Mining of Subtype Markers for the Prognosis of Ovarian Cancer based on Methylation Data

Mining of Subtype Markers for the Prognosis of Ovarian Cancer based on Methylation Data | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Primary research Mining of Subtype Markers for the Prognosis of Ovarian Cancer based on Methylation Data Lili Yin, Ningning Zhang, Qing Yang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.2.23311/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract Aims: Ovarian cancer is one of three major malignancies involving the female reproductive system, and its morbidity and mortality are ranked number 3 and number 1 among gynecological tumors, respectively. DNA methylation (MET), as one of the main epigenetic modes, is closely related to the occurrence and development of ovarian cancer. To guide individualized treatment and improve the prognosis in ovarian cancer patients, it is of great significance to elucidate effective MET subtype markers. Methods: A total of 571 ovarian cancer MET samples were downloaded from the Cancer Genome Atlas (TCGA), and a COX proportional hazards model was established using the MET spectrum and clinically pathological parameters. Subsequently, the consensus clustering of CpG loci with a significant difference in both univariate and multivariate analyses was performed to screen the molecular subtypes, and these CpG loci were subjected to gene function annotation. Finally, CpG MET loci associated with poor prognosis in ovarian cancer patients were further screened by constructing a weighted gene co-expression network analysis (WGCNA). Results: A total of 250 prognosis-related MET loci were obtained by COX regression and 6 molecular subtypes were screened by clustering. There was a remarkable MET difference between most subtypes, of which Cluster 2 had the highest MET level and demonstrated the best prognosis in patients, while Cluster 4 and Cluster 5 had a MET level significantly lower than that of the other subtypes and demonstrated a very poor prognosis. All Cluster 5 samples were at a high grade, while the percentage of Stage IV samples in Cluster 4 was evidently greater than that in the other subtypes. Using the co-expression network, 5 CpG loci were eventually obtained: cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. The clustering analysis shows that the prognosis in patients with hypomethylation was significantly worse than that in patients with hypermethylation. Conclusions: These MET molecular subtypes can be used not only to evaluate the prognosis in ovarian cancer patients but also to fully distinguish the tumor stage and histological grade in these patients. Prognosis-related CpG loci can be applied as biomarkers for individualized treatment in ovarian cancer patients. Cancer Biology Oncology Ovarian cancer TCGA methylation molecular subtype prognosis marker Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Background Ovarian cancer is a disease of high heterogeneity with varying molecular phenotypes, pathogeneses, and prognoses, and its morbidity is ranked number 3 among malignant gynecological tumors. However, ovarian cancer is not easily detected at an early stage since the ovaries are located deep in the pelvis, and at diagnosis, the tumor is advanced with distant metastases in 70% of cases. Most patients will experience recurrence within 2 years, and there is a lack of effective therapies for recurrent ovarian cancer, which ranks the mortality of ovarian cancer as number 1 among gynecological tumors. Therefore, precision therapy is an urgent demand. Epithelial ovarian cancer is the most common pathological type and accounts for 80−95% of ovarian malignancies. It is classified into 5 histological subtypes: high-grade serous adenocarcinoma, endometrioid adenocarcinoma, clear cell adenocarcinoma, mucous adenocarcinoma, and low-grade serous adenocarcinoma 1-5 . In a large, prospective, phase-III clinical study, Kommoss et al., showed that histological grading is necessary in early ovarian cancer but has no significance in prognostic evaluation in advanced ovarian cancer 6 . By combining the clinical pathological and molecular biological characteristics of ovarian cancer, Shih IeM et al., classified epithelial ovarian cancer into types I and II. Type I epithelial ovarian cancer mainly includes low-grade serous carcinoma and low-grade endometrioid carcinoma, and most cases have early onset and good prognosis. Type II epithelial ovarian cancer mainly includes high-grade serous carcinoma, high-grade endometrioid carcinoma, and undifferentiated carcinoma, and most cases have rapid onset and poor prognosis. Type I epithelial ovarian cancer is significantly correlated with mutations in BRAF, KRAS, and PTEN, while type II epithelial ovarian cancer is associated with p53 mutations and also very frequently with BRCA1/2 mutations. The dualistic theory reflects different biological behaviors and clinical prognoses of tumors, and such differences are especially remarkable between low-grade and high-grade serous ovarian carcinomas. However, application of the dualistic theory in non-serous ovarian carcinoma is limited; for instance, clear cell carcinoma has many biological behaviors similar to type II epithelial ovarian cancer, despite belonging to type I epithelial ovarian cancer. It is of profound significance to realize precision molecular typing of ovarian cancer for clinical treatment and prognosis monitoring 7 . As a result of the continuous improvement in human genome sequencing technology and the advancement of biomedical analysis technology, a new trend has been formed that involves molecular targeted therapy and prognosis evaluation based on the molecular typing of malignant tumors. Molecular-targeted therapy has been successfully applied in several tumors, including ER (+) or HER2(+) breast cancer and EGFR-mutated lung cancer 8-11 , presenting great progress in precision medical treatment. By K-means clustering, Tothill et al. 12 detected the gene expression spectrum of 285 cases of endometrioid and serous tumors originating from the ovary, peritoneum, and tuba uterine, finally identified 6 molecular subtypes, of which 4 (high interstitium-reactive, high immunity, hypomethylation trix-reactive, and interstitial low immunity) are features of high-grade serous ovarian carcinoma and can be used for predicting prognosis in these patients. Based on gene expression, TCGA and Tothill et al., divided high-grade serous carcinoma into 4 subtypes: immunoreactive, differentiated, proliferative, and interstitial 13 . Moreover, the clinical trial completed by Kommoss et al. showed that in high-grade serous ovarian carcinoma patients with proliferative and interstitial molecular subtypes, bevacizumab can improve the progression-free survival to different degrees; however, this study was limited to the molecular typing of the gene expression spectrum 14 . Despite the impact of gene changes on cancer occurrence, epigenetic changes such as DNA MET also play an important role. Epigenetic inheritance refers to the hereditary changes that occur under the precondition of no changes in DNA sequence, such as histone modification, DNA MET, RNA editing, and gene silencing. The occurrence and growth of ovarian cancer involve several pathways including DNA repair, cell apoptosis, cell cycle regulation, and changes in protooncogenes tumor suppressor genes. Studies have suggested that epigenetic changes in these pathways play essential roles in the development of ovarian cancer, and the detection of MET signals is helpful for early diagnosis 15,16 . DNA MET mainly occurs in CpG islands; CpG expression can be inhibited by the hypermethylation of tumor suppressor gene promoters and enhanced by the decreased demethylation probability of protooncogenes. Different regulatory effects of protooncogenes and tumor suppressor genes contribute to the occurrence of cancer 17-20 . According to relevant studies, the tumor suppressor gene involved in ovarian cancer exists in a hypermethylated state, and changes in its MET level is an important molecular foundation for cancer occurrence 21,22 . Hu WL et al., 23 built a DNA MET interaction network for ovarian cancer, breast cancer, and glioma, confirming that the number of DNA MET loci was associated with prognosis in cancer patients; however, no DNA MET molecular typing of ovarian cancer was performed. In the present study, the univariate and multivariate COX proportional hazards models were established by analyzing Illumina Infinium® Human Methylation 27K data in the TCGA database and combining the MET level and clinical data of the samples. Subsequently, 6 molecular subtypes associated with the prognosis in ovarian cancer patients were screened by consensus clustering of MET spectra with a significant difference in both models, and the ovarian cancer patients were classified using these subtypes. Further, 5 CpG hypomethylation loci related to poor prognosis in ovarian cancer patients were obtained by constructing a WGCNA co-expression network. These loci are of great significance for the clarification of the pathogenesis of ovarian cancer, and can be used as effective tumor markers to provide a reference for clinical prognosis and individualized treatment. Materials And Methods Preprocessing of ovarian cancer expression datasets and preliminary screening of DNA MET loci TCGA 24 GDC API was utilized to download the latest clinical follow-up information and RNA-Seq data. Illumina Infinium® Human Methylation 27 BeadChip ( Illumina 27K ) microarray was acquired from UCSC Cancer Browser. Samples with complete clinical data and methylation spectrum data were selected. CpG loci with NA (Not Available) > 70% in all samples were deleted. The impute-KNN of R package was used to fill the missing value of methylation spectrum. The unstable genomic methylation loci were further removed, involving CpGs and single nucleotide loci on sex chromosomes as well as CpG loci that were not annotated to the gene promoter region 25 . We divided the datasets into two queues: a training set coupled with a test set. The standards for the subgroups included: 1) Samples were assigned to the training set and the test set randomly; and 2) the data of the two groups should be similar, including age distribution, clinical stage, follow-up time, and mortality ratio. Univariate survival analysis of MET loci in the training set The research objective was to determine the molecular subtypes of ovarian carcinoma as prognostic determinants. Therefore, CpG loci, which had an important impact on survival, were utilized as a classification feature. First and foremost, a univariate COX proportional hazard model was established based on the methylation level of each CpG loci, age, tumor grade and stage, coupled with survival data by coxph function of the R-package survival . Subsequently, we introduced the significant CpG loci obtained from the univariate model into the multivariate COX proportional hazards model, and took the significant age and clinical attributes in the univariate model as covariables. Ultimately, the CpG loci, which were still significant, were employed as classification features [25]. For each CpG island, the multivariate COX proportional hazards model formula was described below: h( t , x ) i = h 0( t )exp (β methy methyi + β age age + β stage stage) (1) In the formula, “methy i ” is the carrier of the CpG locus methylation level in the sample. “Age” and “stage” describe the age and clinical characteristics of the patients, respectively. “β methy ”, “β age ”, and “β stage ” are regression coefficients. The P -values of the COX regression coefficient was adjusted by Benjamini−Hochberg error detection rate. Various comparing processes were carried out. Screening of molecular subtypes by the consensus clustering of methylation profile with a significant difference in both univariate and multivariate analyses Consensus ClusterPlus in the R package 26 was utilized for consensus clustering according to the method described by Zhang et al. 25 The subgroups of epithelial ovarian tumors were identified based on the most variable CpG loci. The algorithm is described as follows. First, double sampling of some items and features from the data matrix was conducted, in which each sub-sample was divided into several groups (max.) using a user-specific clustering algorithm ( k-means , hierarchical clustering, or custom algorithms). The paired consensus value (defined as the proportion of clustering running for the combination of two items) was calculated and stored in the k i consensus matrix. Second, the final coherent sheaf clustering for each k i was completed using the distance of 1- consensus value and pruned into k i group through cutting, which is known as consensus clustering. The algorithm determined the “consensus” clustering by measuring the stability of the clustering results applied to random data subsets from given clustering methods. In each iteration, 80% of the tumors were sampled, and a k-means algorithm with the Euclidean squared distance measures were utilized: [Please see the supplementary files section to access this equation.] (2) There was k = 2−10 groups, and these results were compiled for 100 times. The cluster consensus as well as item consensus results were obtained with Consensus ClusterPlus. R package. The graphical output results included the heat map of consensus matrix, cumulative distribution function (CDF) diagram, and Δ region diagram. The criteria of clustering number included relatively high consistency within the cluster, relatively low coefficient of variation, and insignificant increase of the area under the CDF curve (AUC). The CV (%) was calculated based on the formula below: CV (%) = (SD/MN) x 100 (3) Where SD is the standard deviation, while MN is the average value of the samples. We selected category number as the area under the CDF curve, and there was no significant change. The consensus clustering heatmap was generated using the R package pheatmap . Clustering analysis of the methylation expression profile and analysis of the clinical characteristics of screened molecular subtypes The stable clustering results were selected and the methylation profile was analyzed by clustering analysis. The distance between the MET loci was calculated using the Euclidean distance. Furthermore, the distribution of various molecular subtype samples was analyzed with respect to prognosis, stage, grade, and age. Gene annotation of MET loci As for the genes corresponding to the gene promoter regions annotated by the selected CpG loci, the transcription factor enrichment analysis was performed by the online tool g:profiler 27 . WGCNA co-expression analysis of CpG loci Based on the modification beta value of selected CpG loci, the co-expressed CpG loci were mined by WGCNA co-expression algorithm. The distance between CpG loci was calculated using Pearson Correlation Coefficient. The R package WGCNA was used to construct weighted co-expression network, and select a soft threshold of 4 to filter the CpG co-expression modules. The results showed that the co-expression network conformed to the scale-free network. That is to say, the log(k) of node k presented in the connection is negatively correlated with the log(P(k)) of the probability of node k, with a correlation coefficient larger than 0.8. In order to ensure a scale-free network, we chose β = 4. The next step was to convert the expression matrix into an adjacency matrix, and then transform the adjacency matrix into a topological matrix. Based on TOM, we utilized average-linkage hierarchical clustering method to cluster genes. The minimum number of genes in each IncRNA network module was set at 30 according to the standard merged dynamic tree cutting. After determining the gene modules with the dynamic cutting method, we calculated eigengenes of each module in turn. The modules were clustered, and the adjacent modules were merged into new modules. Construction of prognosis models and data validation of independent test set Unsupervised clustering analysis was conducted on the CpG methylation profile selected in the previous step. The similarity between samples was calculated by using the Euclidean distance. The samples were then divided into two groups according to the methylation level of CpG loci. The prognosis differences between the two groups were further analyzed. The methylation profile of 286 samples in the test set were used for validation. Results Selection of 250 characteristic MET loci The Illumina Infinium® Human Methylation 27 BeadChip microarray contained 613 samples in total, with 571 samples being screened using MET detection. The missing data imputation of the MET spectrum was performed using Impute in the R software