Abstract Background The pathogenesis of recurrent pelvic organ prolapse (POP) is currently unclear. Therefore, developing targeted preventive measures is difficult. This study identified potential key pathways, crucial genes, comorbidities, and therapeutic targets associated with the occurrence and development of recurrent POP. Methods The original microarray data [31]GSE28660, [32]GSE53868, and [33]GSE12852 were downloaded from the GEO database. Identification and validation of differentially expressed genes (DEGs) and hub genes associated with recurrent POP were performed using R software and cytoHubba of Cytoscape. Protein–protein interaction (PPI) networks were constructed using the STRING tool and visualized using Cytoscape. Gene ontology (GO) and Kyoto Encyclopedia of Gene and Genome (KEGG) enrichment analyses were effectively performed using DAVID platforms. In addition, the NetworkAnalyst platform was used to explore and visualize the miRNA–hub gene network, TF–hub gene network, hub gene–disease network, and hub gene–drug/chemical network. Results A total of 110 DEGs and 6 hub genes (ADIPOQ, IL6, PPARG, CEBPA, LPL, and LIPE) were identified in this study. These genes were primarily enriched in the PPAR, AMPK, and adipocytokine, non-alcoholic fatty liver disease, and signaling pathways related to glycerol metabolism. Moreover, 96 miRNAs and 97 TFs were identified to as being associated with recurrent POP. These genes were closely linked to adipocyte metabolism and distribution, energy metabolism, and the longevity regulatory pathway. In addition, 192 diseases or chronic complications were potentially related to the recurrence of POP, including diabetes, hypertension, obesity, inflammatory diseases, and chronic obstructive pulmonary disease. Furthermore, 954 drugs or compounds were shown to have therapeutic potential for recurrent POP, and the most critical target drugs were dexamethasone, bisphenol A, efavirenz, 1-methyl-3-isobutylxanthine, and estradiol. Conclusions The results of this study revealed that ADIPOQ, IL6, PPARG, CEBPA, LPL, and LIPE as potential hub genes associated with recurrent POP, and these hub genes may aid in the understanding of the mechanism underlying POP recurrence and the development of potential molecular drugs. Keywords: Pelvic organ prolapse, Recurrence, Bioinformatic analysis, Differentially expressed genes, Therapeutic target 1Introduction Pelvic organ prolapse (POP) is one of the most common clinical conditions. POP is defined as the falling, slipping, or downward displacement of the uterus and/or vagina as well as its adjacent organs, such as the vaginal vault, the bladder, and the rectum [[34]1]. Epidemiological studies have reported that the overall prevalence rate of POP in women worldwide is 3%–6%, and that POP affects up to as much as 50% of multiparous women [[35]2,[36]3]. Approximately 12% of POP cases show obvious symptoms. Most POP patients typically complain of vaginal bulging and malaise, as well as urination, defecation, and sexual dysfunction. Therefore, symptomatic POP affects the daily activity and quality of life of women and thus requires careful treatment [[37]4]. Based on the symptoms and types of prolapse, the management of POP is mainly performed using any of the three methods: follow-up observation, conservative treatment, or surgical treatment [[38]5]. In cases of mild prolapse, follow-up observation and conservative treatment are appropriate. As age increases, prolapse often necessitates surgical treatment [[39]6,[40]7]. The cumulative risk of women undergoing prolapse surgery at age 80 is 12.6%. As the population ages continuously, the total number of POP operations will continue to increase by 50% over the next 40–50 years [[41]8,[42]9]. The overall success rate of POP surgery is generally 50%–80%, and the rate of repeat surgery for recurrent POP within 10 years is 17% [[43]10]. Therefore, focusing on the POP recurrence is the key to thoroughly resolving the adverse effects and economic burden of POP. The definition of recurrent POP is currently unclear. “Recurrence” refers to the failure of a previous surgery, which may be subjective or objective [[44]11]. Furthermore, estimation of the incidence rate of recurrent POP is challenging due to the lack of a consistent definition and a powered sufficiently statistical analysis [[45]11,[46]12]. In addition, studies investigating the natural history of POP recurrence are limited, and it is difficult to understand the disease process and pathological mechanism of POP recurrence [[47]13,[48]14]. Much of the cognition regarding recurrent POP cannot be completely clarified, but the cause of recurrent POP is clearly multifactorial. Some studies have attempted to identify the risk factors for recurrence after autologous tissue repair or pelvic reconstructive surgery. Risk factors include the preoperative prolapse stage, levator ani avulsion, hiatal region, younger age, and family history [[49][14], [50][15], [51][16]]. A growing number of studies have reported the significance of genetic factors in POP. These genetic factors may lead to inherited weaknesses or repair defects of the pelvic floor, resulting in the occurrence and/or recurrence of POP [[52][17], [53][18], [54][19]]. However, how these factors interact to affect the pelvic floor support structure and cause prolapse recurrence has not been fully elucidated. A