repair of ionizing radiation-induced DNA damage and \nmaintenance of genomic stability. Cell Death Discov. 2023;9(1):426.\n 59. Suwa A, Yoshino M, Kurama T, Shimokawa T, Aramori I. Glucose regulates \nRMI1 expression through the E2F pathways in adipose cells. Endocrine. \n2011;40(1):56–61.\n 60. Suwa A, Yoshino M, Yamazaki C, Naitou M, Fujikawa R, Matsumoto S, \nKurama T, Shimokawa T, Aramori I. RMI1 deficiency in mice protects from \ndiet and genetic-induced obesity. Febs J. 2010;277(3):677–86.\n 61. Gönenc II, Elcioglu NH, Martinez GC, Aras S, Großmann N, Praulich I, \nAltmüller J, Kaulfuß S, Li Y, Nürnberg P , et al. Phenotypic spectrum of BLM- \nand RMI1-related Bloom syndrome. Clin Genet. 2022;101(5–6):559–64.\n 62. Mantovani A, Sozzani S, Locati M, Allavena P , Sica A. Macrophage polariza-\ntion: tumor-associated macrophages as a paradigm for polarized M2 \nmononuclear phagocytes. Trends Immunol. 2002;23(11):549–55.\n 63. Takebayashi A, Kimura F, Kishi Y, Ishida M, Takahashi A, Yamanaka A, Wu D, \nZheng L, Takahashi K, Suginami H, et al. Subpopulations of macrophages \nwithin eutopic endometrium of endometriosis patients. Am J Reprod \nImmunol. 2015;73(3):221–31.\n 64. Thiruchelvam U, Wingfield M, O’Farrelly C. Increased uNK Progenitor \nCells in Women With Endometriosis and Infertility are Associated With \nLow Levels of Endometrial Stem Cell Factor. Am J Reprod Immunol. \n2016;75(4):493–502.\n 65. Lee SK, Kim CJ, Kim DJ, Kang JH. Immune cells in the female reproductive \ntract. Immune Netw. 2015;15(1):16–26.\n 66. Thiruchelvam U, Wingfield M, O’Farrelly C. Natural Killer Cells: Key Players \nin Endometriosis. Am J Reprod Immunol. 2015;74(4):291–301.\n 67. Zhang D, Li Y, Tian J, Zhang H, Wang S. MiR-202 promotes endometriosis \nby regulating SOX6 expression. Int J Clin Exp Med. 2015;8(10):17757–64.\n 68. Shen C, Wang B, Zhang K, Wang C, Wang J, An Z, Shu L. RGMB-AS1/miR-\n4428/PBX1 axis drives the progression of cervical cancer. Transl Cancer \nRes. 2020;9(5):3180–90.\n 69. Wu B, Zhang L, Yu Y, Lu T, Zhang Y, Zhu W, Song Q, Lv C, Guo J, Tian Y, et al. \nmiR-6086 inhibits ovarian cancer angiogenesis by downregulating the \nOC2/VEGFA/EGFL6 axis. Cell Death Dis. 2020;11(5):345.\n 70. Lee NK, Lee JW, Woo JH, Choi YS, Choi JH. Upregulation of SPI1 in Ectopic \nEndometrium Contributes to an Invasive Phenotype. Arch Med Res. \n2023;54(2):86–94.\n 71. Yamagata Y, Takaki E, Shinagawa M, Okada M, Jozaki K, Lee L, Sato S, \nMaekawa R, Taketani T, Asada H, et al. Retinoic acid has the potential to \nsuppress endometriosis development. J Ovarian Res. 2015;8:49.\n 72. Bigdeli B, Goliaei B, Masoudi-Khoram N, Jooyan N, Nikoofar A, Rouhani \nM, Haghparast A, Mamashli F. Enterolactone: A novel radiosensitizer \nfor human breast cancer cell lines through impaired DNA repair and \nincreased apoptosis. Toxicol Appl Pharm. 2016;313:180–94.\n 73. Chen JJ, Xiao ZJ, Meng X, Wang Y, Yu MK, Huang WQ, Sun X, Chen \nH, Duan YG, Jiang X, et al. MRP4 sustains Wnt/β-catenin signaling \nfor pregnancy, endometriosis and endometrial cancer. Theranostics. \n2019;9(17):5049–64.\n 74. Ivanga M, Labrie Y, Calvo E, Belleau P , Martel C, Luu-The V, Morissette J, \nLabrie F, Durocher F. Temporal analysis of E2 transcriptional induction of \nPTP and MKP and downregulation of IGF-I pathway key components in \nthe mouse uterus. Physiol Genomics. 2007;29(1):13–23.\n 75. Wallis B, Bowman KR, Lu P , Lim CS: The Challenges and Prospects of p53-\nBased Therapies in Ovarian Cancer. Biomolecules 2023, 13(1).\n 76. Wang Y, Xue L. Decoding the role of FOXP3 in esophageal cancer: \nUnderlying mechanisms and therapeutic implications. BBA-REV Cancer. \n2024;1879(6): 189211.\nPublisher’s Note\nSpringer Nature remains neutral with regard to jurisdictional claims in pub-\nlished maps and institutional affiliations.","source_license":"CC0","license_restricted":false}