package, and 25,154 MET loci were selected following exclusion of the unstable genomic MET loci. All 571 samples were assigned to either a training set (n = 285) or a validation set (n = 286). The clinical pathological information in the training set and validation test is shown in Table 1 . The MET loci and survival data were analyzed using a univariate COX proportional hazards regression model with p < 0.05 as the threshold. A total of 967 loci demonstrated a significant difference in prognosis ( Table S1 ), of which the top 20 loci with the most significant difference are shown in Table 2 . The prognostic significance of age had a logrank P value of 5.93e-06, while that of stage was 0.0379. The significant MET loci were selected using a univariate COX model followed by multivariate COX proportional hazards regression model analysis, with stage and age as covariates. Finally 250 significant MET loci were obtained ( Table S2 ). Screening 6 molecular subtypes by consensus clustering of the MET loci Consensus clustering of MET loci with a significant difference in both univariate and multivariate analyses was performed using Consensus ClusterPlus in the R package to screen the molecular subtypes. The similarity between samples was calculated using the Euclidean distance, the clustering was performed with k-means, and 80% sampling was conducted 100 times using a double-sampling method. The optimal cluster number was determined by CDF; different colors in the CDF curve represent different cluster numbers ( Figure 1A ), the AUC was larger at 6 and 7 clusters and the clustering effect was better. Further observation of the CDF delta area curve ( Figure 1B ) shows that at 6 clusters, the AUC demonstrates stable clustering results. Following consideration, k = 6 was selected and 6 molecular subtypes were obtained. Clustering analysis of the MET expression spectra of 6 molecular subtypes The composition and number of samples in the 6 clusters were evaluated using the consensus matrix. The color gradient was from white to blue, indicating the consensus of progression. In the matrix permutation, the same clusters were made mutually adjacent; and eventually, a color-coded heat map was displayed and featured by the arrangement of dark blue blocks on a diagonal white background ( Figure 2A ), showing that 285 tumor samples were assigned to these 6 clusters. Furthermore, clustering analysis was performed on 250 MET spectra; the distance between MET loci was calculated using the Euclidean distance, and the heat map was generated by pheatmap using clinical pathological stage and histological type as notes ( Figure 2B ). The pairwise comparison of various subtypes was performed using a t-test, and the results revealed that most MET loci had a low beta value. There was a significant difference in MET level among the majority of subtypes; the MET level in Cluster 2 was remarkably higher than that in the other 5 subtypes, while the MET levels in Cluster 4 and Cluster 5 were evidently lower than those in the other subtypes ( Table S3 ). Analysis of the clinical characteristics of the 6 molecular subtypes We further analyzed the distribution of the 6 molecular subtypes with respect to prognosis, stage, grade, and age ( Figure 3) . There was a significant difference in the prognosis among the 6 subtypes; the prognosis was best in Cluster 2, worst in Cluster 5, and poor in Cluster 4 ( Figure 3A ), indicating that the prognosis of hypomethylation subtypes was inferior to that of hypermethylation subtypes. The samples in Cluster 5 were all stage III, and the percentage of stage IV samples in Cluster 4 was significantly higher than that in the other subtypes ( Figure 3B ). All samples in Cluster 5 were G3 ( Figure 3C , Table S4) , suggesting that the hypomethylation subtypes were mostly high-grade in clinical pathology. The age of the patients of samples in Cluster 5 was remarkably greater than that of the patients of the other subtypes, and the onset age was 70−80 years old ( Table S5 ), while the mean age of the patients of samples in Cluster 2 was the lowest ( Figure 3D ), indicating that the age of patients with hypomethylation subtypes was generally higher than that of patients with hypermethylation subtypes. The above findings suggest, to a certain degree, that these DNA MET subtypes could be used to predict the prognosis, tumor stage, and pathological grade in ovarian cancer patients. Gene annotation and function analysis of the 250 MET loci There was a total of 285 genes corresponding to the gene promoter regions annotated by 250 CpG loci, and these genes were subjected to transcription factor enrichment analysis using the online tool g:profiler. It was found that 42 genes were significantly enriched to Transcription Factor EC (TFEC) (logrank P value = 0.0107) (Figure 4A). At present, the role of TFEC in cancer progression has been studied to a limited extent; thus, to further explore the biological functions in which TFEC may be involved, TFEC-co-expressed molecules in the cBioPortal database were elucidated. Subsequently, the top 300 molecules with the most positive and negative correlations according to Spearman's correlation were selected. Functional enrichment analysis was performed using the DAVID 6.7 database and visualized using “GOplot” in the R software package. Finally, the top 5 biological processes with the most significant functions were chosen: GO:0006955 − immune response, GO:0050776 − regulation of immune response, GO:0006954 − inflammatory response, GO:0045087 − innate immune response, and GO:0007165 − signal transduction. TFEC may promote the occurrence and progression of ovarian cancer by influencing these biological functions(Figure 4B). Screening 5 CpG loci by WGCNA co-expression analysis Using the WGCNA co-expression algorithm, 250 significant CpG loci were mined. Evaluation of the scale-free model was performed at different soft thresholds; a larger value and lower mean connectivity both indicated better compliance with the scale-free distribution. Finally, β = 4 ( Figure 5A, B ) was selected, and the settings of height = 0.25, deepSplit = 3, minModuleSize = 10 were chosen. A total of 7 modules were obtained ( Figure 5C ), of which the grey module is the set of genes that could not be clustered to other modules. The statistics of genes in various modules are shown in Table 3 . The 250 CpG loci were assigned to 7 modules. Pearson’s correlation coefficient between the ME of each module and the characteristics of the samples was calculated; a higher correlation coefficient indicates that the module was more important. In Figure 5D , the row represents the eigengenes of each module and the column represents the feature information of the samples. The greatest correlation can be seen between the yellow module and Cluster 2 (R = 0.68, logrank P = 7e-40), the brown module and Cluster 3 (R = 0.51, logrank P = 1e-20), and the black module and Cluster 5 (R = 0.61, logrank P = 6e-30). Since Cluster 2 demonstrated the best prognosis among all the clusters, all CpG loci in the yellow module, mostly correlated with Cluster 2, were selected, and the interaction network was constructed according to their weighted relationship ( Figure 6A ). In this network, the CpG loci with a network centrality > 10 were cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. Furthermore, the expression relationships among the 22 CpG loci were calculated, and a significantly higher correlation was found among 8 (cg27625732, cg00431050, cg22197830, cg03152385, cg22809047, cg00328227, cg06851207, and cg01777397) ( Figure 6B ). Finally, a total of 5 CpG loci in the intersection were chosen, which had a strong correlation between each other and a centrality > 10 in the weighted network, as the characteristic MET loci of Cluster 2 samples ( Table 4 ). Clustering analysis of the 5 CpG loci Unsupervised clustering analysis was further performed on the MET spectra of the 5 selected CpG loci, and the similarity between samples was calculated using the Euclidean distance. Figure 7A shows that the samples were divided into two groups according to the MET level of the 5 CpG loci: Cluster 1 (hypomethylation group) and Cluster 2 (hypermethylation group). The prognosis difference between these two groups was further analyzed ( Figure 7B ), indicating that the prognosis in the hypomethylation group was significantly poorer than that in the hypermethylation group. Model validation using the test dataset The MET spectra of the 5 CpG loci from the 286 samples in the test dataset were extracted and analyzed by hierarchical clustering ( Figure 8B ). The results show that the MET spectra of the 5 CpG loci were also obviously clustered into two groups: Cluster 1 and Cluster 2. The MET level of Cluster 1 samples was significantly higher than that of Cluster 2 samples. The prognosis difference between Cluster 1 and Cluster 2 ( Figure 8B) was further analyzed, and it was found that the prognosis in the hypermethylation group was remarkably better than that in the hypomethylation group, which is consistent with the results of the training set. Flow chart of all the analysis The flow chart of mining of subtype markers for the prognosis of ovarian cancer based on methylation data is shown in Figure 9 . All the R package covered in this article is listed in Table S6 . Discussion In recent years, an increasing number of studies have focused on exploring the molecular typing of epithelial ovarian cancer to promote the realization of personalized treatment and improve the survival rate in patients; however, the achievements of molecular typing remain in the initial phase. Studies have shown that the occurrence of cancer is associated with genetic changes, and epigenetic abnormalities are also contributors. DNA MET is the major epigenetic modification mode of genomic DNA; it is an important means in regulating the functions of the genome 28 and is closely associated with the occurrence, progression, treatment, and prognosis of ovarian cancer. DNA MET-based molecular typing and subtype markers are of great significance for guiding personalized treatment and prognosis evaluation in ovarian cancer patients. In the present study, 571 ovarian cancer MET samples were downloaded from the TCGA database, 250 MET loci related to the prognosis of ovarian cancer patients were screened by COX regression analysis, and 6 molecular subtypes were selected by clustering with k-means. There was a significant difference in MET loci among most subtypes; the highest MET level and the best prognosis were observed in Cluster 2, and the MET level in Cluster 4 and Cluster 5 was remarkably lower than that in the other subtypes, accompanied by a very poor prognosis. This suggests, to a certain degree, that the prognosis of patients with a hypomethylation subtype was worse than that of patients with a hypermethylation subtype. All samples in Cluster 5 were high-grade, and the mean age of patients in Cluster 5 was higher than that in the other subtypes. The percentage of stage IV samples in Cluster 4 was significantly greater than that in the other subtypes. The above findings suggest that these molecular subtypes can be used not only to evaluate the prognosis in ovarian cancer patients, but also to fully distinguish the tumor stage, histological grade, and age of these patients to guide subsequent treatment. DNA MET molecular typing also plays a very important role in the diagnosis, treatment, and prognosis of other tumors. Zhang et al. 25 screened 9 molecular subtypes by clustering analysis on DNA MET data in 669 breast cancer patients, and the DNA MET mode was reflected in varying races, ages, tumor stages, subject states, histological types, metastatic states, and prognoses. In comparison with PAM50 subtypes using gene expression clustering, DNA MET subtypes are more precise and can be used for the precision treatment of specific histological subtypes of breast cancer. Jurmeister constructed a DNA MET map using the whole genome MET data from 600 cases of primary pulmonary, colorectal, and upper gastrointestinal adenocarcinoma, and successfully distinguished between pulmonary enteric adenocarcinoma and metastatic colorectal cancer 29 . Williams et al. 30 measured the MET level in different histological subtypes of 154 cases of child germ cell tumors using the Illumina Infinium® Human Methylation 450K chip, identifying 4 molecular subtypes. The MET level in the germ cell tumors was low, and these molecular subtypes provided information regarding their etiology. Again using the Illumina Infinium® Human Methylation 450K chip, Wu SP 1 et al. 31 detected the DNA MET state in 482 and 421 CpG loci in 10 samples of Ewing's sarcoma, 11 samples of synovial sarcoma, and 15 samples of osteosarcoma. Moreover, they developed and validated a whole-genome DNA MET classifier to identify osteosarcoma, Ewing's sarcoma, and synovial sarcoma. MET-based molecular typing is of great significance for diagnosing, recognizing, and treating morphologically overlapping solid tumors. Taskesen E et al. integrated the gene expression and DNA MET spectra of 344 samples of acute myeloid leukemia (AML) and established a regression model using Lasso. The results indicated that the subtype prediction of AML cytogenetics and molecular abnormalities could be significantly improved 32 . A study by Rodríguez-Rodero et al., demonstrated that thyroid carcinoma subtypes have promoter-differentiated MET features, and the molecular typing could be realized using abnormal DNA MET expression. Undifferentiated thyroid carcinoma was characterized by abnormal promoter hypomethylation, while differentiated papillary and follicular thyroid carcinoma was characterized by promoter hypermethylation 33 . To further explore the functions of the 250 screened MET loci, gene function annotation of the loci was performed and 42 genes were found to be significantly enriched to TFEC. The TFEC gene is located at 7q31.2 and encodes a polypeptide with a length of 347 amino acids, which is mainly localized in the nucleus and cytoplasm. According to a study by Chung et al. , TFEC plays a role as an activating transcription factor (ATF) for the non-myosin heavy chain II-a gene 34 . At present, evidence for the involvement of TFEC in cancer progression is limited; however, TFEC, MITF, TFEB, and TFE3 are important members of the MIT (microphthalmia-associated transcription factor) family, and recent studies have proven that changes in these transcription factors are related to melanoma, sarcoma, and renal cell carcinoma. With a similar structure to TFEB (another member of the MIT family), TFEC may play an important role in regulating genes related to autophagy and lysosomes 35 . The regulation of genes is a complex network; to investigate the effects of TFFC and its relevant factors on the occurrence and progression of tumors, function enrichment analysis was performed and these genes were found to be remarkably enriched to the following biological functions: GO:0006955 − immune response, GO:0050776 − regulation of the immune response, GO:0006954 − inflammatory response, GO:0045087 − innate immune response, and GO:0007165 − signal transduction. Currently, there are no reports of TFEC in ovarian cancer; thus, further investigation is needed. Finally, 5 CpG loci were screened via the WGCNA co-expression network: cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. The results show that hypomethylation of these 5 CpG loci was associated with poor prognosis in ovarian cancer patients. The gene annotated by the cg22809047 locus was RPL31, and Maruyama et al. 36 have previously shown that in comparison with benign prostate tissues, RPL31 is overexpressed in prostate cancer. In RPL31 siRNA-treated LNCaP and BicR cells, there is an increase in the protein expression levels of the tumor suppressor p53 and its targets, p21 and MDM2. In addition, the inhibition of cell growth and the cell cycle by RPL31 could be recovered by p53 siRNA treatment. RPL31 could be used as the target of molecular treatment for advanced prostate cancer, and we presume that RPL31 could also be used as a target for the treatment of ovarian cancer. ELOVL3 was the gene corresponding to the gene promoter region annotated by the cg00431050 locus. ELOVL3 is a member of the ELOVL (elongase of very long-chain fatty acids) family, which contains a total of 7 members (ELOVL1−7). The proteins encoded by the ELOVL1 − 7 genes are involved in the elongation of fatty acid chains of different lengths, and play an important role in regulating the biological synthesis of lipids, fatty acid metabolism, and certain metabolic diseases. There exist only limited studies of the involvement of ELOVL3 in tumors, while ELOVL2 has been widely described in tumors. A study by