| Objective:Ovarian cancer is one of the most lethal malignant gynecological cancers worldwide,causing 180,000 women deaths each year.Although the five-year survival rate of malignant tumors in China has increased from 30.9% 10 years ago to40.5%,ovarian cancer has only increased by 0.4%.This is due to the lack of effective biomarkers for the early diagnosis of ovarian cancer.CA125,the most widely used blood marker for ovarian cancer in clinical use,lacks sufficient diagnostic specificity.Circulating tumor DNA(ctDNA)was fragmented DNA released from tumor cells’ apoptosis,necrosis,and metabolism,whose features reflect the genetic variation or epigenetic modification of the tumor.It is a very important part of the Circulating cell-free DNA(cfDNA).Methylation is an early epigenetic modification in the development of tumors,which can reflect the pathological and physiological state of the body in real time.Recently,cfDNA methylation detection has gradually become a hot spot in liquid biopsy due to its non-invasive,real-time,high sensitivity,and good specificity.This study aims to screen out candidate methylation markers of cfDNA for detection of ovarian cancer through cfDNA high-throughput sequencing and bioinformatics analysis of plasma samples from patients with ovarian cancer and normal women,and to further construct a methylation diagnosis model of ovarian cancer,followed by the validation of these markers by digital PCR,providing a reference for the early diagnosis of ovarian cancer.Materials and Methods:1.Determination of the differential methylation profiles of cfDNA in ovarian cancer.Firstly,the TruSeq Methyl Capture EPIC high-throughput methylation sequencing was performed on 11 ovarian cancer mixed-sample pools and 10 normal female cfDNA mixed-sample pools,using the "mixed-pool sequencing" strategy,and potential sites were screened according to the P values and Dif methylation differences.Then,the Customized methylated probes were designed for these differential sites,and verified in a customized EPIC cohort of 754 ovarian cancer and 1118 normal female control samples by the Customized EPIC high-throughput methylated sequencing.2.Establishing the EPIC methylation diagnostic model and EPIC methylation comprehensive scoring system for ovarian cancer.Firstly,the 1872 samples of Customized EPIC cohort were divided into the training set and the test set in a ratio of 2:1.In the training set,the Least Absolute Shrinkage and Selection Operator(LASSO)and Support Vector Machine Recursive Feature Elimination(SVM-RFE)algorithms were used for feature screening,and 11 methylation sites that overlapped by the two algorithms were selected to build a logistic regression model.Receiver operating characteristic curves(ROC)in the training set and the test set were drawn to evaluate the model.The EPIC methylation comprehensive scoring system for ovarian cancer was constructed using the unbiased coefficient and the corresponding methylation rate.The sensitivity and specificity were evaluated.Then the sensitivity of the model was compared with that of CA125 and HE4.After that,the predictive efficiency of the model for different stages of tumor development was evaluated,and the correlations between CD score,CA125,and HE4 and tumor development stages were evaluated.3.Verification of the EPIC methylation diagnostic model for ovarian cancer by droplet digital PCR(ddPCR).In a ddPCR cohort containing 218 ovarian cancer cases and 168 normal female cfDNA,the methylation rates of the 11 methylation sites in the EPIC methylation diagnostic model were at first detected by the ddPCR technique to verify their diagnostic potential.Then,it was substituted into the EPIC methylation diagnostic model to evaluate the applicability of this model for the ddPCR detection method,and the predictive efficiency of this model for different tumor development stages was assessed.4.Establishing the ddPCR methylation diagnostic model for ovarian cancer based on ddPCR.The methylation rates measured by ddPCR were used to refit a logistic regression model and the corresponding ddPCR comprehensive diagnostic scoring system for ovarian cancer.The diagnostic efficiency of the model was evaluated by the ROC curve and predictive confusion matrix,and the diagnostic potential of the model for different stages of tumor development was analyzed.Subsequently,the sensitivity of the EPIC methylation diagnostic model,ddPCR methylation diagnostic model,CA125,and HE4 for ovarian cancer diagnosis was compared,and the correlations between them and tumor development stages were evaluated.Finally,a prospective cohort containing 20 early-stage ovarian cancer patients was used to validate the ddPCR diagnostic model for ovarian cancer.Results:1.Through Truseq Methyl Capture EPIC