comprehensive analysis of the specific mechanisms and regulatory pathways of family history, genetic factors, and susceptibility genes in POP recurrence is needed. Currently, high-throughput technology is more extensively used in medical research, diagnostic marker identification, disease molecular typing, prognosis prediction models, and development of new targeted drugs [[55]20,[56]21]. Here, we downloaded the original microarray data [57]GSE28660, [58]GSE53868, and [59]GSE12852 from the GEO database, and identified the differentially expressed genes (DEGs) in the recurrent POP and control groups. Gene ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) were used to determine the functional enrichment of the DEGs. Furthermore, hub genes were identified and validated using the protein–protein interaction (PPI) network, algorithms analysis, and external datasets. Then, the miRNA and TF regulatory networks, disease network, and drug/chemical network of the hub genes were constructed and analyzed. In this study, we used bioinformatic tools to predict potential key pathways, crucial genes, comorbidities, and therapeutic targets associated with the occurrence and development of recurrent POP. These data can be applied to the prevention and management of recurrent POP. 2. Methods 2.1. Microarray data and download This study was performed in accordance with the criteria of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [[60]22] ([61]Supplementary Table S1). The datasets [62]GSE28660, [63]GSE53868, and [64]GSE12852 datasets were acquired from the GEO database of the National Center for Biotechnology Information (NCBI-GEO). The GEO accession ID of the recurrent POP dataset was [65]GSE28660, which was the expression profiling data of uterosacral ligament (USL) tissue expression profiling data obtained via high throughput DNA microarrays of the human whole genome and was contributed by Eyster KM. The sample consisted of three groups of patients with a total of 14 specimens, The sample consisted of recurrent prolapse (n = 5), primary prolapse (n = 5), and normal control (n = 4). The [66]GSE53868 dataset was the expression profiling data for the anterior vaginal wall of 12 pairs of prolapse and normal control groups and was submitted by Kerkhof MH et al. The [67]GSE12852 dataset was the expression profiling data of POP and was uploaded to the GEO database by Brizzolana SS et al., in 2008 [[68]23]. The [69]GSE12852 dataset comprised the microarray analysis results of USL and round ligament samples from 8 women with POP versus 9 controls without prolapse. In this study, the [70]GSE28660 dataset was used as the test dataset, and [71]GSE53868 and [72]GSE12852 datasets severed as the validation sets of the results. The comprehensive summary of the three datasets is shown in [73]Table 1. Table 1. The comprehensive summary information of the three data sets. GEO Platform Tissue (Homo sapiens) Samples Experiment type Attribute Organization Contact Author [74]GSE28660 [75]GPL2895 uterosacral ligament 12 Array University of South Dakota Kathleen M Eyster [76]GSE53868 [77]GPL18142 anterior vaginal wall 24 Array VU University Medical Center Manon Heleen Kerkhof [78]GSE12852 [79]GPL2986 uterosacral ligament 34 Array John A burns School of Mecicine Shawna S Brizzolara [80]Open in a new tab 2.2. DEG identification The raw data files were downloaded using the R software (version 3.5.2) with the GEO query packages, and the mRNA expression data were analyzed to identify differential expression genes (DEGs) by using the R package limma function [[81]24]. Based on the results of paired Student's t-test, hierarchical clustering analysis categorized the samples into two groups with expression patterns in POP and normal controls. Moreover, multiple tests and P values of results were corrected according to Benjamin and Hochberg methods so as to control the error detection rate at <5%. Finally, P < 0.05 and |logFC| ≥ 1 were set as the cut-off criteria for detecting significant DEGs from all datasets. 2.3. Functional enrichment analysis of DEGs In order to understand the biological functions, such as biological processes (BP), cellular components (CC), and molecular functions (MF), and to access the signaling pathways of DEGs involved in recurrent POP, the enrichment analysis results of gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were obtained and visualized using the R package ClusterProfile [[82]25]. Adjusted P < 0.05 and Q < 0.05 were considered standard metrics for statistical significance. 2.4. Protein–protein interaction network and module analysis of DEGs To interpret and gain insights into the gene function involved in recurrent POP, the Protein–protein interaction (PPI) network was established using the STRING database ([83]www.string-db.org). To generate the PPI network of DEGs, the minimum required composite score was established as 0.4 [[84]26]. Then, the PPI network was transferred to Cytoscape (v.3.7.1) software for further visualization and beautification [[85]27]. Subsequently, the MCODE plugin of Cytoscape was used to discover crucial MCODEs, with a false degree cutoff of 2, K-Core of 2, and node score cutoff of 0.2. 