Kang et al., revealed that breast cancer patients with low ELOVL2 expression have a poor prognosis. ELOVL2 expression has been correlated with the malignant phenotype of breast cancer, and its downregulation induced lipid metabolism reprogramming; thus, ELOVL2 is a novel prognostic biomarker 37 . We suggest that ELOVL3 expression may also be involved in the occurrence and progression of ovarian cancer by inducing lipid metabolism reprogramming. Zhang et al., 38 investigated the molecular typing of serous ovarian cancer using the multi-omics data of DNA MET and protein, miRNA, and gene expression, mainly discussing the relationship between molecular typing based on RNA-Seq data and that based on other omics data. They finally screened 9 molecular subtypes based on RNA-Seq data; these subtypes had significant overlap with the molecular subtypes of other omics, but the function analysis results showed that the subtypes based on an omics dataset could not be completely substituted by other omics data. In the present study, the significance of MET in the molecular typing of ovarian cancer was analyzed using MET data, and the markers of subtypes closely related to the prognosis prediction of ovarian cancer were further screened. A MET data-based ovarian cancer prognosis prediction model was subsequently developed to provide a reference for clinical trials and researchers. In summary, the study by Zhang et al., and our study have different focal points, despite both involving molecular typing. Subtyping of ovarian carcinomas based on methylation profiles has been reported in a TCGA seminal article 13 , in which 4 subtypes were identified to be significantly associated with differences in age, BRCA inactivation events, and survival based on consensus clustering of variable DNA methylation. The cluster associated with the worst prognosis is characterized by hypomethylation and is associated with old age, which is in accordance with the present findings; however, our approach is different from that in the aforementioned TCGA paper. Firstly, the samples included in the TCGA paper were 489 cases of high-grade serous ovarian cancer, while the present paper included 571 cases of methylated ovarian cancer, including different clinical stages and grades. Our sample size is larger, and the results are more abundant. Secondly, a multivariate COX proportional hazards model was performed to elucidate that 250 CpG loci were significant predictors of prognosis, and 6 molecular subtypes were clustered based on the methylation level at these 250 CpG loci. The cluster that was characterized by hypomethylation was associated with a worse prognosis, stage, and grade, and an older patient age. Thirdly, weighted gene co-expression network analysis was further applied to identify the 5 most significant CpG loci, and hypomethylation of these 5 loci was demonstrated to be associated with a worse outcome. Conclusion We identified 6 different molecular subtypes using ovarian cancer MET data in the TCGA database and showed that DNA MET molecular typing could accurately support the distinction of tumor stage and pathological grade in ovarian cancer. The specific CpG loci and genes can be used as biomarkers for individualized treatment, which can be carried out in clinical practice according to these prognostic markers. Nevertheless, the present study has certain limitations: only internal validation was performed on MET prognostic loci and no suitable external datasets were obtained; thus, our study results need to be further validated with a larger sample size. Abbreviations TCGA : the Cancer Genome Atlas OS : overall survival Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Availability of data and materials All data during this study are included within this published article and additional files. Any material described in the article can be requested directly from corresponding author on reasonable request. All data in the current study are based on public data available in The Cancer Genome Atlas (TCGA) datasets. Declaration of Interests The authors declare no competing financial interests Non-Financial Conflict of Interest The authors declare no non-financial conflict of interest. Funding Sources No Author Contributions Qing Yang designed and conceptualized this study; Lili Yin and Ningning Zhang were major contributors in experiment; Lili Yin analyzed the data; Lili Yin, and Qing Yang were major contributors in writing the manuscript. All authors read and approved the final manuscript. Acknowledgements Not applicable. References Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. CA: a cancer journal for clinicians. 2006;56(2):106-130. Herzog TJ. Recurrent ovarian cancer: how important is it to treat to disease progression? Clinical cancer research : an official journal of the American Association for Cancer Research. 2004;10(22):7439-7449. Pearce CL, Templeman C, Rossing MA, et al. Association between endometriosis and risk of histological subtypes of ovarian cancer: a pooled analysis of case-control studies. The Lancet Oncology. 2012;13(4):385-394. Sung P-L, Chang Y-H, Chao K-C, Chuang C-M. Global distribution pattern of histological subtypes of epithelial ovarian cancer: a database analysis and systematic review. Gynecologic oncology. 2014;133(2):147-154. Böcker W. [WHO classification of breast tumors and tumors of the female genital organs: pathology and genetics]. Verh Dtsch Ges Pathol. 2002;86:116-119. Kommoss S, Schmidt D, Kommoss F, et al. Histological grading in a large series of advanced stage ovarian carcinomas by three widely used grading systems: consistent lack of prognostic significance. A translational research subprotocol of a prospective randomized phase III study (AGO-OVAR 3 protocol). Virchows Archiv : an international journal of pathology. 2009;454(3):249-256. Shih I-M, Kurman RJ. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. The American journal of pathology. 2004;164(5):1511-1518. Miller JA, Kotecha R, Ahluwalia MS, et al. Overall survival and the response to radiotherapy among molecular subtypes of breast cancer brain metastases treated with targeted therapies. Cancer. 2017;123(12):2283-2293. Roode SC, Rotroff D, Avery AC, et al. Genome-wide assessment of recurrent genomic imbalances in canine leukemia identifies evolutionarily conserved regions for subtype differentiation. Chromosome research : an international journal on the molecular, supramolecular and evolutionary aspects of chromosome biology. 2015;23(4):681-708. Dolly SO, Collins DC, Sundar R, Popat S, Yap TA. Advances in the Development of Molecularly Targeted Agents in Non-Small-Cell Lung Cancer. Drugs. 2017;77(8):813-827. Xu Y, Chen M, Liu C, et al. Association Study Confirmed Three Breast Cancer-Specific Molecular Subtype-Associated Susceptibility Loci in Chinese Han Women. The oncologist. 2017;22(8):890-894. Tothill RW, Tinker AV, George J, et al. Novel molecular subtypes of serous and endometrioid ovarian cancer linked to clinical outcome. Clinical cancer research : an official journal of the American Association for Cancer Research. 2008;14(16):5198-5208. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609-615. Kommoss S, Winterhoff B, Oberg AL, et al. Bevacizumab May Differentially Improve Ovarian Cancer Outcome in Patients with Proliferative and Mesenchymal Molecular Subtypes. Clinical cancer research : an official journal of the American Association for Cancer Research. 2017;23(14):3794-3801. Ushijima T, Asada K. Aberrant DNA methylation in contrast with mutations. Cancer science. 2010;101(2):300-305. Wilting RH, Dannenberg J-H. Epigenetic mechanisms in tumorigenesis, tumor cell heterogeneity and drug resistance. Drug resistance updates : reviews and commentaries in antimicrobial and anticancer chemotherapy. 2012;15(1-2):21-38. Esteller M, Silva JM, Dominguez G, et al. Promoter hypermethylation and BRCA1 inactivation in sporadic breast and ovarian tumors. Journal of the National Cancer Institute. 2000;92(7):564-569. Baylin SB, Ohm JE. Epigenetic gene silencing in cancer - a mechanism for early oncogenic pathway addiction? Nature reviews Cancer. 2006;6(2):107-116. Esteller M. Epigenetics in cancer. The New England journal of medicine. 2008;358(11):1148-1159. Baylin SB, Chen WY. Aberrant gene silencing in tumor progression: implications for control of cancer. Cold Spring Harb Symp Quant Biol. 2005;70:427-433. Petrocca F, Iliopoulos D, Qin HR, et al. Alterations of the tumor suppressor gene ARLTS1 in ovarian cancer. Cancer research. 2006;66(21):10287-10291. Horak P, Pils D, Haller G, et al. Contribution of epigenetic silencing of tumor necrosis factor-related apoptosis inducing ligand receptor 1 (DR4) to TRAIL resistance and ovarian cancer. Molecular cancer research : MCR. 2005;3(6):335-343. Hu W-L, Zhou X-H. Identification of prognostic signature in cancer based on DNA methylation interaction network. BMC medical genomics. 2017;10(Suppl 4):63. Tomczak K, Czerwińska P, Wiznerowicz M. The Cancer Genome Atlas (TCGA): an immeasurable source of knowledge. Contemp Oncol (Pozn). 2015;19(1A):A68-A77. Zhang S, Wang Y, Gu Y, et al. Specific breast cancer prognosis-subtype distinctions based on DNA methylation patterns. Molecular oncology. 2018;12(7):1047-1060. Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics (Oxford, England). 2010;26(12):1572-1573. Reimand J, Arak T, Adler P, et al. g:Profiler-a web server for functional interpretation of gene lists (2016 update). Nucleic acids research. 2016;44(W1):W83-W89. Aran D, Hellman A. DNA methylation of transcriptional enhancers and cancer predisposition. Cell. 2013;154(1):11-13. Jurmeister P, Schöler A, Arnold A, et al. DNA methylation profiling reliably distinguishes pulmonary enteric adenocarcinoma from metastatic colorectal cancer. Modern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc. 2019;32(6):855-865. Williams LA, Mills L, Hooten AJ, et al. Differences in DNA methylation profiles by histologic subtype of paediatric germ cell tumours: a report from the Children's Oncology Group. British journal of cancer. 2018;119(7):864-872. Wu SP, Cooper BT, Bu F, et al. DNA Methylation-Based Classifier for Accurate Molecular Diagnosis of Bone Sarcomas. JCO Precis Oncol. 2017;2017. Taskesen E, Babaei S, Reinders MMJ, de Ridder J. Integration of gene expression and DNA-methylation profiles improves molecular subtype classification in acute myeloid leukemia. BMC bioinformatics. 2015;16 Suppl 4:S5. Rodríguez-Rodero S, Fernández AF, Fernández-Morera JL, et al. DNA methylation signatures identify biologically distinct thyroid cancer subtypes. J Clin Endocrinol Metab. 2013;98(7):2811-2821. Chung MC, Kim HK, Kawamoto S. TFEC can function as a transcriptional activator of the nonmuscle myosin II heavy chain-A gene in transfected cells. Biochemistry. 2001;40(30):8887-8897. Haq R, Fisher DE. Biology and clinical relevance of the micropthalmia family of transcription factors in human cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2011;29(25):3474-3482. Maruyama Y, Miyazaki T, Ikeda K, et al. Short hairpin RNA library-based functional screening identified ribosomal protein L31 that modulates prostate cancer cell growth via p53 pathway. PloS one. 2014;9(10):e108743. Kang YP, Yoon J-H, Long NP, et al. Spheroid-Induced Epithelial-Mesenchymal Transition Provokes Global Alterations of Breast Cancer Lipidome: A Multi-Layered Omics Analysis. Frontiers in oncology. 2019;9:145. Zhang Z, Huang K, Gu C, et al. Molecular Subtyping of Serous Ovarian Cancer Based on Multi-omics Data. Scientific reports. 2016;6:26001. Tables Table 1: The clinical pathological information in the training set and validation test. Validation set Traning set Stage Stage I 5 11 Stage II 15 12 Stage III 222 215 Stage IV Grade 41 45 G1 3 3 G2 38 31 G3 238 243 G4 0 1 Age 60 115 138 Table 2: The top 20 loci with the most significant difference in prognosis. CpGs p.value HR Low 95%CI High 95%CI cg25781123 1.42E-05 144.7952 15.31263 1369.173 cg01278291 1.68E-05 0.013765 0.001955 0.096903 cg21291896 2.93E-05 1.24E + 13 8962882 1.72E + 19 cg08946332 5.46E-05 0.21193 0.099747 0.450282 cg13804316 8.51E-05 213979.4 469.6279 97496703 cg16179125 9.40E-05 8.255374 2.862145 23.81123 cg13060646 0.000201 4.08882 1.945923 8.591523 cg03750606 0.000282 31.81681 4.916032 205.92 cg15341340 0.000317 3559149 966.7441 1.31E + 10 cg08013810 0.00033 0.087427 0.023115 0.330674 cg06797533 0.000383 4.351053 1.932616 9.795869 cg21022435 0.000396 0.00496 0.000263 0.093416 cg22916109 0.000475 1.07E + 11 69831.7 1.65E + 17 cg10415235 0.00048 6.56E + 08 7370.126 5.83E + 13 cg05955301 0.000514 8.98125 2.60262 30.99294 cg16016036 0.00056 0.243586 0.109202 0.543343 cg23486067 0.000569 20.19834 3.654797 111.6267 cg25634666 0.00057 0.255673 0.117688 0.555439 cg17332016 0.000602 104162.1 141.5735 76636837 cg03190825 0.000654 15.75275 3.227124 76.89485 Table 3: The CpG loci in different modules. Module Count black 12 blue 34 brown 22 green 16 red 14 turquoise 49 yellow 22 Table 4 ： The annotation of 5 CpG loci. CpG Chrom Start End GeneSymbol Feature_Type cg03152385 chr16 15094739 15094740 RP11-72I8.1 S_Shore cg27625732 chr9 1.29E + 08 1.29E + 08 TBC1D13 N_Shore cg22197830 chr5 1.35E + 08 1.35E + 08 TXNDC15 N_Shore cg22809047 chr2 1.01E + 08 1.01E + 08 AC016738.4 Island cg03152385 chr16 15094739 15094740 RRN3 S_Shore cg00431050 chr10 1.02E + 08 1.02E + 08 ELOVL3 N_Shore cg22809047 chr2 1.01E + 08 1.01E + 08 RPL31 Island Supplementary File Information Table S1: A total of 967 loci demonstrated a significant difference in prognosis Table S2: 250 significant MET loci Table S3: Distribution of methylation levels in each subtype Table S4: The distribution of grade and stage in each subtype Table S5: Age distribution of samples in each subtype Table S6: The R package covered in this article Supplementary Files Equation.docx TableS5.docx TableS3.docx TableS4.docx TableS1.docx TableS6.docx TableS2.docx Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-13880","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Primary research","associatedPublications":[],"authors":[{"id":364920,"identity":"9b1c79f8-6a17-4b87-ab7f-e83be93a930a","order_by":1,"name":"Lili Yin","email":"","orcid":"","institution":"China Medicine Universitry","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Lili","middleName":"","lastName":"Yin","suffix":""},{"id":364921,"identity":"f9742b0f-00d9-4bb6-a5ee-2a137ffa2aba","order_by":2,"name":"Ningning Zhang","email":"","orcid":"","institution":"China Medical Universitry","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Ningning","middleName":"","lastName":"Zhang","suffix":""},{"id":364922,"identity":"2a7a0a16-ec59-4ded-98f8-288d549b15cd","order_by":3,"name":"Qing Yang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIie3RPQrCMBTA8SdCXYJZ00U9QqAgCGKv8qSQqUpHx052EG/QQ3gDgxlcWrp2cFCE4tBBR0HEKs792ATzHzKE90sIAdDpfjb+WadHXIybEYcfI9HsLsM8LXc1zt/H6pJ7h4lPZbZAQwINVlhOorkYhTxzfIYiRXIAFsWbciLdoUW4coBBQVgGnM0qSJJ/CZXCQ65qkNS1zgWZAKAAxBrETPNhK+QKgaHDUApS+ZZu4lrX/KFsoNH0dn+OezRYl5OBBIOR4ue3kuB7g5SOv+v70L4WYzZAR1ZO63Q63X/2AtglTjf9ag3vAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-7324-6103","institution":"Shengjing Hospital of China Medical University","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Qing","middleName":"","lastName":"Yang","suffix":""}],"badges":[],"createdAt":"2020-02-10 16:04:03","currentVersionCode":2,"declarations":"","doi":"10.21203/rs.2.23311/v2","doiUrl":"https://doi.org/10.21203/rs.2.23311/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":529331,"identity":"6948fc1e-7671-4452-ab12-82d03c831644","added_by":"auto","created_at":"2020-02-21 18:13:19","extension":"tif","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":90362,"visible":true,"origin":"","legend":"A: CDF curve; different colors reflect different cluster numbers; the horizontal axis represents the consensus index; the vertical axis represents the cumulative distribution function (CDF); and a larger AUC indicates better clustering. B: CDF delta area curve of consensus clustering indicating the relative change in the area under the CDF curve for each category number k as compared with k – 1; the horizontal axis represents the category number k and the vertical axis represents the relative change in the area under the CDF curve.","description":"","filename":"Figure1.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure1.tif"},{"id":529334,"identity":"7bb323d8-1264-4b61-b19b-21f70d079bf4","added_by":"auto","created_at":"2020-02-21 18:13:20","extension":"tif","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":803193,"visible":true,"origin":"","legend":"A: Clustering heat map of samples at consensus k = 6. Different colors reflect different cluster numbers; the color gradient is from white to blue, indicating the consensus of progression. B: Clustering results of 250 MET loci, clinical stage, and histological grade in 6 clusters of samples; red represents high expression, blue represents low expression.","description":"","filename":"Figure2.