high-throughput sequencing using the 21 ovarian cancer and normal female cfDNA "mixing pools",2757 potential differential methylation sites were screened out.Then we successfully identified165 differential methylation sites with the single sample-based high-throughput methylation sequencing using the customized EPIC cohort,which were visualized by hierarchical cluster analysis.2.In the training set of the customized EPIC cohort,11 diagnostic methylation markers of methylation were selected by feature selection,and EPIC methylation diagnostic model of ovarian cancer was established based on these variables.The model was proved with a specificity of 94.1% and a sensitivity of 84.4% in the training set,and a specificity of 94.8% and a sensitivity of 83.4% in the test set.Meanwhile,the Area under the predicted ROC curve(AUROC)in the training set and test set were 0.952 and 0.949,respectively,showing strong robustness.Importantly,the diagnostic sensitivity of the model was also higher than 70%(74.1%)in patients with early-stage ovarian cancer.Further analysis showed that the diagnostic sensitivity of the model in ovarian cancer patients(84.90%)was much higher than that of the clinical blood biomarkers CA125 and HE4(both less than50%),and the combination of them(60.56%).Moreover,the EPIC methylation comprehensive score--cdscore was significantly correlated with tumor development stages in ovarian cancer.3.In the ddPCR cohort,all of the 11 methylation diagnostic sites showed good differentiation potential between ovarian cancer and normal female control samples,of which OV1 showing the best predictive potential,which was consistent with the EPIC cohort.However,when the methylation rates measured by ddPCR were applied to the EPIC methylation diagnostic model,the ROC curve and diagnostic sensitivity were only slightly improved,suggesting that the EPIC methylation diagnostic model may not be applicable to the ddPCR cohort.4.The methylation rates of the 11 methylation sites detected by ddPCR were then used to construct a ddPCR diagnostic model of ovarian cancer and the corresponding ddPCR comprehensive diagnostic scoring system.The specificity and sensitivity of this model were 90.48% and 84.86% in 218 cases of ovarian cancer and 168 normal female control cfDNA samples,and the area under the ROC curve reached 0.918.The diagnostic efficiency was close to that of the customized EPIC cohort based on customized EPIC sequencing,and the sensitivity was much higher than that of the EPIC methylation diagnostic model detected by ddPCR and clinical blood biomarkers CA125 and HE4.In addition,in the ddPCR cohort,ovarian cancer EPIC methylation diagnostic scoring system--cdscore,ovarian cancer ddPCR methylation diagnostic scoring system--cdscore,CA125,and HE4 all showed significantly correlated with tumor development stage,which may be related to the sample size and sample composition.Finally,in a prospective ddPCR cohort of 20 ovarian cancer samples,the ddPCR methylation diagnostic model for ovarian cancer successfully predicted 17 early-stage cancer samples,including one benign control.Conclusion:1.Differential analysis of plasma free DNA methylation based on the "mixed-sample sequencing" strategy can be used to develop diagnostic markers for ovarian cancer.In this study,2757 candidate differential methylation sites were screened by high-throughput sequencing of cfDNA samples from ovarian cancer and normal female,and 165 of them were successfully verified by the single cfDNA sample-based high-throughput sequencing using the customized EPIC probes.Subsequent analysis showed the feasibility of this method in the development of diagnostic markers for cfDNA methylation.2.The high-throughput methylation sequencing-based EPIC methylation diagnostic model and the corresponding EPIC methylation comprehensive scoring system for ovarian cancer were successfully constructed.The diagnostic sensitivity of this model is much higher than the current clinical blood markers CA125 and HE4,and it also has high sensitivity in patients with early ovarian cancer,which may be used for early screening of high-risk patients with ovarian cancer.The EPIC methylation comprehensive score system has an advantage over CA125 and HE4 in predicting different stages of tumor development.3.The ddPCR detection-based ddPCR methylation diagnostic model and the corresponding ddPCR methylation comprehensive scoring system of ovarian cancer were successfully established.When ddPCR was used to detect the cfDNA methylation rate,the predictive sensitivity of this model in ovarian cancer patients is much higher than that of CA125 and HE4,and it is a new promising method that is more efficient,convenient and cheap than cfDNA high-throughput sequencing. |
PDF Full Text Request