2.5. Identification and enrichment analysis of hub genes Cytohubba is a plug-in to the Cytoscape software that identifies the top 20 hub genes for recurrent POPs based on the shortest available path using BottleNeck, Closeness, Degree, DMNC, EcCeniciy and MCC algorithms. The common hub genes were screened and identified through the Venn diagram [[86]28]. Then, the determined hub genes were imported into the GeneMANIA online database ([87]http://genemania.org/) to build a putative PPI network [[88]29]. The total genes derived from the network were used for conducting the GO and KEGG pathway enrichment analyses by using Metascape ([89]https://metascape.org/gp/index.html#/main/step1). 2.6. Validation of hub genes The hub genes were validated using data from [90]GSE208261 and [91]GSE12852. The R package edgeR and limma functions were used for the differential expression analysis of RNA-seq and microarray data, respectively. Subsequently, the receiver operating characteristic (ROC) curve analysis was performed using the pROC package in R, and the AUC was calculated to estimate the diagnostic value of hub genes [[92]30]. The threshold for statistical significance was set at P < 0.05, and AUC >0.7 was considered to indicate acceptable predictive performance in discriminating between the recurrent POP group and the control group. 2.7. Regulatory network analysis of hub genes The NetworkAnalyst platform ([93]http://www.networkanalyst.ca) was used to explore and visualize comprehensive experimentally validated hub gene–miRNA interactions by using TarBase v8.0 [[94]31]. Combined with the human transcription factor (TF) information recorded using the ENCODE database, TFs of hub genes were explored, which were derived from the ENCODE ChIP-seq data [[95]32]. Both TF–hub gene and miRNA–hub gene interaction networks were illustrated and visualized using the NetworkAnalyst platform. Enrichment of the TF–hub and miRNA–hub gene networks by The KEGG pathway was performed, with P < 0.05 set as a threshold criterion. 2.8. Gene–disease association analysis DisGeNET ([96]https://www.disgenet.org/) is a comprehensive gene-disease association database, which is used to identify hub gene-related diseases and their chronic complications [[97]33]. The relevant steps were completed through the gene release association function of the NetworkAnalyst platform. 2.9. Protein–drug and protein–chemical interaction analysis The DrugBank database (Version 5.0) [[98]34] and Comparative Toxicogenomics Database (CTD, [99]http://ctdbase.org/) [[100]35] were applied to explore protein–drug and protein–chemical interactions in the hub genes identified as potential targets for existing drugs or chemical compounds through the NetworkAnalyst platform. The interaction networks were used to jointly visualize interaction networks. 3. Results 3.1. DEG identification in recurrent POP The [101]GSE28660 dataset was acquired from the NCBI-GEO, which was the expression profiling data of 14 USL tissue specimens of patients from the recurrent POP (n = 5), primary POP (n = 5), and normal control (n = 4) groups. The normalization, principal component analysis, and density diagram of profiling data are presented in [102]Fig. 1. As shown in [103]Fig. 2A–D, upon setting the cut-off criterion as genes with P < 0.05 and |logFC| ≥ 1, the cohort of recurrent POP vs. primary POP in [104]GSE28660 datasets contains 249 DEGS, including 178 upregulated genes and 71 downregulated genes. The cohort of recurrent POP vs. normal control contained 179 DEGs with 142 upregulated genes and 37 downregulated genes. We further identified the overlapping DEGs among the two cohorts. Venn diagram depicting the 110 overlapping DEGs visualization, including 14 downregulated and 94 upregulated genes ([105]Fig. 2E–F; [106]Supplementary Table S2). Fig. 1. [107]Fig. 1 [108]Open in a new tab Normalization of the microarray datasets of [109]GSE288660. (A) The expression value histogram of each sample in [110]GSE288660 dataset before normalization. (B) The normalized expression value histogram of each sample in [111]GSE288660 dataset. (C) The UMAP dimension reduction analysis diagram of each sample in [112]GSE288660 dataset. (D) The expression density diagram of each sample in [113]GSE288660 dataset. Fig. 2. [114]Fig. 2 [115]Open in a new tab The hierarchical clustering heatmap and volcano plot of DEGs between recurrent POP group and control group. (A) The hierarchical clustering heatmap of DEGs between recurrent POP group and primary POP group in [116]GSE28660 datasets. (B) The volcano plot of DEGs between recurrent POP group and primary POP group in [117]GSE28660 datasets.(C) The hierarchical clustering heatmap of DEGs between recurrent POP group and normal group in [118]GSE28660 datasets. (D) The volcano plot of DEGs between recurrent POP group and normal group in [119]GSE28660 datasets. (E) Venn diagram of the intersection of up-regulated genes and down-regulated genes among recurrent POP group, primary POP group and normal control group. Dataset 1 is the intersection gene of recurrent POP group and primary POP group, Dataset 2 is the intersection gene of recurrent POP group and normal POP group.(F)The up-regulated genes and down-regulated gene intersection genes were found in the recurrent POP group, the primary POP group and the normal control group. 