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure2.tif"},{"id":529335,"identity":"d5679f90-10d7-4676-b5fb-2c58b1483bf0","added_by":"auto","created_at":"2020-02-21 18:13:20","extension":"tif","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":404877,"visible":true,"origin":"","legend":"A: Prognosis differences among the 6 subtypes of samples; different colors represent different molecular subtypes; the horizontal axis reflects the survival time and the vertical axis represents the survival rate. B: Percentages of samples of different clinical stages in the 6 subtypes. C. Percentages of samples of different grades in the 6 subtypes; the horizontal axis represents different molecular subtypes and the vertical axis represents the percentage. D. Age distribution of the patients of the samples in the 6 subtypes; the horizontal axis represents different molecular subtypes and the vertical axis represents age.","description":"","filename":"Figure3.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure 3.tif"},{"id":529336,"identity":"7003a20a-2cfe-4a6a-b97e-fca22111c0f8","added_by":"auto","created_at":"2020-02-21 18:13:20","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2495100,"visible":true,"origin":"","legend":"A: Transcription factor enrichment results of genes corresponding to the gene promoter regions annotated by 250 CpG loci; green represents the transcription factors, pink represents the genes annotated by MET loci, and purple represents MET loci. B: Chord diagram showing the top 5 enriched GO clusters for genes associated with TFEC. In each chord diagram, enriched GO clusters are shown on the right and genes contributing to enrichment are shown on the left. Positively correlated molecules are displayed in red and negatively correlated molecules are displayed in blue. Each GO term is represented by one colored line.","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure 4.png"},{"id":529340,"identity":"9586dbda-f337-4282-85d8-760337c04852","added_by":"auto","created_at":"2020-02-21 18:13:21","extension":"tif","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":346969,"visible":true,"origin":"","legend":"A: Evaluation of the scale-free model at different soft thresholds; a larger value indicates better compliance with the features of the biological network. B: Mean connectivity at different soft thresholds; the horizontal axis represents the soft threshold and the vertical axis represents the mean connectivity. C: Gene dendrogram and module colors; different colors represent the genes in different modules. D: Module-feature correlation; the row represents the eigengenes of each module and the column represents the feature information of the samples. Red to green represents a high to low correlation coefficient. The digit in each grid indicates the correlation coefficient between gene modules and the corresponding features, and the digit in the bracket represents the P value.","description":"","filename":"Figure5.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure5.tif"},{"id":529341,"identity":"992d2fbc-00af-45a9-9edd-26f65a020bda","added_by":"auto","created_at":"2020-02-21 18:13:22","extension":"tif","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":341891,"visible":true,"origin":"","legend":"A: Weighted interaction network of 22 CpG loci in the yellow module; the circle represents CPG loci, the connection line between two circles represents the interaction relationship, and a redder color indicates a larger node degree. B: Correlation coefficient clustering of 22 CpG loci; a more purple color indicates a smaller correlation coefficient and a redder color indicates a larger correlation coefficient.","description":"","filename":"Figure6.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure6.tif"},{"id":529343,"identity":"bec54222-83b9-4e02-b86e-0c1c156ff517","added_by":"auto","created_at":"2020-02-21 18:13:22","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":211201,"visible":true,"origin":"","legend":"A: MET spectrum clustering results of the 5 CpG loci. B: Prognosis difference between the hypermethylation and hypomethylation groups formed by clustering. The horizontal axis represents the survival time (months), the vertical axis represents the survival rate, red indicates the hypomethylation group, and blue indicates the hypermethylation group.","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure7.png"},{"id":529344,"identity":"fdc076f9-063f-407e-9d18-a50b1615f01e","added_by":"auto","created_at":"2020-02-21 18:13:22","extension":"tif","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":178714,"visible":true,"origin":"","legend":"A: Expression spectrum clustering results of the 5 CpG loci. B: Prognosis difference between the hypermethylation and hypomethylation groups formed by clustering. The horizontal axis represents the survival time (months), the vertical axis represents the survival rate, red indicates the hypomethylation group, and blue indicates the hypermethylation group.","description":"","filename":"Figure8.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure8.tif"},{"id":529345,"identity":"f1644aa3-f9e8-4a39-b89d-3068ffb37c8e","added_by":"auto","created_at":"2020-02-21 18:13:22","extension":"tif","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":50373,"visible":true,"origin":"","legend":"Flow chart of all the analysis","description":"","filename":"Figure9.tif","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Figure 9.tif"},{"id":13490153,"identity":"8f053e90-9879-487f-9459-828292498d8e","added_by":"auto","created_at":"2021-09-16 22:22:21","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3863106,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-13880/v2/9813c976-980d-45d2-a0e8-f9f1d99f6c67.pdf"},{"id":529342,"identity":"da3ccb81-5be8-476d-b612-74fd771ee9c0","added_by":"auto","created_at":"2020-02-21 18:13:22","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":26264,"visible":true,"origin":"","legend":"","description":"","filename":"Equation.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Equation.docx"},{"id":529338,"identity":"275fc606-c9a2-49cb-89a0-b181826c8317","added_by":"auto","created_at":"2020-02-21 18:13:21","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":320005,"visible":true,"origin":"","legend":"","description":"","filename":"TableS5.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S5.docx"},{"id":529337,"identity":"022d484f-7c49-4e3b-b788-32a29256a9bb","added_by":"auto","created_at":"2020-02-21 18:13:21","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":357365,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S3.docx"},{"id":529339,"identity":"02eda076-fb30-4f1b-8cfb-378b57535175","added_by":"auto","created_at":"2020-02-21 18:13:21","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":494413,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S4.docx"},{"id":529332,"identity":"b9101094-a6b4-4455-a975-d5b0cf4dad43","added_by":"auto","created_at":"2020-02-21 18:13:20","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":132869,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S1.docx"},{"id":529333,"identity":"67b430da-4c8b-4ad2-a4d3-01187b5fe882","added_by":"auto","created_at":"2020-02-21 18:13:20","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":19188,"visible":true,"origin":"","legend":"","description":"","filename":"TableS6.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S6.docx"},{"id":529330,"identity":"45c0771b-268b-4ff5-a907-bc02766b689d","added_by":"auto","created_at":"2020-02-21 18:13:19","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":37082,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.docx","url":"https://assets-eu.researchsquare.com/files/e9307d24-1a36-45a1-8117-c32a5eb29684/v2/Table S2.docx"}],"financialInterests":"","formattedTitle":"Mining of Subtype Markers for the Prognosis of Ovarian Cancer based on Methylation Data","fulltext":[{"header":"Background","content":"\u003cp\u003eOvarian cancer is a disease of high heterogeneity with varying molecular phenotypes, pathogeneses, and prognoses, and its morbidity is ranked number 3 among malignant gynecological tumors. However, ovarian cancer is not easily detected at an early stage since the ovaries are located deep in the pelvis, and at diagnosis, the tumor is advanced with distant metastases in 70% of cases. Most patients will experience recurrence within 2 years, and there is a lack of effective therapies for recurrent ovarian cancer, which ranks the mortality of ovarian cancer as number 1 among gynecological tumors. Therefore, precision therapy is an urgent demand.\u003c/p\u003e\n\u003cp\u003eEpithelial ovarian cancer is the most common pathological type and accounts for 80\u0026minus;95% of ovarian malignancies. It is classified into 5 histological subtypes: high-grade serous adenocarcinoma, endometrioid adenocarcinoma, clear cell adenocarcinoma, mucous adenocarcinoma, and low-grade serous adenocarcinoma \u003csup\u003e1-5\u003c/sup\u003e. In a large, prospective, phase-III clinical study, Kommoss et al., showed that histological grading is necessary in early ovarian cancer but has no significance in prognostic evaluation in advanced ovarian cancer \u003csup\u003e6\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eBy combining the clinical pathological and molecular biological characteristics of ovarian cancer, \u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Shih%20IeM%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=15111296\"\u003eShih IeM\u003c/a\u003e et al., classified epithelial ovarian cancer into types I and II. Type I epithelial ovarian cancer mainly includes low-grade serous carcinoma and low-grade endometrioid carcinoma, and most cases have early onset and good prognosis. Type II epithelial ovarian cancer mainly includes high-grade serous carcinoma, high-grade endometrioid carcinoma, and undifferentiated carcinoma, and most cases have rapid onset and poor prognosis. Type I epithelial ovarian cancer is significantly correlated with mutations in BRAF, KRAS, and PTEN, while type II epithelial ovarian cancer is associated with p53 mutations and also very frequently with BRCA1/2 mutations. The dualistic theory reflects different biological behaviors and clinical prognoses of tumors, and such differences are especially remarkable between low-grade and high-grade serous ovarian carcinomas. However, application of the dualistic theory in non-serous ovarian carcinoma is limited; for instance, clear cell carcinoma has many biological behaviors similar to type II epithelial ovarian cancer, despite belonging to type I epithelial ovarian cancer. It is of profound significance to realize precision molecular typing of ovarian cancer for clinical treatment and prognosis monitoring \u003csup\u003e7\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eAs a result of the continuous improvement in human genome sequencing technology and the advancement of biomedical analysis technology, a new trend has been formed that involves molecular targeted therapy and prognosis evaluation based on the molecular typing of malignant tumors. Molecular-targeted therapy has been successfully applied in several tumors, including ER (+) or HER2(+) breast cancer and EGFR-mutated lung cancer \u003csup\u003e8-11\u003c/sup\u003e, presenting great progress in precision medical treatment. By K-means clustering, Tothill et al. \u003csup\u003e12\u003c/sup\u003e detected the gene expression spectrum of 285 cases of endometrioid and serous tumors originating from the ovary, peritoneum, and tuba uterine, finally identified 6 molecular subtypes, of which 4 (high interstitium-reactive, high immunity, hypomethylation trix-reactive, and interstitial low immunity) are features of high-grade serous ovarian carcinoma and can be used for predicting prognosis in these patients.\u003c/p\u003e\n\u003cp\u003eBased on gene expression, TCGA and Tothill et al., divided high-grade serous carcinoma into 4 subtypes: immunoreactive, differentiated, proliferative, and interstitial \u003csup\u003e13\u003c/sup\u003e. Moreover, the clinical trial completed by Kommoss et al. showed that in high-grade serous ovarian carcinoma patients with proliferative and interstitial molecular subtypes, bevacizumab can improve the progression-free survival to different degrees; however, this study was limited to the molecular typing of the gene expression spectrum \u003csup\u003e14\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eDespite the impact of gene changes on cancer occurrence, epigenetic changes such as DNA MET also play an important role. Epigenetic inheritance refers to the hereditary changes that occur under the precondition of no changes in DNA sequence, such as histone modification, DNA MET, RNA editing, and gene silencing. The occurrence and growth of ovarian cancer involve several pathways including DNA repair, cell apoptosis, cell cycle regulation, and changes in protooncogenes tumor suppressor genes. Studies have suggested that epigenetic changes in these pathways play essential roles in the development of ovarian cancer, and the detection of MET signals is helpful for early diagnosis \u003csup\u003e15,16\u003c/sup\u003e. DNA MET mainly occurs in CpG islands; CpG expression can be inhibited by the hypermethylation of tumor suppressor gene promoters and enhanced by the decreased demethylation probability of protooncogenes. Different regulatory effects of protooncogenes and tumor suppressor genes contribute to the occurrence of cancer \u003csup\u003e17-20\u003c/sup\u003e. According to relevant studies, the tumor suppressor gene involved in ovarian cancer exists in a hypermethylated state, and changes in its MET level is an important molecular foundation for cancer occurrence\u003csup\u003e21,22\u003c/sup\u003e. \u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Hu%20WL%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=29322932\"\u003eHu WL\u003c/a\u003e et al., \u003csup\u003e23\u003c/sup\u003ebuilt a DNA MET interaction network for ovarian cancer, breast cancer, and glioma, confirming that the number of DNA MET loci was associated with prognosis in cancer patients; however, no DNA MET molecular typing of ovarian cancer was performed.\u003c/p\u003e\n\u003cp\u003eIn the present study, the univariate and multivariate COX proportional hazards models were established by analyzing Illumina Infinium\u0026reg; Human Methylation 27K data in the TCGA database and combining the MET level and clinical data of the samples. Subsequently, 6 molecular subtypes associated with the prognosis in ovarian cancer patients were screened by consensus clustering of MET spectra with a significant difference in both models, and the ovarian cancer patients were classified using these subtypes. Further, 5 CpG hypomethylation loci related to poor prognosis in ovarian cancer patients were obtained by constructing a WGCNA co-expression network. These loci are of great significance for the clarification of the pathogenesis of ovarian cancer, and can be used as effective tumor markers to provide a reference for clinical prognosis and individualized treatment.\u003c/p\u003e"},{"header":"Materials And Methods","content":"\u003col\u003e\n\u003cli\u003e\u003cstrong\u003e Preprocessing of ovarian cancer expression datasets and preliminary screening of DNA MET loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eTCGA \u003csup\u003e24\u003c/sup\u003e GDC API was utilized to download the latest clinical follow-up information and RNA-Seq data. Illumina Infinium\u0026reg; Human Methylation 27 BeadChip (\u003cem\u003eIllumina\u003c/em\u003e \u003cem\u003e27K\u003c/em\u003e) microarray was acquired from UCSC Cancer Browser. Samples with complete clinical data and methylation spectrum data were selected. CpG loci with NA (Not Available) \u0026gt; 70% in all samples were deleted. The \u003cstrong\u003e\u003cem\u003eimpute-KNN\u003c/em\u003e\u003c/strong\u003e of R package was used to fill the missing value of methylation spectrum. The unstable genomic methylation loci were further removed, involving CpGs and single nucleotide loci on sex chromosomes as well as CpG loci that were not annotated to the gene promoter region \u003csup\u003e25\u003c/sup\u003e. We divided the datasets into two queues: a training set coupled with a test set. The standards for the subgroups included: 1) Samples were assigned to the training set and the test set randomly; and 2) the data of the two groups should be similar, including age distribution, clinical stage, follow-up time, and mortality ratio.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"2\"\u003e\n\u003cli\u003e\u003cstrong\u003e Univariate survival analysis of MET loci in the training set\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe research objective was to determine the molecular subtypes of ovarian carcinoma as prognostic determinants. Therefore, CpG loci, which had an important impact on survival, were utilized as a classification feature. First and foremost, a univariate COX proportional hazard model was established based on the methylation level of each CpG loci, age, tumor grade and stage, coupled with survival data by coxph function of the R-package\u003cem\u003e \u003cstrong\u003esurvival\u003c/strong\u003e\u003c/em\u003e. Subsequently, we introduced the significant CpG loci obtained from the univariate model into the multivariate COX proportional hazards model, and took the significant age and clinical attributes in the univariate model as covariables. Ultimately, the CpG loci, which were still significant, were employed as classification features [25]. For each CpG island, the multivariate COX proportional hazards model formula was described below:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eh(\u003cem\u003et\u003c/em\u003e, \u003cem\u003ex\u003c/em\u003e\u003cem\u003e)\u003c/em\u003e\u003cem\u003ei \u003c/em\u003e= \u003cem\u003eh\u003c/em\u003e0(\u003cem\u003et\u003c/em\u003e)exp\u003cem\u003e(\u0026beta;\u003csub\u003emethy\u003c/sub\u003emethyi + \u0026beta;\u003csub\u003eage\u003c/sub\u003eage + \u0026beta;\u003csub\u003estage\u003c/sub\u003estage) \u003c/em\u003e(1)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the formula, \u0026ldquo;methy\u003csub\u003ei\u003c/sub\u003e\u0026rdquo; is the carrier of the CpG locus methylation level in the sample. \u0026ldquo;Age\u0026rdquo; and \u0026ldquo;stage\u0026rdquo; describe the age and clinical characteristics of the patients, respectively. \u0026ldquo;\u0026beta;\u003csub\u003emethy\u003c/sub\u003e\u0026rdquo;, \u0026ldquo;\u0026beta;\u003csub\u003eage\u003c/sub\u003e\u0026rdquo;, and \u0026ldquo;\u0026beta;\u003csub\u003estage\u003c/sub\u003e\u0026rdquo; are regression coefficients. The \u003cem\u003eP\u003c/em\u003e-values of the COX regression coefficient was adjusted by Benjamini\u0026minus;Hochberg error detection rate. Various comparing processes were carried out.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"3\"\u003e\n\u003cli\u003e\u003cstrong\u003e Screening of molecular subtypes by the consensus clustering of methylation profile with a significant difference in both univariate and multivariate analyses\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eConsensus ClusterPlus\u003c/em\u003e\u003c/strong\u003e in the R package \u003csup\u003e26\u003c/sup\u003e was utilized for consensus clustering according to the method described by Zhang et al. \u003csup\u003e25\u003c/sup\u003e The subgroups of epithelial ovarian tumors were identified based on the most variable CpG loci. The algorithm is described as follows. First, double sampling of some items and features from the data matrix was conducted, in which each sub-sample was divided into several groups (max.) using a user-specific clustering algorithm (\u003cem\u003ek-means\u003c/em\u003e, hierarchical\u0026nbsp;clustering, or custom algorithms). The paired consensus value (defined as the proportion of clustering running for the combination of two items) was calculated and stored in the k\u003csub\u003ei\u003c/sub\u003e consensus matrix. Second, the final coherent sheaf clustering for each k\u003csub\u003ei\u003c/sub\u003e was completed using the distance of 1- consensus value and pruned into k\u003csub\u003ei\u003c/sub\u003e group through cutting, which is known as consensus clustering. The algorithm determined the \u0026ldquo;consensus\u0026rdquo; clustering by measuring the stability of the clustering results applied to random data subsets from given clustering methods. In each iteration, 80% of the tumors were sampled, and a \u003cem\u003ek-means\u003c/em\u003e algorithm with the Euclidean squared distance\u0026nbsp;measures were utilized:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e[Please see the supplementary files section to access this equation.] (2)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThere was \u003cem\u003ek\u003c/em\u003e = 2\u0026minus;10 groups, and these results were compiled for 100 times. The cluster consensus as well as item consensus results were obtained with \u003cstrong\u003e\u003cem\u003eConsensus ClusterPlus.\u003c/em\u003e\u003c/strong\u003eR package. The graphical output results included the heat map of consensus matrix, cumulative distribution function (CDF) diagram, and \u0026Delta; region diagram. The criteria of clustering number included relatively high consistency within the cluster, relatively low coefficient of variation, and insignificant increase of the area under the CDF curve (AUC). The CV (%) was calculated based on the formula below:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCV (%) = (SD/MN) x 100 (3)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhere SD is the standard deviation, while MN is the average value of the samples. We selected category number as the area under the CDF curve, and there was no significant change. The consensus clustering heatmap was generated using the R package \u003cstrong\u003e\u003cem\u003epheatmap\u003c/em\u003e\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"4\"\u003e\n\u003cli\u003e\u003cstrong\u003e Clustering analysis of the methylation expression profile and analysis of the clinical characteristics of screened molecular subtypes\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe stable clustering results were selected and the methylation profile was analyzed by clustering analysis. The distance between the MET loci was calculated using the Euclidean distance. Furthermore, the distribution of various molecular subtype samples was analyzed with respect to prognosis, stage, grade, and age.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"5\"\u003e\n\u003cli\u003e\u003cstrong\u003e Gene annotation of MET loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eAs for the genes corresponding to the gene promoter regions annotated by the selected CpG loci, the transcription factor enrichment analysis was performed by the online tool g:profiler \u003csup\u003e27\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"6\"\u003e\n\u003cli\u003e\u003cstrong\u003eWGCNA co-expression analysis of CpG loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eBased on the modification beta value of selected CpG loci, the co-expressed CpG loci were mined by WGCNA co-expression algorithm. The distance between CpG loci was calculated using Pearson Correlation Coefficient. The R package\u003cstrong\u003e\u003cem\u003e WGCNA\u003c/em\u003e\u003c/strong\u003e was used to construct weighted co-expression network, and select a soft threshold of 4 to filter the CpG co-expression modules. The results showed that the co-expression network conformed to the scale-free network. That is to say, the log(k) of node k presented in the connection is negatively correlated with the log(P(k)) of the probability of node k, with a correlation coefficient larger than 0.8. In order to ensure a scale-free network, we chose \u0026beta; = 4. The next step was to convert the expression matrix into an adjacency matrix, and then transform the adjacency matrix into a topological matrix. Based on TOM, we utilized average-linkage hierarchical\u0026nbsp;clustering method to cluster genes. The minimum number of genes in each IncRNA network module was set at 30 according to the standard merged dynamic tree cutting. After determining the gene modules with the dynamic cutting method, we calculated eigengenes of each module in turn. The modules were clustered, and the adjacent modules were merged into new modules.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"7\"\u003e\n\u003cli\u003e\u003cstrong\u003e Construction of prognosis models and data validation of independent test set\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eUnsupervised clustering analysis was conducted on the CpG methylation profile selected in the previous step. The similarity between samples was calculated by using the Euclidean distance. The samples were then divided into two groups according to the methylation level of CpG loci. The prognosis differences between the two groups were further analyzed. The methylation profile of 286 samples in the test set were used for validation.\u003c/p\u003e"},{"header":"Results","content":"\u003col\u003e\n\u003cli\u003e\u003cstrong\u003e Selection of\u003c/strong\u003e \u003cstrong\u003e250 characteristic MET loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eThe\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e \u003cem\u003eIllumina\u003c/em\u003e \u003cem\u003eInfinium\u0026reg;\u003c/em\u003e\u003cem\u003e \u003cem\u003eHuman Methylation 27\u003c/em\u003e\u003c/em\u003e BeadChip\u003cem\u003e microarray\u003c/em\u003e \u003c/strong\u003econtained 613 samples in total, with 571 samples being screened using MET detection. The missing data imputation of the MET spectrum was performed using Impute in the R software package, and 25,154 MET loci were selected following exclusion of the unstable genomic MET loci. All 571 samples were assigned to either a training set (n = 285) or a validation set (n = 286). The clinical pathological information in the training set and validation test is shown in \u003cstrong\u003eTable 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThe MET loci and survival data were analyzed using a univariate COX proportional hazards regression model with \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.05 as the threshold. A total of 967 loci demonstrated a significant difference in prognosis (\u003cstrong\u003eTable S1\u003c/strong\u003e), of which the top 20 loci with the most significant difference are shown in \u003cstrong\u003eTable 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThe prognostic significance of age had a logrank \u003cem\u003eP\u003c/em\u003e value of 5.93e-06, while that of stage was 0.0379. The significant MET loci were selected using a univariate COX model followed by multivariate COX proportional hazards regression model analysis, with stage and age as covariates. Finally 250 significant MET loci were obtained (\u003cstrong\u003eTable S2\u003c/strong\u003e). \u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"2\"\u003e\n\u003cli\u003e\u003cstrong\u003eScreening 6 molecular subtypes by consensus clustering of the MET loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eConsensus clustering of MET loci with a significant difference in both univariate and multivariate analyses was performed using \u003cstrong\u003e\u003cem\u003eConsensus ClusterPlus\u003c/em\u003e\u003c/strong\u003e in the R package to screen the molecular subtypes. The similarity between samples was calculated using the Euclidean distance, the clustering was performed with\u003cem\u003e k-means,\u003c/em\u003e and 80% sampling was conducted 100 times using a double-sampling method. The optimal cluster number was determined by CDF; different colors in the CDF curve represent different cluster numbers (\u003cstrong\u003eFigure 1A\u003c/strong\u003e), the AUC was larger at 6 and 7 clusters and the clustering effect was better. Further observation of the CDF delta area curve (\u003cstrong\u003eFigure 1B\u003c/strong\u003e) shows that at 6 clusters, the AUC demonstrates stable clustering results. Following consideration, \u003cem\u003ek\u003c/em\u003e = 6 was selected and 6 molecular subtypes were obtained.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"3\"\u003e\n\u003cli\u003e\u003cstrong\u003e Clustering analysis of the MET expression spectra of 6 molecular subtypes \u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe composition and number of samples in the 6 clusters were evaluated using the consensus matrix. The color gradient was from white to blue, indicating the consensus of progression. In the matrix permutation, the same clusters were made mutually adjacent; and eventually, a color-coded heat map was displayed and featured by the arrangement of dark blue blocks on a diagonal white background (\u003cstrong\u003eFigure 2A\u003c/strong\u003e), showing that 285 tumor samples were assigned to these 6 clusters. Furthermore, clustering analysis was performed on 250 MET spectra; the distance between MET loci was calculated using the Euclidean distance, and the heat map was generated by pheatmap using clinical pathological stage and histological type as notes (\u003cstrong\u003eFigure 2B\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eThe pairwise comparison of various subtypes was performed using a t-test, and the results revealed that most MET loci had a low beta value. There was a significant difference in MET level among the majority of subtypes; the MET level in Cluster 2 was remarkably higher than that in the other 5 subtypes, while the MET levels in Cluster 4 and Cluster 5 were evidently lower than those in the other subtypes (\u003cstrong\u003eTable S3\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"4\"\u003e\n\u003cli\u003e\u003ca name=\"_Toc29772_WPSOffice_Level2\"\u003e\u003c/a\u003e\u003cstrong\u003e Analysis of the clinical characteristics of the 6 molecular subtypes\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eWe further analyzed the distribution of the 6 molecular subtypes with respect to prognosis, stage, grade, and age (\u003cstrong\u003eFigure 3)\u003c/strong\u003e. There was a significant difference in the prognosis among the 6 subtypes; the prognosis was best in Cluster 2, worst in Cluster 5, and poor in Cluster 4 (\u003cstrong\u003eFigure 3A\u003c/strong\u003e), indicating that the prognosis of hypomethylation subtypes was inferior to that of hypermethylation subtypes. The samples in Cluster 5 were all stage III, and the percentage of stage IV samples in Cluster 4 was significantly higher than that in the other subtypes (\u003cstrong\u003eFigure 3B\u003c/strong\u003e). All samples in Cluster 5 were G3 (\u003cstrong\u003eFigure 3C\u003c/strong\u003e, \u003cstrong\u003eTable S4)\u003c/strong\u003e, suggesting that the hypomethylation subtypes were mostly high-grade in clinical pathology. The age of the patients of samples in Cluster 5 was remarkably greater than that of the patients of the other subtypes, and the onset age was 70\u0026minus;80 years old (\u003cstrong\u003eTable S5\u003c/strong\u003e), while the mean age of the patients of samples in Cluster 2 was the lowest (\u003cstrong\u003eFigure 3D\u003c/strong\u003e), indicating that the age of patients with hypomethylation subtypes was generally higher than that of patients with hypermethylation subtypes. The above findings suggest, to a certain degree, that these DNA MET subtypes could be used to predict the prognosis, tumor stage, and pathological grade in ovarian cancer patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003col start=\"5\"\u003e\n\u003cli\u003e\u003ca name=\"_Toc26470_WPSOffice_Level2\"\u003e\u003c/a\u003e\u003cstrong\u003e Gene annotation and function analysis of the 250\u003c/strong\u003e\u003cstrong\u003e MET loci\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThere was a total of 285 genes corresponding to the gene promoter regions annotated by 250 CpG loci, and these genes were subjected to transcription factor enrichment analysis using the online tool g:profiler. It was found that 42 genes were significantly enriched to Transcription Factor EC (TFEC) (logrank \u003cem\u003eP\u003c/em\u003e value = 0.0107) (Figure 4A). At present, the role of TFEC in cancer progression has been studied to a limited extent; thus, to further explore the biological functions in which TFEC may be involved, TFEC-co-expressed molecules in the cBioPortal database were elucidated. Subsequently, the top 300 molecules with the most positive and negative correlations according to Spearman's correlation were selected. Functional enrichment analysis was performed using the DAVID 6.7 database and visualized using \u0026ldquo;GOplot\u0026rdquo; in the R software package. Finally, the top 5 biological processes with the most significant functions were chosen: GO:0006955 \u0026minus; immune response, GO:0050776 \u0026minus; regulation of immune response, GO:0006954 \u0026minus; inflammatory response, GO:0045087 \u0026minus; innate immune response, and GO:0007165 \u0026minus; signal transduction. TFEC may promote the occurrence and progression of ovarian cancer by influencing these biological functions(Figure 4B).