3.2. Functional enrichment analysis of DEGs To further determine the potential biological functions of the 110 DEGs, GO and KEGG pathway enrichment analyses were performed. According to the GO analysis results, BPs closely related to recurrent POP were glucose homeostasis, lipid metabolism, response to bacteria, response to nutrients, and triglyceride biosynthesis ([120]Fig. 3 and [121]Table 2). CCs mainly related to recurrent POP were lipid particles, hemoglobin complex, extracellular space, extracellular region, and endoplasmic reticulum. MFs mainly related to recurrent POP were oxidoreductase activity, hormone activity, oxygen transporter activity, protein homodimerization activity, and oxygen binding. According to the results of the KEGG pathway enrichment analysis, the PPAR, AMPK, and adipocytokine signaling pathways; non-alcoholic fatty liver disease (NAFLD); and glycerolipid metabolism might be involved in recurrent POP. Fig. 3. [122]Fig. 3 [123]Open in a new tab GO terms and KEGG pathway enrichment between recurrent POP group and control group. (A) Biological process (BP) of DEGs. (B) Cellular component of DEGs. (C) Molecular function (MF) of DEGs. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG) of DEGs. The dot size reflects the count of concentrated DEG and the point color represents a negative Log10-p value. Table 2. Top 10 enrichment analysis items of DEGs between recurrent POP group and control group. Category Term Process % PValue FDR BP GO:0042593 glucose homeostasis 8.74 0.00 0.00 BP GO:0006629 lipid metabolic process 8.74 0.00 0.00 BP GO:0009617 response to bacterium 6.80 0.00 0.01 BP GO:0007584 response to nutrient 5.83 0.00 0.01 BP GO:0019432 triglyceride biosynthetic process 3.88 0.00 0.02 BP GO:0042572 retinol metabolic process 4.85 0.00 0.03 BP GO:0019915 lipid storage 3.88 0.00 0.05 BP GO:0071300 cellular response to retinoic acid 4.85 0.00 0.06 BP GO:0061037 negative regulation of cartilage development 2.91 0.00 0.06 BP GO:0045471 response to ethanol 5.83 0.00 0.07 CC GO:0005811 lipid particle 7.77 0.00 0.00 CC GO:0005833 hemoglobin complex 3.88 0.00 0.00 CC GO:0005615 extracellular space 20.39 0.00 0.04 CC GO:0005576 extracellular region 20.39 0.00 0.08 CC GO:0005783 endoplasmic reticulum 13.59 0.00 0.08 CC GO:0005581 collagen trimer 3.88 0.01 0.21 CC GO:0005789 endoplasmic reticulum membrane 11.65 0.01 0.23 CC GO:0009897 external side of plasma membrane 6.80 0.02 0.35 CC GO:0030176 integral component of endoplasmic reticulum membrane 3.88 0.02 0.35 CC GO:0031526 brush border membrane 2.91 0.04 0.46 MF GO:0016655 oxidoreductase activity 2.91 0.00 0.29 MF GO:0005179 hormone activity 4.85 0.00 0.30 MF GO:0005344 oxygen transporter activity 2.91 0.00 0.30 MF GO:0042803 protein homodimerization activity 10.68 0.00 0.30 MF GO:0019825 oxygen binding 2.91 0.01 0.81 MF GO:0018636 phenanthrene 9,10-monooxygenase activity 1.94 0.02 0.81 MF GO:0047086 ketosteroid monooxygenase activity 1.94 0.03 0.89 MF GO:0004771 sterol esterase activity 1.94 0.03 0.89 MF GO:0047023 androsterone dehydrogenase activity 1.94 0.04 0.89 MF GO:0071813 lipoprotein particle binding 1.94 0.04 0.89 KEGG hsa03320 PPAR signaling pathway 8.74 0.00 0.00 KEGG hsa04152 AMPK signaling pathway 7.77 0.00 0.00 KEGG hsa04920 Adipocytokine signaling pathway 4.85 0.00 0.10 KEGG hsa04932 Non-alcoholic fatty liver disease 5.83 0.01 0.26 KEGG hsa00561 Glycerolipid metabolism 3.88 0.01 0.37 KEGG hsa04550 Signaling pathways regulating pluripotency of stem cells 4.85 0.02 0.63 KEGG hsa04512 ECM-receptor interaction 3.88 0.03 0.67 KEGG hsa00564 Glycerophospholipid metabolism 3.88 0.04 0.74 KEGG hsa04975 Fat digestion and absorption 2.91 0.04 0.74 KEGG hsa04931 Insulin resistance 3.88 0.05 0.76 [124]Open in a new tab 3.3. PPI network and module analysis Using the STRING database and Cytoscape (v 3.9.0) software, we generated and visualized the PPI network for DEGs with 71 nodes and 228 edges ([125]Fig. 4). Seven genes were down-regulated in the PPI network, and the vast majority of nodes were DEGs up-regulated in the network ([126]Fig. 4A). The MCODE plugin of Cytoscape was used to discover critical MCODEs, with a false degree cutoff of 2, a K-Core of 2, and a node score cutoff of 0.2. Four significant modules (Modules 1–4) with a score of ≥5 were screened out through MCODE. Module 1 contains 18 crucial node genes: IL6, AGPAT2, SCD, PCK1, LPL, MLXIPL, CIDEA, PLIN1, PPARG, LEP, GPAM, FABP4, LIPE, DGAT2, ACSL1, CEBPA, CD36, and ADIPOQ ([127]Fig. 4B). Module 2 contains SHH, SIX1, ISL1, and NEUROG1 as hub nodes ([128]Fig. 4C). The two additional modules Module 3 (key nodes were THBS4, ITGA8, and COL4A5) and Module 4 (key nodes were HBM, AHSP, and HBD) were also important support networks in DEGs ([129]Fig. 4D and E). The significant modules were primarily involved in multiple pathways using the DAVID platform, including the PPAR signaling pathway, the regulation of lipocytosis in adipocytes and NAFLD by the KEGG-enriched method; translational regulation of white adipocyte differentiation, developmental biology, and the triglyceride mainland by the REACTOME enriched method; and the PPAR signaling pathway, AMP activated protein kinase signaling, and regulation of transcription factors in adipogenesis by the method enriched with WIKIPATHWAYS ([130]Fig. 4F). Fig. 4. [131]Fig. 4 [132]Open in a new tab The Protein-protein interaction (PPI) network and module analysis of the DEGs. (A) The PPI network of all DEGs.(B) The PPI of Module 1. It contains 18 nodes, namely IL6, AGPAT2, SCD, PCK1, LPL, MLXIPL, CIDEA, PLIN1, PPARG, LEP, GPAM, FABP4, LIPE, DGAT2, ACSL1, CEBPA, CD36 and ADIPOQ.(C) The PPI of Module 2. It contains SHH, SIX1, ISL1 and NEUROG1 nodes.