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"6\"\u003e\n\u003cli\u003e\u003cstrong\u003e Screening 5 CpG loci by WGCNA co-expression analysis\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eUsing the WGCNA co-expression algorithm, 250 significant CpG loci were mined. Evaluation of the scale-free model was performed at different soft thresholds; a larger value and lower mean connectivity both indicated better compliance with the scale-free distribution. Finally, \u0026beta; = 4 (\u003cstrong\u003eFigure 5A, B\u003c/strong\u003e) was selected, and the settings of height = 0.25, deepSplit = 3, minModuleSize = 10 were chosen. A total of 7 modules were obtained (\u003cstrong\u003eFigure 5C\u003c/strong\u003e), of which the grey module is the set of genes that could not be clustered to other modules. The statistics of genes in various modules are shown in \u003cstrong\u003eTable 3\u003c/strong\u003e. The 250 CpG loci were assigned to 7 modules. Pearson\u0026rsquo;s correlation coefficient between the ME of each module and the characteristics of the samples was calculated; a higher correlation coefficient indicates that the module was more important. In \u003cstrong\u003eFigure 5D\u003c/strong\u003e, the row represents the eigengenes of each module and the column represents the feature information of the samples. The greatest correlation can be seen between the yellow module and Cluster 2 (R = 0.68, logrank \u003cem\u003eP\u003c/em\u003e = 7e-40), the brown module and Cluster 3 (R = 0.51, logrank \u003cem\u003eP\u003c/em\u003e = 1e-20), and the black module and Cluster 5 (R = 0.61, logrank \u003cem\u003eP\u003c/em\u003e = 6e-30).\u003c/p\u003e\n\u003cp\u003eSince Cluster 2 demonstrated the best prognosis among all the clusters, all CpG loci in the yellow module, mostly correlated with Cluster 2, were selected, and the interaction network was constructed according to their weighted relationship (\u003cstrong\u003eFigure 6A\u003c/strong\u003e). In this network, the CpG loci with a network centrality \u0026gt; 10 were cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. Furthermore, the expression relationships among the 22 CpG loci were calculated, and a significantly higher correlation was found among 8 (cg27625732, cg00431050, cg22197830, cg03152385, cg22809047, cg00328227, cg06851207, and cg01777397) (\u003cstrong\u003eFigure 6B\u003c/strong\u003e). Finally, a total of 5 CpG loci in the intersection were chosen, which had a strong correlation between each other and a centrality \u0026gt; 10 in the weighted network, as the characteristic MET loci of Cluster 2 samples (\u003cstrong\u003eTable 4\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"7\"\u003e\n\u003cli\u003e\u003cstrong\u003e Clustering analysis of the 5 CpG loci \u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eUnsupervised clustering analysis was further performed on the MET spectra of the 5 selected CpG loci, and the similarity between samples was calculated using the Euclidean distance. \u003cstrong\u003eFigure 7A \u003c/strong\u003eshows that the samples were divided into two groups according to the MET level of the 5 CpG loci: Cluster 1 (hypomethylation group) and Cluster 2 (hypermethylation group). The prognosis difference between these two groups was further analyzed (\u003cstrong\u003eFigure 7B\u003c/strong\u003e), indicating that the prognosis in the hypomethylation group was significantly poorer than that in the hypermethylation group.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003col start=\"8\"\u003e\n\u003cli\u003e\u003cstrong\u003e Model validation using the test dataset\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe MET spectra of the 5 CpG loci from the 286 samples in the test dataset were extracted and analyzed by hierarchical clustering (\u003cstrong\u003eFigure 8B\u003c/strong\u003e). The results show that the MET spectra of the 5 CpG loci were also obviously clustered into two groups: Cluster 1 and Cluster 2. The MET level of Cluster 1 samples was significantly higher than that of Cluster 2 samples. The prognosis difference between Cluster 1 and Cluster 2 (\u003cstrong\u003eFigure 8B)\u003c/strong\u003e was further analyzed, and it was found that the prognosis in the hypermethylation group was remarkably better than that in the hypomethylation group, which is consistent with the results of the training set.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003col start=\"9\"\u003e\n\u003cli\u003e\u003cstrong\u003e Flow chart of all the analysis\u003c/strong\u003e\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eThe flow chart of mining of subtype markers for the prognosis of ovarian cancer based on methylation data is shown in \u003cstrong\u003eFigure 9\u003c/strong\u003e. All the R package covered in this article is listed in\u003cstrong\u003e Table S6\u003c/strong\u003e.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn recent years, an increasing number of studies have focused on exploring the molecular typing of epithelial ovarian cancer to promote the realization of personalized treatment and improve the survival rate in patients; however, the achievements of molecular typing remain in the initial phase. Studies have shown that the occurrence of cancer is associated with genetic changes, and epigenetic abnormalities are also contributors. DNA MET is the major epigenetic modification mode of genomic DNA; it is an important means in regulating the functions of the genome \u003csup\u003e28\u003c/sup\u003e and is closely associated with the occurrence, progression, treatment, and prognosis of ovarian cancer. DNA MET-based molecular typing and subtype markers are of great significance for guiding personalized treatment and prognosis evaluation in ovarian cancer patients.\u003c/p\u003e\n\u003cp\u003eIn the present study, 571 ovarian cancer MET samples were downloaded from the TCGA database, 250 MET loci related to the prognosis of ovarian cancer patients were screened by COX regression analysis, and 6 molecular subtypes were selected by clustering with k-means. There was a significant difference in MET loci among most subtypes; the highest MET level and the best prognosis were observed in Cluster 2, and the MET level in Cluster 4 and Cluster 5 was remarkably lower than that in the other subtypes, accompanied by a very poor prognosis. This suggests, to a certain degree, that the prognosis of patients with a hypomethylation subtype was worse than that of patients with a hypermethylation subtype. All samples in Cluster 5 were high-grade, and the mean age of patients in Cluster 5 was higher than that in the other subtypes. The percentage of stage IV samples in Cluster 4 was significantly greater than that in the other subtypes. The above findings suggest that these molecular subtypes can be used not only to evaluate the prognosis in ovarian cancer patients, but also to fully distinguish the tumor stage, histological grade, and age of these patients to guide subsequent treatment.\u003c/p\u003e\n\u003cp\u003eDNA MET molecular typing also plays a very important role in the diagnosis, treatment, and prognosis of other tumors. Zhang et al. \u003csup\u003e25\u003c/sup\u003e screened 9 molecular subtypes by clustering analysis on DNA MET data in 669 breast cancer patients, and the DNA MET mode was reflected in varying races, ages, tumor stages, subject states, histological types, metastatic states, and prognoses. In comparison with PAM50 subtypes using gene expression clustering, DNA MET subtypes are more precise and can be used for the precision treatment of specific histological subtypes of breast cancer.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Jurmeister%20P%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=30723296\"\u003eJurmeister \u003c/a\u003econstructed a DNA MET map using the whole genome MET data from 600 cases of primary pulmonary, colorectal, and upper gastrointestinal adenocarcinoma, and successfully distinguished between pulmonary enteric\u0026nbsp;adenocarcinoma and metastatic colorectal cancer \u003csup\u003e29\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Williams%20LA%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=30287918\"\u003eWilliams \u003c/a\u003eet al.\u003csup\u003e30\u003c/sup\u003emeasured the MET level in different histological subtypes of 154 cases of child germ cell tumors using the Illumina Infinium\u0026reg; Human Methylation 450K chip, identifying 4 molecular subtypes. The MET level in the germ cell tumors was low, and these molecular subtypes provided information regarding their etiology.\u003c/p\u003e\n\u003cp\u003eAgain using the Illumina Infinium\u0026reg; Human Methylation 450K chip, \u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Wu%20SP%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=29354796\"\u003eWu SP\u003c/a\u003e1 et al. \u003csup\u003e31\u003c/sup\u003e detected the DNA MET state in 482 and 421 CpG loci in 10 samples of Ewing's sarcoma, 11 samples of synovial sarcoma, and 15 samples of osteosarcoma. Moreover, they developed and validated a whole-genome DNA MET classifier to identify osteosarcoma, Ewing's sarcoma, and synovial sarcoma. MET-based molecular typing is of great significance for diagnosing, recognizing, and treating morphologically overlapping solid tumors.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Taskesen%20E%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=25734246\"\u003eTaskesen E\u003c/a\u003e et al. integrated the gene expression and DNA MET spectra of 344 samples of acute myeloid leukemia (AML) and established a regression model using Lasso. The results indicated that the subtype prediction of AML cytogenetics and molecular abnormalities could be significantly improved \u003csup\u003e32\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Rodr%C3%ADguez-Rodero%20S%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=23666970\"\u003eA study by Rodr\u0026iacute;guez-Rodero \u003c/a\u003eet al., demonstrated that thyroid carcinoma subtypes have promoter-differentiated MET features, and the molecular typing could be realized using abnormal DNA MET expression. Undifferentiated thyroid carcinoma was characterized by abnormal promoter hypomethylation, while differentiated papillary and follicular thyroid carcinoma was characterized by promoter hypermethylation \u003csup\u003e33\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eTo further explore the functions of the 250 screened MET loci, gene function annotation of the loci was performed and 42 genes were found to be significantly enriched to TFEC. The \u003cem\u003eTFEC\u003c/em\u003e gene is located at 7q31.2 and encodes a polypeptide with a length of 347 amino acids, which is mainly localized in the nucleus and cytoplasm. According to a study by \u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Chung%20MC%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=11467950\"\u003eChung et al.\u003c/a\u003e, TFEC plays a role as an activating transcription factor (ATF) for the non-myosin\u0026nbsp;heavy chain II-a gene \u003csup\u003e34\u003c/sup\u003e. At present, evidence for the involvement of TFEC in cancer progression is limited; however, TFEC, MITF, TFEB, and TFE3 are important members of the MIT (microphthalmia-associated transcription factor) family, and recent studies have proven that changes in these transcription factors are related to melanoma, sarcoma, and renal cell carcinoma. With a similar structure to TFEB (another member of the MIT family), TFEC may play an important role in regulating genes related to autophagy and lysosomes \u003csup\u003e35\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eThe regulation of genes is a complex network; to investigate the effects of TFFC and its relevant factors on the occurrence and progression of tumors, function enrichment analysis was performed and these genes were found to be remarkably enriched to the following biological functions: GO:0006955 \u0026minus; immune response, GO:0050776 \u0026minus; regulation of the immune response, GO:0006954 \u0026minus; inflammatory response, GO:0045087 \u0026minus; innate immune response, and GO:0007165 \u0026minus; signal transduction. Currently, there are no reports of TFEC in ovarian cancer; thus, further investigation is needed.\u003c/p\u003e\n\u003cp\u003eFinally, 5 CpG loci were screened via the WGCNA co-expression network: cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. The results show that hypomethylation of these 5 CpG loci was associated with poor prognosis in ovarian cancer patients. The gene annotated by the cg22809047 locus was RPL31, and \u003ca href=\"https://www.ncbi.nlm.nih.gov/pubmed/?term=Maruyama%20Y%5BAuthor%5D\u0026amp;cauthor=true\u0026amp;cauthor_uid=25285958\"\u003eMaruyama \u003c/a\u003eet al. \u003csup\u003e36\u003c/sup\u003e have previously shown that in comparison with benign prostate tissues, RPL31 is overexpressed in prostate cancer. In RPL31 siRNA-treated LNCaP and BicR cells, there is an increase in the protein expression levels of the tumor suppressor p53 and its targets, p21 and MDM2. In addition, the inhibition of cell growth and the cell cycle by RPL31 could be recovered by p53 siRNA treatment. RPL31 could be used as the target of molecular treatment for advanced prostate cancer, and we presume that RPL31 could also be used as a target for the treatment of ovarian cancer. ELOVL3 was the gene corresponding to the gene promoter region annotated by the cg00431050 locus. ELOVL3 is a member of the ELOVL (elongase of very long-chain fatty acids) family, which contains a total of 7 members (ELOVL1\u0026minus;7). The proteins encoded by the \u003cem\u003eELOVL1\u003c/em\u003e\u003cstrong\u003e\u003cem\u003e\u0026minus;\u003c/em\u003e\u003c/strong\u003e\u003cem\u003e7\u003c/em\u003e genes are involved in the elongation of fatty acid chains of different lengths, and play an important role in regulating the biological synthesis of lipids, fatty acid metabolism, and certain metabolic diseases. There exist only limited studies of the involvement of ELOVL3 in tumors, while ELOVL2 has been widely described in tumors. A study by Kang et al., revealed that breast cancer patients with low ELOVL2 expression have a poor prognosis. ELOVL2 expression has been correlated with the malignant phenotype of breast cancer, and its downregulation induced lipid metabolism reprogramming; thus, ELOVL2 is a novel prognostic biomarker \u003csup\u003e37\u003c/sup\u003e. We suggest that ELOVL3 expression may also be involved in the occurrence and progression of ovarian cancer by inducing lipid metabolism reprogramming.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp; Zhang et al., \u003csup\u003e38\u003c/sup\u003e investigated the molecular typing of serous ovarian cancer using the multi-omics data of DNA MET and protein, miRNA, and gene expression, mainly discussing the relationship between molecular typing based on RNA-Seq data and that based on other omics data. They finally screened 9 molecular subtypes based on RNA-Seq data; these subtypes had significant overlap with the molecular subtypes of other omics, but the function analysis results showed that the subtypes based on an omics dataset could not be completely substituted by other omics data.\u003c/p\u003e\n\u003cp\u003eIn the present study, the significance of MET in the molecular typing of ovarian cancer was analyzed using MET data, and the markers of subtypes closely related to the prognosis prediction of ovarian cancer were further screened. A MET data-based ovarian cancer prognosis prediction model was subsequently developed to provide a reference for clinical trials and researchers. In summary, the study by Zhang et al., and our study have different focal points, despite both involving molecular typing.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSubtyping of ovarian carcinomas based on methylation profiles has been reported in a TCGA seminal article\u003csup\u003e13\u003c/sup\u003e, in which 4 subtypes were identified to be significantly associated with differences in age, BRCA inactivation events, and survival based on consensus clustering of variable DNA methylation. The cluster associated with the worst prognosis is characterized by hypomethylation and is associated with old age, which is in accordance with the present findings; however, our approach is different from that in the aforementioned TCGA paper.\u003c/p\u003e\n\u003cp\u003eFirstly, the samples included in the TCGA paper were 489 cases of high-grade serous ovarian cancer, while the present paper included 571 cases of methylated ovarian cancer, including different clinical stages and grades. Our sample size is larger, and the results are more abundant. Secondly, a multivariate COX proportional hazards model was performed to elucidate that 250 CpG loci were significant predictors of prognosis, and 6 molecular subtypes were clustered based on the methylation level at these 250 CpG loci. The cluster that was characterized by hypomethylation was associated with a worse prognosis, stage, and grade, and an older patient age. Thirdly, weighted gene co-expression network analysis was further applied to identify the 5 most significant CpG loci, and hypomethylation of these 5 loci was demonstrated to be associated with a worse outcome.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eWe identified 6 different molecular subtypes using ovarian cancer MET data in the TCGA database and showed that DNA MET molecular typing could accurately support the distinction of tumor stage and pathological grade in ovarian cancer. The specific CpG loci and genes can be used as biomarkers for individualized treatment, which can be carried out in clinical practice according to these prognostic markers. Nevertheless, the present study has certain limitations: only internal validation was performed on MET prognostic loci and no suitable external datasets were obtained; thus, our study results need to be further validated with a larger sample size.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003e\u003cstrong\u003eTCGA\u003c/strong\u003e: the Cancer Genome Atlas\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOS\u003c/strong\u003e: overall survival\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data during this study are included within this published article and additional files. Any material described in the article can be requested directly from corresponding author on reasonable request. All data in the current study are based on public data available in The Cancer Genome Atlas (TCGA) datasets.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing financial interests\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNon-Financial Conflict of Interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no non-financial conflict of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Sources\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eQing Yang designed and conceptualized this study; Lili Yin and Ningning Zhang were major contributors in experiment; Lili Yin analyzed the data; Lili Yin, and Qing Yang were major contributors in writing the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eJemal A, Siegel R, Ward E, et al. Cancer statistics, 2006. \u003cem\u003eCA: a cancer journal for clinicians. \u003c/em\u003e2006;56(2):106-130.\u003c/li\u003e\n\u003cli\u003eHerzog TJ. Recurrent ovarian cancer: how important is it to treat to disease progression? \u003cem\u003eClinical cancer research : an official journal of the American Association for Cancer Research. \u003c/em\u003e2004;10(22):7439-7449.\u003c/li\u003e\n\u003cli\u003ePearce CL, Templeman C, Rossing MA, et al. Association between endometriosis and risk of histological subtypes of ovarian cancer: a pooled analysis of case-control studies. \u003cem\u003eThe Lancet Oncology. \u003c/em\u003e2012;13(4):385-394.\u003c/li\u003e\n\u003cli\u003eSung P-L, Chang Y-H, Chao K-C, Chuang C-M. Global distribution pattern of histological subtypes of epithelial ovarian cancer: a database analysis and systematic review. \u003cem\u003eGynecologic oncology. \u003c/em\u003e2014;133(2):147-154.\u003c/li\u003e\n\u003cli\u003eB\u0026ouml;cker W. [WHO classification of breast tumors and tumors of the female genital organs: pathology and genetics]. \u003cem\u003eVerh Dtsch Ges Pathol. \u003c/em\u003e2002;86:116-119.\u003c/li\u003e\n\u003cli\u003eKommoss S, Schmidt D, Kommoss F, et al. Histological grading in a large series of advanced stage ovarian carcinomas by three widely used grading systems: consistent lack of prognostic significance. A translational research subprotocol of a prospective randomized phase III study (AGO-OVAR 3 protocol). \u003cem\u003eVirchows Archiv : an international journal of pathology. \u003c/em\u003e2009;454(3):249-256.\u003c/li\u003e\n\u003cli\u003eShih I-M, Kurman RJ. Ovarian tumorigenesis: a proposed model based on morphological and molecular genetic analysis. \u003cem\u003eThe American journal of pathology. \u003c/em\u003e2004;164(5):1511-1518.\u003c/li\u003e\n\u003cli\u003eMiller JA, Kotecha R, Ahluwalia MS, et al. Overall survival and the response to radiotherapy among molecular subtypes of breast cancer brain metastases treated with targeted therapies. \u003cem\u003eCancer. \u003c/em\u003e2017;123(12):2283-2293.\u003c/li\u003e\n\u003cli\u003eRoode SC, Rotroff D, Avery AC, et al. Genome-wide assessment of recurrent genomic imbalances in canine leukemia identifies evolutionarily conserved regions for subtype differentiation. \u003cem\u003eChromosome research : an international journal on the molecular, supramolecular and evolutionary aspects of chromosome biology. \u003c/em\u003e2015;23(4):681-708.\u003c/li\u003e\n\u003cli\u003eDolly SO, Collins DC, Sundar R, Popat S, Yap TA. Advances in the Development of Molecularly Targeted Agents in Non-Small-Cell Lung Cancer. \u003cem\u003eDrugs. \u003c/em\u003e2017;77(8):813-827.\u003c/li\u003e\n\u003cli\u003eXu Y, Chen M, Liu C, et al. Association Study Confirmed Three Breast Cancer-Specific Molecular Subtype-Associated Susceptibility Loci in Chinese Han Women. \u003cem\u003eThe oncologist. \u003c/em\u003e2017;22(8):890-894.\u003c/li\u003e\n\u003cli\u003eTothill RW, Tinker AV, George J, et al. Novel molecular subtypes of serous and endometrioid ovarian cancer linked to clinical outcome. \u003cem\u003eClinical cancer research : an official journal of the American Association for Cancer Research. \u003c/em\u003e2008;14(16):5198-5208.\u003c/li\u003e\n\u003cli\u003eIntegrated genomic analyses of ovarian carcinoma. \u003cem\u003eNature. \u003c/em\u003e2011;474(7353):609-615.\u003c/li\u003e\n\u003cli\u003eKommoss S, Winterhoff B, Oberg AL, et al. Bevacizumab May Differentially Improve Ovarian Cancer Outcome in Patients with Proliferative and Mesenchymal Molecular Subtypes. \u003cem\u003eClinical cancer research : an official journal of the American Association for Cancer Research. \u003c/em\u003e2017;23(14):3794-3801.\u003c/li\u003e\n\u003cli\u003eUshijima T, Asada K. Aberrant DNA methylation in contrast with mutations. \u003cem\u003eCancer science. \u003c/em\u003e2010;101(2):300-305.\u003c/li\u003e\n\u003cli\u003eWilting RH, Dannenberg J-H. Epigenetic mechanisms in tumorigenesis, tumor cell heterogeneity and drug resistance. \u003cem\u003eDrug resistance updates : reviews and commentaries in antimicrobial and anticancer chemotherapy. \u003c/em\u003e2012;15(1-2):21-38.\u003c/li\u003e\n\u003cli\u003eEsteller M, Silva JM, Dominguez G, et al. Promoter hypermethylation and BRCA1 inactivation in sporadic breast and ovarian tumors. \u003cem\u003eJournal of the National Cancer Institute. \u003c/em\u003e2000;92(7):564-569.\u003c/li\u003e\n\u003cli\u003eBaylin SB, Ohm JE. Epigenetic gene silencing in cancer - a mechanism for early oncogenic pathway addiction? \u003cem\u003eNature reviews Cancer. \u003c/em\u003e2006;6(2):107-116.\u003c/li\u003e\n\u003cli\u003eEsteller M. Epigenetics in cancer. \u003cem\u003eThe New England journal of medicine. \u003c/em\u003e2008;358(11):1148-1159.\u003c/li\u003e\n\u003cli\u003eBaylin SB, Chen WY. Aberrant gene silencing in tumor progression: implications for control of cancer. \u003cem\u003eCold Spring Harb Symp Quant Biol. \u003c/em\u003e2005;70:427-433.\u003c/li\u003e\n\u003cli\u003ePetrocca F, Iliopoulos D, Qin HR, et al. Alterations of the tumor suppressor gene ARLTS1 in ovarian cancer. \u003cem\u003eCancer research. \u003c/em\u003e2006;66(21):10287-10291.\u003c/li\u003e\n\u003cli\u003eHorak P, Pils D, Haller G, et al. Contribution of epigenetic silencing of tumor necrosis factor-related apoptosis inducing ligand receptor 1 (DR4) to TRAIL resistance and ovarian cancer. \u003cem\u003eMolecular cancer research : MCR. \u003c/em\u003e2005;3(6):335-343.\u003c/li\u003e\n\u003cli\u003eHu W-L, Zhou X-H. Identification of prognostic signature in cancer based on DNA methylation interaction network. \u003cem\u003eBMC medical genomics. \u003c/em\u003e2017;10(Suppl 4):63.\u003c/li\u003e\n\u003cli\u003eTomczak K, Czerwińska P, Wiznerowicz M. The Cancer Genome Atlas (TCGA): an immeasurable source of knowledge. \u003cem\u003eContemp Oncol (Pozn). \u003c/em\u003e2015;19(1A):A68-A77.\u003c/li\u003e\n\u003cli\u003eZhang S, Wang Y, Gu Y, et al. Specific breast cancer prognosis-subtype distinctions based on DNA methylation patterns. \u003cem\u003eMolecular oncology. \u003c/em\u003e2018;12(7):1047-1060.\u003c/li\u003e\n\u003cli\u003eWilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. \u003cem\u003eBioinformatics (Oxford, England). \u003c/em\u003e2010;26(12):1572-1573.\u003c/li\u003e\n\u003cli\u003eReimand J, Arak T, Adler P, et al. g:Profiler-a web server for functional interpretation of gene lists (2016 update). \u003cem\u003eNucleic acids research. \u003c/em\u003e2016;44(W1):W83-W89.\u003c/li\u003e\n\u003cli\u003eAran D, Hellman A. DNA methylation of transcriptional enhancers and cancer predisposition. \u003cem\u003eCell. \u003c/em\u003e2013;154(1):11-13.\u003c/li\u003e\n\u003cli\u003eJurmeister P, Sch\u0026ouml;ler A, Arnold A, et al. DNA methylation profiling reliably distinguishes pulmonary enteric adenocarcinoma from metastatic colorectal cancer. \u003cem\u003eModern pathology : an official journal of the United States and Canadian Academy of Pathology, Inc. \u003c/em\u003e2019;32(6):855-865.\u003c/li\u003e\n\u003cli\u003eWilliams LA, Mills L, Hooten AJ, et al. Differences in DNA methylation profiles by histologic subtype of paediatric germ cell tumours: a report from the Children's Oncology Group. \u003cem\u003eBritish journal of cancer. \u003c/em\u003e2018;119(7):864-872.\u003c/li\u003e\n\u003cli\u003eWu SP, Cooper BT, Bu F, et al. DNA Methylation-Based Classifier for Accurate Molecular Diagnosis of Bone Sarcomas. \u003cem\u003eJCO Precis Oncol. \u003c/em\u003e2017;2017.\u003c/li\u003e\n\u003cli\u003eTaskesen E, Babaei S, Reinders MMJ, de Ridder J. Integration of gene expression and DNA-methylation profiles improves molecular subtype classification in acute myeloid leukemia. \u003cem\u003eBMC bioinformatics. \u003c/em\u003e2015;16 Suppl 4:S5.\u003c/li\u003e\n\u003cli\u003eRodr\u0026iacute;guez-Rodero S, Fern\u0026aacute;ndez AF, Fern\u0026aacute;ndez-Morera JL, et al. DNA methylation signatures identify biologically distinct thyroid cancer subtypes. \u003cem\u003eJ Clin Endocrinol Metab. \u003c/em\u003e2013;98(7):2811-2821.\u003c/li\u003e\n\u003cli\u003eChung MC, Kim HK, Kawamoto S. TFEC can function as a transcriptional activator of the nonmuscle myosin II heavy chain-A gene in transfected cells. \u003cem\u003eBiochemistry. \u003c/em\u003e2001;40(30):8887-8897.\u003c/li\u003e\n\u003cli\u003eHaq R, Fisher DE. Biology and clinical relevance of the micropthalmia family of transcription factors in human cancer. \u003cem\u003eJournal of clinical oncology : official journal of the American Society of Clinical Oncology. \u003c/em\u003e2011;29(25):3474-3482.\u003c/li\u003e\n\u003cli\u003eMaruyama Y, Miyazaki T, Ikeda K, et al. Short hairpin RNA library-based functional screening identified ribosomal protein L31 that modulates prostate cancer cell growth via p53 pathway. \u003cem\u003ePloS one. \u003c/em\u003e2014;9(10):e108743.\u003c/li\u003e\n\u003cli\u003eKang YP, Yoon J-H, Long NP, et al. Spheroid-Induced Epithelial-Mesenchymal Transition Provokes Global Alterations of Breast Cancer Lipidome: A Multi-Layered Omics Analysis. \u003cem\u003eFrontiers in oncology. \u003c/em\u003e2019;9:145.\u003c/li\u003e\n\u003cli\u003eZhang Z, Huang K, Gu C, et al. Molecular Subtyping of Serous Ovarian Cancer Based on Multi-omics Data. \u003cem\u003eScientific reports. \u003c/em\u003e2016;6:26001.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp style=\"text-align: left; line-height: 200%;\"\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: 'Times New Roman',serif;\"\u003eTable 1: The clinical pathological information in the training set and validation test. \u003c/span\u003e\u003c/p\u003e\n\u003cdiv class=\"gridtable\" style=\"box-sizing: border-box; color: #212529; font-family: 'Source Sans Pro', sans-serif; font-size: 16px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: #ffffff; text-decoration-style: initial; text-decoration-color: initial;\"\u003e\n\u003ctable id=\"Tab1\" style=\"box-sizing: border-box; border-collapse: collapse;\" border=\"1\"\u003e\u003ccaption style=\"box-sizing: border-box; padding-top: 0.75rem; padding-bottom: 0.75rem; color: #6c757d; text-align: left; caption-side: bottom;\"\u003e\u0026nbsp;\u003c/caption\u003e\u003ccolgroup style=\"box-sizing: border-box;\"\u003e\u003c/colgroup\u003e\n\u003cthead style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eValidation\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eset Traning set\u003c/div\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e\u003cspan class=\"Bold\" style=\"box-sizing: border-box;\"\u003eStage\u003c/span\u003e\u003c/div\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eStage I\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e5\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e11\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eStage II\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e12\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eStage III\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e222\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e215\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eStage IV\u003c/div\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e\u003cspan class=\"Bold\" style=\"box-sizing: border-box;\"\u003eGrade\u003c/span\u003e\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e41\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e45\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eG1\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e3\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e3\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eG2\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e38\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e31\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eG3\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e238\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e243\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eG4\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e\u003cspan class=\"Bold\" style=\"box-sizing: border-box;\"\u003eAge\u003c/span\u003e\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e\u0026lt;=60\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e171\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e147\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e\u0026gt;\u0026thinsp;60\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e115\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e138\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp style=\"text-align: left; line-height: 200%;\"\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: 'Times New Roman',serif;\"\u003eTable 2: The top 20 loci with the most significant difference in prognosis. \u003c/span\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\" style=\"box-sizing: border-box; color: #212529; font-family: 'Source Sans Pro', sans-serif; font-size: 16px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: #ffffff; text-decoration-style: initial; text-decoration-color: initial;\"\u003e\n\u003ctable id=\"Tab2\" style=\"box-sizing: border-box; border-collapse: collapse;\" border=\"1\"\u003e\u003ccaption style=\"box-sizing: border-box; padding-top: 0.75rem; padding-bottom: 0.75rem; color: #6c757d; text-align: left; caption-side: bottom;\"\u003e\u0026nbsp;\u003c/caption\u003e\u003ccolgroup style=\"box-sizing: border-box;\"\u003e\u003c/colgroup\u003e\n\u003cthead style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eCpGs\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ep.value\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eHR\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eLow 95%CI\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eHigh 