(D) The PPI of Module 3. It contains THBS4, ITGA8 and COL4A5 nodes.(E) The PPI of Module 4. It contains HBM, AHSP and HBD nodes.(F)The key genes of the Module pathway enrichment analysis results. 3.4. Identification and functional analysis of hub genes The cytoHubba plugin was used to score each node gene by six randomly selected algorithms, namely BottleNeck, Closeness, Degree, DMNC, EcCeniciy, and MCC. The first 20 hub genes in each algorithm were identified. Then, common hub genes in the six algorithms were selected as final hub genes by using the Venn diagram ([133]Fig. 5A and [134]Supplementary Table S3). Finally, we identified six hub genes as crucial targets for further analysis: ADIPOQ, IL6, PPARG, CEBPA, LPL, and LIPE ([135]Fig. 5B). Using the GeneMANIA platform, a hypothetical hub PPI network was built. In addition to the six crucial hub genes, 20 genes were closely related to the hub genes in Co-expression, Pathway, Physical Interactions, and Predictions: MAFF, FABP4, LEP, CEBPB, MGLL, EBF1, PIN1, GPD1, ELANE, CIDEC, PRKAR2B, NFKBIZ, STEAP4, FOXO1, CCND3, NR2F2, EHMT1, KLF3, NQO2, and MPI ([136]Fig. 5C). Enrichment pathways of the six hub genes and their associated genes are presented in [137]Fig. 5D. A significant enrichment of hub genes was observed in the transcriptional regulation of white lipid formation, the lipid metabolism pathway, the regulation of protein kinase activity, the myomechanical relaxation and contract pathways, and the electric transport chain ([138]Supplementary Table S4). Fig. 5. [139]Fig. 5 [140]Open in a new tab Determination and function enrichment analysis of hub gene. (A) Venn diagram of the intersection of the top 20 hub genes determined by six random algorithms, which include BottleNeck, Closeness, Degree, DMNC, EcCeniciy and MCC. (B) Six hub genes have been determined, ADIPOQ, IL6, PPARG, CEBPA, LPL and LIPE. (C)Interaction network of six hub genes and their closely related genes. (D)Functional enrichment analysis of six hub genes and their closely related genes. 3.5. Validation of candidate hub genes In the [141]GSE28660 test set, all six overlapping central genes in the recurrent POP samples were significantly upregulated and had a good diagnostic value ([142]Fig. 6). Among them, the area under the curve (AUC) value of ADIPOQ, IL6, PPARG, LPL, and LIPE was 100% and that of CEBPA was 96.9%. However, in the [143]GSE12852 and [144]GSE53868 validation sets of primary POP, IL6 expression was significantly upregulated in the pelvic floor tissue specimens of primary POP, and the expression of other hub genes exhibited no significant difference in the primary POP and normal control groups. Moreover, the ROC curve analysis displayed that IL6 (AUC = 0.806) and PPARG (AUC = 0.708) had a good prediction ability in the [145]GSE12852 training dataset. The validation results of the [146]GSE53868 dataset further substantiated that IL6 (AUC = 0.792) had excellent predictive power ([147]Supplementary Fig. S1). These results suggest that the six hub genes had a limited discrimination value for primary POP, but may have a specific predictive value for recurrent POP. Fig. 6. [148]Fig. 6 [149]Open in a new tab Validation and ROC curve analyses of six hub genes. (A) The expression of ADIPOQ, IL6, PPARG, CEBPA, LPL and LIPE in the dataset [150]GSE28660. (B) The ROC curve analyses of ADIPOQ, IL6, PPARG, CEBPA, LPL and LIPE in the dataset [151]GSE28660. (C) The expression of ADIPOQ, IL6, PPARG, CEBPA, LPL and LIPE in the dataset [152]GSE12852. (D) The ROC curve analyses of ADIPOQ, IL6, PPARG, CEBPA, LPL and LIPE in the dataset [153]GSE12852. 3.6. Target hub gene–miRNA/TF regulatory network To identify regulatory mechanisms at the transcriptional level and regulatory molecules of hub proteins, a network-based approach was used to decode regulatory TFs and miRNAs. The interaction of miRNA regulators with the six hub genes is displayed in [154]Fig. 7A. Using TarBase v8.0 of the NetworkAnalyst platform, 96 miRNAs related to hub genes were discovered, including hsa-mir-27a-3p, hsa-mir-124–3p, hsa-mir-182–5p, hsa-mir-210–3p, hsa-let-7b-5p, hsa-mir-1-3p, hsa-mir-155–5p, and hsa-mir-1343–3p ([155]Supplementary Table S5). The enrichment and analysis of miRNA–hub gene network nodes revealed that the hub genes were mainly involved in the PPAR signaling pathway, AMPK signaling pathway, NAFLD, transcriptional regulation in cancer, and longevity regulatory pathway after miRNA regulation ([156]Fig. 7B and [157]Table 3). In total, 97 TFs were collected from the hub gene–TF regulatory network ([158]Fig. 7C). TFs of CREM, GTF2A2, SUZ12, ZNF2, and ZNF512 in the network exhibited a high degree of interaction ([159]Supplementary Table S6). Furthermore, the enrichment results revealed that the nodes of the TF–hub gene network were mainly involved in transcriptional misregulation in cancer, herpes simplex infection, Huntington's disease, HTLV-I infection, and cell cycle regulation ([160]Fig. 7D and [161]Table 3). In addition, intersection processing was performed on the related signal pathways of miRNA–hub genes and TF–hub genes. This intersection processing revealed that six pathways were closely related to the transcriptional regulation of recurrent POP: PPAR signaling pathway, AMPK signaling pathway, NAFLD, transcriptive misregulation in cancer, longevity regulatory pathway, and pathways in cancer. Fig. 7. [162]Fig. 7 [163]Open in a new tab The miRNA-hub gene regulatory network and TF-hub gens regulatory network. (A)The miRNA-hub gene network. The dot represents hub genes, and the square represents miRNA.(B) Functional enrichment analysis of miRNA-hub gene network.(C)The TF-hub gene network. The dot represents hub genes, and the Diamond represents TFs.(D) Functional enrichment analysis of TF-hub gene network.(E) Venn diagram of the intersection of pathway enrichment analysis of miRNA hub gene network and TF hub gene network.(F) six intersection pathways of miRNA-hub gene network and TF-hub gene network. Table 3. Enrichment analysis items of miRNA-hub gene regulatory network and TF-hub gens regulatory network. Pathway Total Expected Hits P.Value FDR miRNA PPAR signaling pathway 74 0.06 3 0.00 0.01 AMPK signaling pathway 120 0.09 3 0.00 0.01 Non-alcoholic fatty liver disease (NAFLD) 149 0.12 3 0.00 0.01 Transcriptional misregulation in cancer 186 0.14 3 0.00 0.02 Longevity regulating pathway 89 0.07 2 0.00 0.12 Pathways in cancer 530 0.41 3 0.01 0.29 Antifolate resistance 31 0.02 1 0.02 0.71 Prion diseases 35 0.03 1 0.03 0.71 African trypanosomiasis 37 0.03 1 0.03 0.71 Thyroid cancer 37 0.03 1 0.03 0.71 Graft-versus-host disease 41 0.03 1 0.03 0.71 Type II diabetes mellitus 46 0.04 1 0.04 0.71 Intestinal immune network for IgA production 49 0.04 1 0.04 0.71 Malaria 49 0.04 1 0.04 0.71 Regulation of lipolysis in adipocytes 55 0.04 1 0.04 0.71 Legionellosis 55 0.04 1 0.04 0.71 Glycerolipid metabolism 61 0.05 1 0.05 0.71 Cytosolic DNA-sensing pathway 63 0.05 1 0.05 0.71 Inflammatory bowel disease (IBD) 65 0.05 1 0.05 0.71 TF Transcriptional misregulation in cancer 186 1.20 10 0.00 0.00 Herpes simplex infection 492 3.18 14 0.00 0.00 Huntington's disease 193 1.25 8 0.00 0.00 HTLV-I infection 219 1.42 7 0.00 0.04 Cell cycle 124 0.80 5 0.00 0.07 PPAR signaling pathway 74 0.48 4 0.00 0.07 Viral carcinogenesis 201 1.30 6 0.00 0.08 Longevity regulating pathway 89 0.58 4 0.00 0.10 Hepatitis B 163 1.05 5 0.00 0.14 Insulin resistance 108 0.70 4 0.01 0.16 TNF signaling pathway 110 0.71 4 0.01 0.16 AMPK signaling pathway 120 0.78 4 0.01 0.19 Non-alcoholic fatty liver disease (NAFLD) 149 0.96 4 0.02 0.38 RNA polymerase 31 0.20 2 0.02 0.38 Pathways in cancer 530 3.43 8 0.02 0.41 Prion diseases 35 0.23 2 0.02 0.42 Thyroid cancer 37 0.24 2 0.02 0.43 Prostate cancer 97 0.63 3 0.02 0.43 Tuberculosis 179 1.16 4 0.03 0.45 Alcoholism 180 1.16 4 0.03 0.45 Th17 cell differentiation 107 0.69 3 0.03 0.47 Vasopressin-regulated water reabsorption 44 0.28 2 0.03 0.47 Cocaine addiction 49 0.32 2 0.04 0.55 [164]Open in a new tab 3.7. Identification of disease association The DisGeNET database in the NetworkAnalyst platform was used to find hub gene-related diseases or chronic complications. A total of 192 diseases or chronic complications were identified from the hub gene–gene–disease association network ([165]Fig. 8). Insulin resistance, atherosclerosis, Crohn's disease, diabetes mellitus, non-insulin-dependent, heart failure, hypertensive disease, inflammation, and obesity exhibited the highest degree of interaction and coordination with hub genes ([166]Supplementary Table S7). The findings implied that the six predicted hub genes might be involved in various pathological processes of metabolic diseases, especially obesity. Fig. 8. [167]Fig. 8 [168]Open in a new tab The hub genes-diseases associations network constructed with six hub genes and 192 diseases or chronic complications.(A)The hub genes-disease associations network. The dot represents hub genes, and the diamond represents diseases. (B)The top 15 diseases associated with hub genes. 3.8. Identification of candidate drugs or compounds In order to investigate the interaction between the six predicted hub genes and available therapeutic drugs or compounds for recurrent POP, the hub gene–drug or compound interaction network was constructed using DrugBank and CTD based on the NetworkAnalyst platform. AST-120 and Tyloxapol can be used as crucial target drugs for regulating LPL expression ([169]Fig. 9 and [170]Supplementary Table S8). Other hub genes, including ADIPOQ, IL6, PPARG, CEBPA, and LIPE, did not have any directly related target drugs. Various compounds could affect the expression of these six predicted hub genes. There were interactions between dexamethasone and all of the hub genes, as well as interactions between 1-methyl-3-isobutylxanthine, bisphenol A, efavirenz, estradiol, glucose, hesperetin, nevirapine, sodium arsenite, troglitazone, valproic acid, rosiglitazone, niacin, and bis (4-hydroxyphenyl)sulfone and five hub genes. These potential drugs or chemical compounds may have a therapeutic effect on recurrent operations by acting on the hub genes. [171]Table 4 presents 10 potential compounds showing the largest correlation with the hub genes from the DSigDB database. Fig. 9. [172]Fig. 9 [173]Open in a new tab The gene-drug or compounds interaction network constructed with six hub genes and 954 drugs or compounds. The dot represents hub genes, and the pentagon represents drugs or compounds. Table 4. The top 10 drugs or compounds most closely related to hub genes. Id Label Degree Betweenness Chemical Formula Molecular Structure D003907 Dexamethasone 6 2626.14 C22H29FO5 Image 1 D015056 1-Methyl-3-isobutylxanthine 5 2232.3 C10H14N4O2 Image 2 C006780 bisphenol A 5 2232.3 