95%CI\u003c/div\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg25781123\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.42E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e144.7952\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15.31263\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1369.173\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg01278291\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.68E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.013765\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.001955\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.096903\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg21291896\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e2.93E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.24E\u0026thinsp;+\u0026thinsp;13\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e8962882\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.72E\u0026thinsp;+\u0026thinsp;19\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg08946332\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e5.46E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.21193\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.099747\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.450282\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg13804316\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e8.51E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e213979.4\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e469.6279\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e97496703\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg16179125\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e9.40E-05\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e8.255374\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e2.862145\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e23.81123\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg13060646\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000201\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e4.08882\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.945923\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e8.591523\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg03750606\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000282\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e31.81681\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e4.916032\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e205.92\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg15341340\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000317\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e3559149\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e966.7441\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.31E\u0026thinsp;+\u0026thinsp;10\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg08013810\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.00033\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.087427\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.023115\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.330674\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg06797533\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000383\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e4.351053\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.932616\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e9.795869\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg21022435\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000396\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.00496\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000263\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.093416\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg22916109\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000475\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.07E\u0026thinsp;+\u0026thinsp;11\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e69831.7\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.65E\u0026thinsp;+\u0026thinsp;17\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg10415235\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.00048\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e6.56E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e7370.126\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e5.83E\u0026thinsp;+\u0026thinsp;13\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg05955301\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000514\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e8.98125\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e2.60262\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e30.99294\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg16016036\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.00056\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.243586\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.109202\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.543343\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg23486067\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000569\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e20.19834\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e3.654797\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e111.6267\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg25634666\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.00057\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.255673\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.117688\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.555439\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg17332016\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000602\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e104162.1\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e141.5735\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e76636837\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg03190825\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e0.000654\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15.75275\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e3.227124\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e76.89485\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cp style=\"text-align: left; line-height: 200%;\"\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: 'Times New Roman',serif;\"\u003eTable 3: The CpG loci in different modules. \u003c/span\u003e\u003c/p\u003e\n\u003cdiv class=\"gridtable\" style=\"box-sizing: border-box; color: #212529; font-family: 'Source Sans Pro', sans-serif; font-size: 16px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: #ffffff; text-decoration-style: initial; text-decoration-color: initial;\"\u003e\n\u003ctable id=\"Tab3\" style=\"box-sizing: border-box; border-collapse: collapse;\" border=\"1\"\u003e\u003ccaption style=\"box-sizing: border-box; padding-top: 0.75rem; padding-bottom: 0.75rem; color: #6c757d; text-align: left; caption-side: bottom;\"\u003e\u0026nbsp;\u003c/caption\u003e\u003ccolgroup style=\"box-sizing: border-box;\"\u003e\u003c/colgroup\u003e\n\u003cthead style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eModule\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 50px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eCount\u003c/div\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eblack\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e12\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eblue\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e34\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ebrown\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e22\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003egreen\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e16\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ered\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e14\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eturquoise\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e49\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 70px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eyellow\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 50px;\" align=\"char\" char=\".\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e22\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\" style=\"box-sizing: border-box; color: #212529; font-family: 'Source Sans Pro', sans-serif; font-size: 16px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400; letter-spacing: normal; orphans: 2; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-stroke-width: 0px; background-color: #ffffff; text-decoration-style: initial; text-decoration-color: initial;\"\u003e\n\u003cp style=\"text-align: left; line-height: 200%;\"\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: 'Times New Roman',serif;\"\u003eTable 4\u003c/span\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: SimSun;\"\u003e：\u003c/span\u003e\u003cspan style=\"font-size: 12.0pt; line-height: 200%; font-family: 'Times New Roman',serif;\"\u003eThe annotation of 5 CpG loci.\u003c/span\u003e\u003c/p\u003e\n\u003cdiv class=\"gridtable\" style=\"box-sizing: border-box;\"\u003e\n\u003ctable id=\"Tab4\" style=\"box-sizing: border-box; border-collapse: collapse;\" border=\"1\"\u003e\u003ccaption style=\"box-sizing: border-box; padding-top: 0.75rem; padding-bottom: 0.75rem; color: #6c757d; text-align: left; caption-side: bottom;\"\u003e\u0026nbsp;\u003c/caption\u003e\u003ccolgroup style=\"box-sizing: border-box;\"\u003e\u003c/colgroup\u003e\n\u003cthead style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eCpG\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eChrom\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eStart\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eEnd\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eGeneSymbol\u003c/div\u003e\n\u003c/th\u003e\n\u003cth style=\"box-sizing: border-box; text-align: inherit; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eFeature_Type\u003c/div\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody style=\"box-sizing: border-box;\"\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg03152385\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr16\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15094739\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15094740\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eRP11-72I8.1\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eS_Shore\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg27625732\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr9\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.29E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.29E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eTBC1D13\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eN_Shore\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg22197830\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr5\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.35E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.35E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eTXNDC15\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eN_Shore\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg22809047\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr2\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.01E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.01E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eAC016738.4\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eIsland\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg03152385\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr16\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15094739\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e15094740\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eRRN3\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eS_Shore\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg00431050\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr10\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.02E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.02E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eELOVL3\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eN_Shore\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr style=\"box-sizing: border-box;\"\u003e\n\u003ctd style=\"box-sizing: border-box; width: 91px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003ecg22809047\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 55px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003echr2\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.01E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 79px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003e1.01E\u0026thinsp;+\u0026thinsp;08\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 99px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eRPL31\u003c/div\u003e\n\u003c/td\u003e\n\u003ctd style=\"box-sizing: border-box; width: 110px;\" align=\"left\"\u003e\n\u003cdiv class=\"SimplePara\" style=\"box-sizing: border-box;\"\u003eIsland\u003c/div\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003c/div\u003e"},{"header":"Supplementary File Information","content":"\u003cp\u003eTable S1: A total of 967 loci demonstrated a significant difference in prognosis\u003c/p\u003e\n\u003cp\u003eTable S2: 250 significant MET loci\u003c/p\u003e\n\u003cp\u003eTable S3: Distribution of methylation levels in each subtype\u003c/p\u003e\n\u003cp\u003eTable S4: The distribution of grade and stage in each subtype\u003c/p\u003e\n\u003cp\u003eTable S5: Age distribution of samples in each subtype\u003c/p\u003e\n\u003cp\u003eTable S6: The R package covered in this article\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Ovarian cancer, TCGA, methylation, molecular subtype, prognosis marker","lastPublishedDoi":"10.21203/rs.2.23311/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.2.23311/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAims: Ovarian cancer is one of three major malignancies involving the female reproductive system, and its morbidity and mortality are ranked number 3 and number 1 among gynecological tumors, respectively. DNA methylation (MET), as one of the main epigenetic modes, is closely related to the occurrence and development of ovarian cancer. To guide individualized treatment and improve the prognosis in ovarian cancer patients, it is of great significance to elucidate effective MET subtype markers. \u003c/p\u003e\u003cp\u003eMethods: A total of 571 ovarian cancer MET samples were downloaded from the Cancer Genome Atlas (TCGA), and a COX proportional hazards model was established using the MET spectrum and clinically pathological parameters. Subsequently, the consensus clustering of CpG loci with a significant difference in both univariate and multivariate analyses was performed to screen the molecular subtypes, and these CpG loci were subjected to gene function annotation. Finally, CpG MET loci associated with poor prognosis in ovarian cancer patients were further screened by constructing a weighted gene co-expression network analysis (WGCNA). \u003c/p\u003e\u003cp\u003eResults: A total of 250 prognosis-related MET loci were obtained by COX regression and 6 molecular subtypes were screened by clustering. There was a remarkable MET difference between most subtypes, of which Cluster 2 had the highest MET level and demonstrated the best prognosis in patients, while Cluster 4 and Cluster 5 had a MET level significantly lower than that of the other subtypes and demonstrated a very poor prognosis. All Cluster 5 samples were at a high grade, while the percentage of Stage IV samples in Cluster 4 was evidently greater than that in the other subtypes. Using the co-expression network, 5 CpG loci were eventually obtained: cg27625732, cg00431050, cg22197830, cg03152385, and cg22809047. The clustering analysis shows that the prognosis in patients with hypomethylation was significantly worse than that in patients with hypermethylation. \u003c/p\u003e\u003cp\u003eConclusions: These MET molecular subtypes can be used not only to evaluate the prognosis in ovarian cancer patients but also to fully distinguish the tumor stage and histological grade in these patients. Prognosis-related CpG loci can be applied as biomarkers for individualized treatment in ovarian cancer patients.\u003c/p\u003e","manuscriptTitle":"Mining of Subtype Markers for the Prognosis of Ovarian Cancer based on Methylation Data","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2020-02-21 18:13:19","doi":"10.21203/rs.2.23311/v2","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}},{"code":1,"date":"2020-02-12 17:22:54","doi":"10.21203/rs.2.23311/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a50b7eaf-978a-40e5-85f4-33c89a3debdf","owner":[],"postedDate":"February 21st, 2020","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":61305,"name":"Cancer Biology"},{"id":61306,"name":"Oncology"}],"tags":[],"updatedAt":"","versionOfRecord":[],"versionCreatedAt":"2020-02-21 18:13:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-13880","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"identity":"rs-13880","version":["v2"]},"buildId":"cBFmMYwuxLRRLfASyISRj","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}