C15H16O2 Image 3 C098320 efavirenz 5 2232.3 C14H9ClF3NO2 Image 4 D004958 Estradiol 5 2232.3 C18H24O2 Image 5 D005947 Glucose 5 2232.3 C6H12O6 Image 6 C013015 hesperetin 5 2232.3 C16H14O6 Image 7 D019829 Nevirapine 5 2232.3 C15H14N4O Image 8 C017947 sodium arsenite 5 2232.3 C2H3NaO2 Image 9 C057693 troglitazone 5 2232.3 C24H27NO5S Image 10 [174]Open in a new tab 4. Discussion Recurrent POP has recently attracted increasing attention, although unified diagnostic criteria, standard evaluation tools, outcome measures, a cut-off level of physical examination, determined follow-up time, and a treatment plan are lacking [[175]11,[176]12]. The currently accepted concept of recurrent POP refers to objective recurrent disease, patients with POP directly or indirectly (POP-Q ≥ stage 2 b) at or below the level of the hymen and recurrent POP symptoms are due to subjective recurrence [[177]11]. Based on this diagnostic criterion, studies have reported a POP recurrence rate of 30%–36%, especially in young patients with anterior vaginal wall prolapse, the recurrence rate is higher [[178]11,[179]36,[180]37]. At present, knowledge about the natural history of recurrent POP is limited. Academics often consider the incidence of recurrent reoperation, which is relatively easy to determine, as a substitute indicator of POP recurrence [[181]12,[182]13]. According to two recent meta-analyses, the risk factors for POP reoperation are pre-preoperative prostate stage, levator ani avulsion, hiatal area, younger age, and family history [[183]14,[184]15]. In particular, family history suggests that genetics are closely linked to the recurrence of POP [[185]16]. Developing preventive and predictive measures for other POP reoperation-related factors, such as preoperative prolapse, hiatal area, and younger age, is difficult [[186][38], [187][39], [188][40]]. Therefore, the mechanism and function of genetic factors, including core genes and key pathways, in POP recurrence must be urgently determined. Here we used bioinformatics methods to integrate and analyze 110 DEGs with potential molecular diagnostic and therapeutic value in the recurrent POP and control groups. These 110 DEGs included 14 downregulated and 94 upregulated genes. GO enrichment revealed that these genes were mainly related to lipid metabolism, energy metabolism, oxidative stress, and hormone regulation. This result is different from the results of the expression profile analysis and bioinformatics analysis of primary POP. The GO function enrichment revealed that the primary POP-related genes were mainly involved in tissue repair, mineral metabolism, immune response, complement activation, cell apoptosis, etc. [[189]41,[190]42]. In addition, the pathway enrichment analysis revealed that the PPAR signaling pathway, AMPK signaling pathway, adipocytokine signaling pathway, NAFLD, and glycerol metabolism may be related to recurrent POP. These pathways are closely related to tissue repair, energy metabolism, and lipid metabolism and have been reported in previous studies on primary POP. For example, through bioinformatics, Zhou et al. found that the MAPK signaling pathway is involved in primary POP development [[191]41]. Yu et al. discovered that the metabolites of primary POP are mainly enriched in glycophorophospholipid; nicotine and nicotine; glycine, serine, and threonine metabolism; arginine and proline metabolism; and pyridine and purine metabolism [[192]43]. To further determine the biological functions of DEGs, the PPI network and four functional DEG modules were constructed and analyzed. Three pathway enrichment methods, namely KEGG, REACTOME, and WIKIPATHWAYS enrichment analyses, revealed that the PPAR signaling pathway, AMP signaling pathway, and adipocyte metabolism and differentiation were closely related to POP recurrence-related modules. Notably, on the basis of the histological quantitative system of the USL of women with POP, Orlicky et al. found that adipose accumulated abnormally in the USL of these patients and that this abnormal adipose distribution was significantly associated with vaginal delivery [[193]44]. Based on these findings, we hypothesized that the adipose composition and function of pelvic floor support structures may be key risk factor for primary and recurrent POPs development. This is also an important aspect of our recent research. We further determined the crucial hub genes (ADIPOQ, IL6, PPARG, CEBPA, LPL, and LIPE) of recurrent POP using Cytoscape software and six randomly selected algorithms. These six key genes were verified using two primary POP datasets. Significant differential expression of IL6 was observed, with better prediction abilities for primary POP identification in IL6 (AUC = 0.792 or 0.806) and PPARG (AUC = 0.708). At present, data on recurrent POP are relatively lacking, and the expression and function of the six genes are worthy of further research and verification in recurrent POP. On the whole, these six genes were mainly involved in white lipid formation, lipid metabolism pathway, regulation of protein kinase activity, myomechanical relaxation and contract pathways, and the electric transport chain. This result is similar to the DEG enrichment results of recurrent POP. However, in studies related to primary POPs [[194]45,[195]46], there are concerns about mechanical releases and contract pathways, as well as electronic transport chains. With regard to the single gene IL6, in addition to its biological role in inflammation, immune response, cytokine production, and tissue repair, several studies have confirmed that IL6 is involved in the progress of POP [[196]41,[197][47], [198][48], [199][49]]. Evidence on the relationship between other genes and POP is limited. These five remaining hub genes with research prospects deserve further exploration, especially the regulatory effects of ADIPOQ on hormone progress [[200]50], PPARG on adipocyte differentiation [[201]51], CEBPA on body weight homeostasis [[202]52], and LPL and LIPE on triglyceride metabolism need to be determined [[203]53]. Here, we constructed and analyzed the miRNA regulatory network, TF regulatory network, disease network, and drug/chemical network of hub genes and further explored the regulatory mechanism and potential therapeutic value of hub genes. In the miRNA–hub gene network, 96 miRNAs were found to be associated with hub genes. Previous basic research has confirmed that among these miRNAs, miRNA-19, miRNA-30, miRNA-181, miRNA-138, and miRNA-221/222 could regulate the expression of critical genes supporting the pelvic floor, such as IGF-1, HOXA11, FBLN5, and estrogen receptor [[204][54], [205][55], [206][56], [207][57]]. In the TF–hub gene network, TFs of CREM, GTF2A2, SUZ12, ZNF2, and ZNF512 in the network exhibited a high degree of interaction. Evidence showed that these TFs are associated with POP is lacking; however, the role of the TF–hub gene network in cell cycle regulation in prolapse needs to be investigated [[208]58]. In the hub–disease network, 192 diseases or chronic complications were found to be possibly associated with POP recurrence. In particular, diabetes, epidemiological studies have confirmed that hypertension, obesity, inflammatory diseases, and chronic obstructive pulmonary disorder (COPD) are related to POP. However, whether a relationship exists with recurrent POP needs to be confirmed using a larger sample size and in a more carefully designed clinical study. In a retrospective analysis of 353 women who received apical prolapse repair through a standard surgery, obesity was not a risk factor for POP recurrence after surgery, but an effective hierarchical analysis was required to determine the prolapse site, prolapse degree, follow-up time, surgical methods, and operators [[209]59,[210]60]. Referring to DrugBank and CTD, 954 target therapeutic drugs or compounds for the six predicted hub genes were identified. Further drug relocation, molecular docking, and basic test validation analysis are required for important targets such as dexamethasone, 1-methyl-3-isobutylxanthine, bisphenol A, efavirenz, estradiol, glucose, hesperetin, nevirapine, sodium arsenite, troglitazone, valproic acid, rosiglitazone, niacin, and bis(4-hydroxyphenyl) sulfone. This is also the focus of our future research. To our knowledge, this is the first in-depth bioinformatics analysis of recurrent POP. The present study has several limitations. First, owing to the difficulty involved in sample collection, this study does not involve pure bioinformatics analysis and lacks experimental verification. Second, the primary POP dataset used in the validation set weakens the persuasiveness. Third, for recurrent POP, related miRNAs, TFs, diseases, and compounds/drugs are predicted using preliminary algorithms, which need to be verified through subsequent strict clinical and basic research. Nevertheless, using limited data information and scientific methods, our research findings possibly help to predict the potential key pathways, crucial genes, concomitant diseases, and therapeutic targets associated with recurrent POP. 5. Conclusions Based on the integrated bioinformatic analysis, this study identified 110 DEGs and 6 hub genes (ADIPOQ, IL6, PPARG, CEBPA, LPL, and LIPE) associated with recurrent POP development and prognosis. The enrichment analysis revealed that the PPAR signaling pathway, AMPK signaling pathway, adipocytokine signaling pathway, NAFLD, and glycerol metabolism may be closely related to POP recurrence. The miRNA–hub gene network and TF–hub gene network may be involved in the progress of recurrent POP. Moreover, patients having prolapse along with diabetes, hypertension, obesity, inflammatory diseases, and COPD need to be observant of the risk of recurrence, and molecular drugs and molecular prediction models need to be further explored. Funding This work was supported by the Hubei Provincial Health Commission Joint Fund Project (WJ2019H500), Hubei Natural Science Foundation (2020CFB643) and Yichang Medical and Health Research Project (A22-2-038). Ethics approval and consent to participate Not necessary. Consent for publication Not applicable. Author contribution statement Quan Zhou: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper. Man Lu: Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data. AbrGuo-Sheng Li: Contributed reagents, materials, analysis tools or data. Gan-Lu Peng: Contributed reagents, materials, analysis tools or data; Wrote the paper. Yan-Feng Song: Conceived and designed the experiments; Analyzed and interpreted the data. Data availability statement Data associated with this study has been deposited at The data of this study are derived from the GEO database ([211]https://www.ncbi.nlm.nih.gov/geo/). Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements