Improved Tumor-Type informed compared to Tumor-Informed Mutation Tracking for ctDNA Detection and Microscopic Residual Disease Assessment in Epithelial Ovarian Cancer

Improved Tumor-Type informed compared to Tumor-Informed Mutation Tracking for ctDNA Detection and Microscopic Residual Disease Assessment in Epithelial Ovarian Cancer | 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 Research Article Improved Tumor-Type informed compared to Tumor-Informed Mutation Tracking for ctDNA Detection and Microscopic Residual Disease Assessment in Epithelial Ovarian Cancer Mehdi Ben Sassi, Henri Azais, Charles Marcaillou, Sylvain Guibert, and 11 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6031886/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 12 Jun, 2025 Read the published version in Journal of Experimental & Clinical Cancer Research → Version 1 posted 9 You are reading this latest preprint version Abstract Background Epithelial ovarian cancer (EOC) is a leading cause of cancer mortality in women, often diagnosed at advanced stages. While first-line treatments improve survival, relapses remain common, with 5-year survival rates below 40%. Circulating tumor DNA (ctDNA) is a promising biomarker for non-invasive EOC detection and monitoring. It may help assess treatment response, notably microscopic residual diseases. Our objective was to compare two ctDNA characterization strategies in EOC for assessing tumor burden during first-line treatment: a tumor-informed approach based on somatic mutations and a tumor-type informed approach utilizing DNA methylation patterns. Methods In the tumor-informed approach, whole exome sequencing (WES) was performed on EOC tumor DNA and matched PBMCs from 12 patients aiming to identify tumor-specific mutations. Custom panels were designed to target patient’s specific tumor mutations, which were then tracked in cfDNA from plasma samples. In the tumor-type informed approach, differentially methylated loci (DMLs) were identified by comparing EOC samples, healthy ovarian tissues, and PBMCs. A unique custom methylation panel was designed, and a support vector machine classifier was trained to distinguish between healthy and cancerous plasma samples. Plasma samples were collected from 47 advanced-stage EOC patients during neoadjuvant chemotherapy, alongside plasma from healthy subjects. Results For the tumor-informed approach, WES identified an average of 74 somatic mutations per patient. CtDNA was detected in 11 out of 12 patients at baseline (mean VAF: 1.29%). For the tumor-type informed approach, 52,173 DMLs were identified. The classifier trained on these DMLs detected ctDNA in baseline plasma samples for 11 out of the 12 patients demonstrating equivalent sensitivity (mean VAF: 1.17%. In end-of-treatment samples, the tumor-type-informed approach detected ctDNA twice as often as the tumor-informed method. Detection using this more sensitive approach correlated with relapse and shorter progression-free survival (log-rank p = 0.017, Hazard ratio = 8.24; 95% CI [1.06–64.4]) and was associated with poorer overall survival (log-rank p = 0.036). Conclusion The tumor-type informed classifier demonstrated sensitivity and specificity for ctDNA detection, outperforming the tumor-informed approach in monitoring EOC progression. Requiring fewer sequencing data, it offers a practical, efficient solution for clinical management of EOC. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1 Introduction Epithelial ovarian cancer (EOC) is the eighth most common cancer in women worldwide ( 1 ). Due to unspecific symptoms, patients are often diagnosed at advanced stage ( 2 ). According to the International Federation of Gynecology and Obstetrics recommendations, the first-line treatment combines chemotherapy and cytoreductive surgery. Bevacizumab or Poly-ADP-ribose-polymerase inhibitors are commonly employed as maintenance treatments ( 3 ). After surgery, the absence of macroscopic residual disease is the major prognostic factor for disease curing ( 4 ). Despite first-line treatment and maintenance therapy, the 5-year survival rate remains below 40% with frequent relapses ( 5 ). Current follow-up methods, such as imaging and CA125 blood monitoring lack of sensitivity, hampering early recurrence detection and therapeutic response evaluation ( 6 ). Indeed, assessing microscopic residual disease (MRD) after treatment could improve prognostication and therapeutic management of patients ( 7 ). A reliable, non-invasive and sensitive monitoring approach of MRD is critically needed. Circulating tumor DNA (ctDNA) shed by tumor cells into the bloodstream, offers a promising approach for both tumor burden assessment and characterization of tumor genomic alterations ( 8 ). It could improve the disease monitoring, including screening, treatment response assessment, MRD evaluation after surgery and early detection of recurrence ( 8 , 9 ). The ctDNA detection is achieved through either tumor-informed or agnostic strategies. The tumor-informed approach, first identifies patient-specific tumor alterations by analyzing the tumor tissue and then tracks these alterations in DNA extracted from plasma ( 10 ). This method offers high sensitivity but is resource intensive. In contrast, the tumor-agnostic approach bypasses the need for tumor tissue by leveraging knowledge of common cancer alterations to create a standardized, one-size-fits-all test ( 11 ). This method can employ broad genomic panels, such as whole-exome sequencing (WES) and whole-genome sequencing (WGS) or focus on specific genes using PCR based methods such as digital PCR. While this approach offers advantages in turnaround time and cost-effectiveness and has proven high sensitivity for therapeutic response analysis, it generally sacrifices sensitivity compared to the tumor-informed approach ( 11 ). In 2020, Zviran et al. modeled the probability of detecting ctDNA based on the number of genetic targets, the cfDNA input and the ctDNA fraction ( 8 ). Their study demonstrated a clear benefit of increasing the number of targets, with detection of ctDNA fractions as low as 10⁻ 5 achievable using tens of nanograms of cfDNA input when up to 10,000 targets are analyzed. In EOC, the average tumor mutation burden is between 2 and 3 mutations per mega base, enabling the identification of around 9,000 mutations through WGS ( 12 ). However, WGS is cost-prohibitive for many applications. WES, which reduces the volume of sequencing data, directly correlated to cost, identifies only a hundred mutations, significantly limiting the sensitivity of tumor-informed strategies for ctDNA detection around a ctDNA fraction of 10⁻⁴. We aimed to develop a method that focuses on common alterations specific to a particular cancer type. By incorporating a sufficient number of these alterations, this approach could achieve sensitivity comparable to tumor-informed methods based on WGS identification of alteration, while maintaining the versatility of a one-size-fits-all test for that specific cancer type. This can be completed by studying epigenetic marks specific to a tumor type. This method can be named as tumor-type informed and tumor-patient agnostic. DNA methylation offers an avenue for ctDNA detection ( 13 ). Recent advances identified distinct methylation signatures across cancer types, endeavor to develop ctDNA detection strategies ( 14 – 17 ). Tumor-type informed approach was applied in three key steps. First, we identified specific epigenetic markers of EOC tissue using enzymatic conversion of unmethylated cytosines ( 18 ), an enrichment panel, and next-generation sequencing (NGS). This approach enabled the identification of several thousand of differentially methylated CpGs within a reduced genomic window, distinguishing EOC tissue from healthy ovaries and peripheral blood mononuclear cells (PBMCs). Second, we evaluated the performance of these markers in a series of plasma samples from EOC patients and compared the results to a tumor-informed approach, which involved WES of solid tumors and mutation tracking in plasma samples. Finally, we assessed the potential of this method to detect MRD following the completion of first-line therapy. 2 Materials & Methods Sample Collection The patients included in this study derived from the BIOVAIRE cohort (Paris, France). Of the 67 patients of the BIOVAIRE cohort, 47 were included in this study. Twelve patients were excluded due to a non-EOC diagnosis confirmed after tissue biopsy analysis, and 8 were excluded because of unavailable baseline plasma sample. All participants provided informed consent, and the study received ethical approval (CPP n°: 2013-10-01 and ID-RCB: 2013-A01108-37). Tumor tissues were stored in RNAlater at -80°C, and blood samples were collected in Streck tubes and were processed for plasma and PBMC isolation. Normal ovarian tissues were obtained from OriGene (Herford, Germany) and classified as non-cancerous by two independent pathologists. Samples were flash-frozen in optimal cutting temperature compound (OCT) and stored at -80°C. Publicly available whole-genome bisulfite sequencing (WGBS) datasets of normal ovarian tissues were downloaded from the ENCODE database (Accession: ENCSR417YFD and ENCSR803SIO). Blood collected in Streck tubes from healthy women donors were provided by Biopredic International (UP, India) or l’Établissement Français du Sang (EFS, Evry, France). DNA from ovarian tissues and PBMC was extracted using the Qiagen DNeasy Blood & Tissue Kit. Enzymatic-targeted methylation protocol and DMR analysis Libraries were prepared with the NEBNext Enzymatic Methyl-seq kit with 100 ng of input DNA, followed by targeted hybrid-capture using the Twist Human Methylome Panel (Twist Bioscience). Sequencing was performed on an Illumina NovaSeq 6000 using 2x100 bp paired-end mode. Reads were processed using trim_galore (v. 0.6.6), BWAmeth (v. 0.2.7), Picard MarkDuplicates (v. 2.22.8), and methylation calling with MethylDackel (v. 0.6.0). CpG methylation status was compared between ovarian tumor and normal ovarian tissues, as well as between ovarian tumor and PBMCs from the same individuals. Differentially methylated loci (DMLs) were identified using DSS R package ( 19 ) and MethylKit R package ( 20 ), while differentially methylated regions (DMRs) were detected using DMRseq R package ( 21 ). A threshold of ≥ 30% methylation difference was applied, with significance criteria set at q-values < 0.001 for DMLs and < 0.01 for DMRs. CpG sites that were found differentially methylated in at least two of the three packages were selected for panel design. Subsequently, DMRs were redefined by clustering the selected differentially methylated CpGs, requiring a minimum of four CpGs per region, each separated by less than 100 bp. Regions within 50bp of each other were merged. For the panel design, DMRs overlapping repetitive or low-complexity regions were excluded. Targeted Methylation Sequencing of Plasma Samples CfDNA was extracted from 3 mL of plasma using the QIAamp Circulating Nucleic Acid Kit. For cfDNA quantification, 1 µL of the 30 µL elution volume was analyzed on a Fragment Analyzer (Agilent), using the NGS Fragment kit (1-6000 bp). Half of the remaining elution volume, corresponding to 1.1 to 25 ng (mean 8.8 ng, IQR 3.5–10.6) was processed using the NEBNext Enzymatic-Methyl-seq kit, with UMI methylated adapters (Twist Bioscience). Libraries were hybrid-captured with the custom-panel targeting identified CpG sites, sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode and aligned to human genome (hg38) with BWAmeth (v. 0.2.7) after data normalization to 1.5 Gb per sample. Duplicates reads were managed using fgbio (v. 2.3.0) to distinguish real duplicates from collisions (i.e., distinct fragments sharing the same genomic coordinates) thanks to molecular barcodes. Methylation patterns at the resolution of individual reads ( 22 ) were identified using wgbstools ( 23 ). A fragment is considered to originate from a tumor if at least 85% of the differentially methylated loci (DMLs) within the fragment share the same status as defined by the analysis of DNA extracted from EOCs. For each DMR, a beta value was computed, representing the proportion of plasma-derived fragments associated with a tumor-derived origin. The standard deviation of this beta value was then calculated for each DMR across a cohort of 24 healthy plasma donors. Only DMRs with a standard deviation ≤ 0.05 were retained. Subsequently, the area under the receiver operating characteristic curve (AUC) was determined for each DMR by comparing the beta values between the 24 healthy plasma donors and 35 plasma samples from patients at baseline, which were presumed to contain ctDNA. DMRs with an AUC ≥ 0.65 were selected for further analysis. The Support Vector Machine classifier was trained using beta values from selected DMRs as features. The training dataset consisted of the same 24 plasma samples from healthy donors and 35 baseline plasma samples used for DMR selection. Once trained, the model was used to classify new plasma samples based on their predicted values. A sample was classified as ctDNA-positive if its predicted value exceeded 0, while a predicted value below 0 indicated a ctDNA-negative sample. ctDNA proportion assessment with the tumor-type informed approach To assess the ctDNA proportion we selected DMRs with a background noise level below 1 in 10,000 reads across the 54 healthy plasma donors, resulting in a final selection of 201 DMRs. Then we calculated the ratio of sequencing reads exhibiting the EOC-specific methylation pattern to the total sequencing reads overlapping the genomic coordinates of the 201 DMRs in samples classified as ctDNA-positive DNA mixes used for ctDNA quantification validation in the tumor-type informed approach Twelve DNA extracted from ovarian tumor were combined with DNA extracted from matched PBMC at a 1:99 ratio. The tumor cellularity in the 12 tumor samples used for the mixes was determined by copy number alteration (CNA) analysis. CNA analysis was performed on sequencing data generated for the identification of genetic alterations (see below) using GATK (v. 4.1.4.1). The expected tumor DNA fraction in the 12 mixes was calculated using the formula: tumor cellularity divided by 100. These samples were used to validate ctDNA quantification methods in the tumor-type informed approach. DNA mixes were processed using the NEBNext Enzymatic-Methyl-seq kit, with UMI methylated adapters (Twist Bioscience) and 20 ng of fragmented DNA input. Libraries were hybrid-captured with the custom-panel targeting identified CpG sites, sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode and followed the same analysis as described for plasma samples. Identification of Genetic Alterations Tumor and matched PBMC DNA from 12 patients underwent WES using the Twist Human Core Exome Panel (Twist Bioscience) with an input of 100 ng. Libraries were sequenced in paired-end 2x100 mode on an Illumina Novaseq 6000. Reads were aligned to the human genome (hg38) with BWA (v. 0.7.15), and duplicates removed using Sambamba ( 24 ). Variant calling for constitutional DNA (PBMCs) used GATK HaplotypeCaller (v. 3.8.1), while MuTect2 (v. 2.0) was used for tumor DNA, followed by post-processing to filter artifacts and germline variants. Variants were annotated with VEP (v. 101). Single nucleotide variants (SNVs) for the ctDNA detection panel were selected based on somatic score, tumor variant allele frequency (VAF, ≥ 5%), and low population frequency (≤ 0.5%). Additionally, each SNVs was evaluated in each patient’s PBMC to exclude those potentially linked to clonal hematopoiesis. Detection of ctDNA in Plasma based on genetic alterations Cell-free DNA from the remaining elution volume of plasma extractions, corresponding to a mean input of 9 ng (IQR 2.6–9.7) was used for mutation analysis with the Twist Mechanical Fragmentation Kit (Twist Bioscience). To enhance the detection of low-frequency SNVs, we added molecular barcodes during library preparation, enabling the generation of consensus reads after sequencing. Libraries were target-enriched for tumor-specific SNVs genetic coordinates using designed panels with Twist MRD500 panels solution and sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode. Data processing involved molecular barcodes management with fgbio (v. 2.3.0), alignment with BWA (v. 2.2.1), and variant calling using samtools mpileup (v. 1.9). To assess the reliability of ctDNA detection, the frequency of reads carrying tumor-specific SNVs was compared to the frequency of reads carrying unspecific SNVs at genomic positions confirmed to be non-mutated in the tumor and PBMC of each patient. For each tumor-SNV position, a p-value was calculated using the binomial probability distribution 𝑃(𝑋=nb.tumor-SNV ∣ Depth, Freq.unspecific-SNV) where: nb.tumor-SNV represents the number of reads carrying a tumor-specific SNV, Depth is the total number of reads overlapping the genomic coordinate of the tumor-SNV Freq.unspecific-SNV corresponds to the sequencing error rate, estimated for each base type and each library using genomic regions confirmed to be free of SNVs in both the tumor and PBMC of the patient (i.e. the distribution under null hypothesis H 0 ). Tumor-SNV with p-value ≥ 10 − 3 were considered non-relevant. A sample was classified as ctDNA positive when at least two tumor-SNVs were detected with p-values ≤ 10 − 3 , corresponding to a false positive probability ≤ 10 − 6 . This method enables a statistical evaluation of tumor-derived ctDNA detection while accounting for background sequencing noise. ctDNA proportion assessment with the tumor-informed approach CtDNA quantification using the tumor-informed approach was performed by calculating the ratio of sequencing reads carrying tumor-SNVs to the total number of reads overlapping the genomic coordinates of all tumor-SNVs identified in the patient’s tumor tissue, for samples classified as ctDNA-positive. Statistical Analysis All statistical analyses were performed in R (version 4.0.2) environments. Machine Learning model was developed in Python (version 3.7.2) with scikit-learn 1.0.2 package, numpy 1.21.6, pandas 1.3.5 and seaborn 0.12.2. Deconvolution of methylation for cellular origin calculations was performed using UXM_deconvolution ( 25 ). Progression-free survival (PFS) and overall survival (OS) were compared between patient groups based on ctDNA presence. PFS was defined as the time from the end of first-line therapy to tumor recurrence. OS were defined as the time from the initial diagnosis to the death from any cause. Kaplan-Meier curves were used to visualize differences in PFS and OS, and a log-rank test was performed to compare the two groups. 3 Results Characteristics of the study population Patients Baseline characteristics of the 47 patients included in this study are presented in Table 1 . The median age at diagnosis was 66 (IQR 60–71). Patients received standard-of-care treatment consisting of primary cytoreductive surgery followed by adjuvant chemotherapy (n = 23), neoadjuvant chemotherapy before interval cytoreductive surgery followed by adjuvant chemotherapy (n = 17), or chemotherapy alone for 7 patients. The median follow-up time for the 47 patients is 84 months (IQR [35.8-not reached]). Among included patients, 33 were analyzed at multiple time points, resulting in a total of 63 plasma samples collected while chemotherapy (V2-V7, Supplemental Fig. 1). Heathly subjects Plasmas obtained from Biopredic International were sourced from women with a median age of 64 years (IQR 53–75) and without gynecological diseases. Additional plasma samples were provided by l’Établissement Français du Sang (EFS), collected from women aged 40 to 50 years, with no reported diseases. Healthy ovarian tissue samples were acquired from Origene (n = 7), with donors having a median age of 52 years (IQR 45–59). To enhance the dataset, we included publicly available whole-genome bisulfite sequencing (WGBS) data from the ENCODE database, derived from two ovarian tissue donors aged 30 and 51 years. Characterizing genetic and epigenetic alterations of ovarian tumors, PBMC and healthy ovaries for custom panels design We aimed to compare the capabilities of patient tumor-informed and tumor-type informed approaches in detecting and quantifying ctDNA. For the patient tumor-informed approach, we performed WES on DNA extracted from ovarian tumors and matched PBMCs from 12 patients from the BIOVAIRE cohort. We found between 16 and 204 SNVs per patient (mean = 74). Eight out of the 12 patients showed a mutated allele of TP53 (67%), 4 had mutations in TTN (33%), and 3 in CSMD3 (25%). However, no recurrent hotspot mutation was identified (Fig. 1 A, upper panel; see Methods for details). Table 1 Characteristics of the 47 study patients. N = 47 1 Age at diagnostic (median (IQR)) 66 (60, 71) BMI (median (IQR)) 22.48 (20.8 ; 27.66) Menopausal 41 (87%) BRCA mutation Negative 37 (78.7%) BRCA 1 2 (4.3%) BRCA 2 2 (4.3%) Non investigated 6 (12.7%) ASA score 1 10 (21.3%) 2 21 (44.7%) 3 7 (14.9%) No data available 9 (19.1%) Histology Clear cell adenocarcinoma 1 (2%) Endometrioid carcinoma (grade I and II) 2 (4.3%) Low grade serous carcinoma 2 (4.3%) High grade serous carcinoma 42 (89.4%) FIGO stage FIGO II B 1 (2%) FIGO III 3 (6.5%) FIGO III B 4 (8.5%) FIGO III C 25 (53.2%) FIGO IV 14 (29.8%) Surgical resection 44 (88%) CC0 38 (80.9%) CC1 1 (2%) CC2 1 (2%) 30-day postoperative mortality 2 (4.3%) Therapeutic sequence Upfront surgery followed by adjuvant chemotherapy 23 (49%) Neoadjuvant chemotherapy followed by interval surgery V3/V4/V5 16 (34%) Neoadjuvant chemotherapy followed by surgery after V6 1 (2%) Chemotherapy alone 7 (15%) Total number of chemotherapy cycles (median; IQR) 6 (6 ; 7) Follow-up in months (median; IQR) 84 (35.75 ; not reached) Recurrence rate 26 (55.3%) Time from last chemotherapy to recurrence (median; IQR) 11.7 (9.43 ; 16.1) Mortality rate 17 (36%) Lost to follow-up 1 (2%) Time from death to last chemotherapy (median; IQR) 24.2 (17.6 ; 46.6) 1 Median (iqr); n (%) IQR: interquartile range; BMI: body mass index; FIGO: international federation of gynecology and obstetrics; BRCA: breast cancer gene; CC: cytoreduction score. For the tumor-type informed approach, we endeavored to identify specific methylation patterns of EOC using enzymatic conversion and library enrichment on a panel targeting 3,98 million CpGs by comparing tumor samples (n = 12) with healthy ovarian tissues (n = 7) and PBMCs (n = 12). PBMC samples were analyzed because the majority of cfDNA present in plasma samples originates from these cells. The two WGBS datasets were added to the healthy ovarian tissues group. Based on this approach, we identified 52,173 CpGs that were consistently differentially methylated in the same direction (either hypomethylated or hypermethylated) when comparing ovarian tumors to both healthy ovarian tissues and PBMCs (Fig. 1 A, bottom panel). The distribution of DMLs across genomic regions revealed distinct patterns of hypermethylation and hypomethylation (Fig. 1 B). In CpG islands, 5,179 CpGs were hypermethylated, compared to 1,574 that were hypomethylated, highlighting a pronounced bias toward hypermethylation in these regions. In contrast, open sea regions exhibited a predominant pattern of hypomethylation, with 23,239 CpGs hypomethylated versus 6,730 hypermethylated. In the shores and shelves of CpG islands, the distributions of hypermethylated and hypomethylated CpGs were balanced. Specifically, CpG island shores contained 4,756 hypermethylated and 4,751 hypomethylated CpGs, while shelves showed a slight bias toward hypomethylation, with 3,042 hypomethylated and 2,902 hypermethylated CpGs. These findings underscore the region-specific nature of DNA methylation alterations in ovarian tumors, with hypermethylation predominantly occurring in CpG islands and hypomethylation prevailing in open sea regions. To assess whether the identified DMLs could differentiate ovarian tumor samples from healthy ovarian tissues, we selected the 2,000 DMLs with the highest standard deviation of methylation value across samples to generate unsupervised clustering. This clustering revealed three distinct clusters of samples, with healthy ovarian tissues clustering together. The distance between healthy samples and tumor samples increases with tumor cellularity and tumor percentage of ovarian epithelium component (Fig. 1 C, Supplemental Fig. 2). Thus, the selected DMLs enabled a clear classification of ovarian tissues and were selected for the design of a unique custom panel of 2.90 Mb. Tumor-informed approach for ctDNA characterization After identifying tumor-SNVs and designing the custom panels, we analyzed plasma samples (n = 28) from the twelve patients at different time-points: baseline (V1, n = 12), during treatment (V2-V5, n = 11), and at the end of treatment (V6-V7, n = 5). We generated an average of 11.66 Gb of data per sample (range: 7.65–15.56 Gb). The presence of ctDNA in plasma samples was assessed by comparing the frequency of reads carrying tumor-specific SNVs to the background sequencing error rate for each nucleotide, which was determined at genomic positions without variations in both tumor and PBMC samples from the corresponding patients (Supplemental Fig. 3). This method provided a robust framework for evaluating the reliability of ctDNA detection in each sample. Among the 28 plasma samples analyzed, 19 were ctDNA positive with at least 2 tumor-SNVs with a p-value < 10 − 3 . Figure 2A illustrates the baseline distribution of tumor-SNVs VAF in plasma for each patient. Among the 12 patients, 11 were classified as ctDNA-positive, with a mean VAF of tumor-SNVs of 1.30% (range: 0.03–3.39%). Patients undergoing neoadjuvant chemotherapy exhibited a higher mean VAF of 2.70% (95% CI [1.45–3.95]) compared to those undergoing primary cytoreductive surgery, who had a mean VAF of 0.46% (95% CI [0.01–0.91]) (Wilcoxon test, p-value = 9 × 10⁻⁴, Fig. 2B). Among plasma samples collected at different time points (n = 16) from 8 of the 12 patients, 8 samples were ctDNA-positive. Importantly, all patients showed a significant decrease in ctDNA levels during chemotherapy (Fig. 2C). Two out of 5 patients remained ctDNA-positive at the end-of-treatment time points (V6–V7). Tumor-type informed approach for ctDNA characterization Tumor-type informed approach is based on the training and validation of a support vector machine (SVM) classifier. We defined a training group of 35 baseline plasmas from EOC patients and 24 plasmas from healthy subjects, used for DMR refinement and SVM model training. The validation group consisted of 28 plasma samples, from the twelve same patients analyzed with the tumor-informed approach and 30 new plasma samples from different healthy subjects. This validation group was used both to assess the sensitivity and specificity of the SVM classifier and to compare the tumor informed and tumor-type informed strategies (Fig. 3 A). Using DMLs from the tumor-type informed custom panel, we defined DMRs based on the principle of methylation coherence among closely spaced CpGs ( 26 ). Briefly, we extended genomic regions by 100bp in each direction from DMLs, and merged the regions within 50bp of each other. A DMR was considered only if it contained at least 4 CpGs. Using this procedure 11,260 DMRs were defined, spanning a genomic region of 3,4 Mb. After the targeted-methylation enrichment and sequencing of cfDNA extracted from the plasma samples, we performed horizontal analysis of cell-free DNA, focusing on the individual fragment methylation patterns rather than the average methylation across all fragments at each DML (Fig. 3 B). The tumor-derived cfDNA is identified based on the similarity of the methylation status of the different DMLs it contains. A fragment is attributed to a tumor-derived origin if at least 85% of the present DMLs share the same status as that defined by the analysis of EOCs. The beta value represents the proportion of tumor-derived fragments in the plasma sample for each DMR. In the first step, we analyzed the plasma samples from healthy subjects in the training group to calculate the beta value standard deviation for the selected DMRs. We then excluded 1,528 DMRs with a beta value standard deviation greater than 0.05. In the second step, we evaluated the ability of each DMR's beta value to differentiate between healthy and cancer plasma samples by calculating the area under the curve (AUC). A total of 6,286 DMRs, 1,470 hypermethylated DMRs and 4,816 hypomethylated DMRs with an AUC ≥ 0.65 were retained (Fig. 3 C). On these selected DMRs, beta values were increased for baseline patient samples compared with healthy controls (Supplemental Fig. 4). These DMRs were used to train a SVM model for samples classification between healthy and cancer patients which is considered as a surrogate marker of the absence or the presence of ctDNA, respectively. To quantify ctDNA by the tumor-type informed method, we used healthy samples from the training group to identify the background noise of beta value and selected a reduced set of DMRs (n = 201). We validated the method using synthetic samples (n = 12) with known tumor DNA quantities (Supplemental Fig. 5, see Methods). The correlation between the expected and measured tumor DNA fractions yielded an R of 0.85 (p < 10⁻³). The SVM model was then tested on the validation group. All healthy subjects were classified as ctDNA negative whereas all but one cancer baseline plasma (V1) patients were considered as ctDNA positive, leading to a theorical sensitivity of 91.7%, a specificity of 100% and an accuracy of 97.6% (Fig. 4 A). The mean ctDNA fraction of positive samples was 1.27% (range: 0.04–3.77%). Interestingly, the patient with a tumor-informed ctDNA negative status at baseline was considered as ctDNA positive with the tumor-type informed method, with ctDNA fractions of 0.08. The negative one by this latter method was considered as ctDNA positive by the tumor-informed approach with a tumor-SNV VAF of 0.15%. The correlation of ctDNA fraction obtained with tumor-type informed approach with the tumor-informed approach is shown on Fig. 4 B, with an R of 0.53 (Spearman correlation, p = 3.4 x 10 − 3 ). When we considered the follow-up plasmas of this cohort of validation collected at different time points during chemotherapy (V2–V5, n = 11), 7 samples were classified as ctDNA-positive by the SVM classifier. Four out of 5 patients remained ctDNA-positive at the final time point (V6–V7). Like the tumor-informed approach, all patients exhibited a significant decrease in ctDNA levels during chemotherapy (Fig. 4 C). Of the patient plasma samples, 67.8% were concordant for ctDNA status between the tumor-informed and tumor-type informed strategies (n = 19), 21.4% were positive only with the tumor-type informed approach (n = 6), and three were positive only with the tumor-informed approach (Table 2 ). Table 2 Contingency table comparing results of tumor-informed and tumor-type informed approaches for ctDNA detection. Positives samples tumor-informed Negative samples tumor-informed Total Positive samples tumor-type informed 16 6 22 Negative samples tumor-type informed 3 3 6 Total 19 9 28 In the tumor-informed approach, samples are classified as ctDNA-positive when at least 2 tumor-SNVs with a p-value < 10 − 3 are detected. For the tumor-type informed approach, samples are classified as ctDNA-positive if the predicted value from the SVM classifier is positive. These results highlighted the comparable sensitivity of the developed tumor-type informed approach with the tumor-informed one (McNemar test, p = 0.51), with a tendency to a better detection of ctDNA at the last time points with the tumor-type informed approach. We illustrated that using horizontal analysis on specific DMRs, we were able to detect ctDNA with a similar sensitivity than with the tumor-informed approach with a reduced quantity of data. Over-fitting assessment of the Support Vector Machine classifier We evaluated our machine learning model for overfitting using 47 unused follow-up plasma samples from 25 patients. These samples came from the same patients whose baseline plasma samples were used to train the SVM model. We showed that 70.2% of the samples (33 out of 47) were classified as ctDNA positive, comparable to that obtained on the follow-up plasma from the validation group (68.8%; 11 out of the 16 samples). Next, combining all samples except those used in the training of the SVM model, we examined whether cfDNA input in library preparation could influence the predicted value of the model for a sample and found no significant correlation (Spearman’s test R = -5.9 × 10 − 2 , p = 0.95 ). The model tended to separate healthy plasma samples from plasma patients with better accuracy when cfDNA input increases (Supplemental Fig. 6). Finally, on the 54 healthy plasma samples, we assessed for each DMR the association between beta values and age of healthy subjects and found no significant correlation (Supplemental Fig. 7). Clinical validation of MRD assessment using the SVM classifier On the 47 supplementals follow-up plasmas samples from 25 patients, (Fig. 3 A), we assessed the capacity of the model to detect MRD. Results of ctDNA detection and quantification from this group of plasma not analyzed with the tumor-informed approach are presented in Supplemental Fig. 8. Most patients showed a decrease in ctDNA fraction along treatment (21 out of the 25 patients tested). We assessed the clinical relevance of our tumor-type informed approach for predicting EOC outcomes by analyzing plasma samples collected at the end of treatment. Among 22 patients tested for MRD detection at V6 (n = 7) or V7 (n = 15), 16 (72.7%) were classified as ctDNA-positive, with a mean ctDNA fraction of 0.17%. Of these positive patients, 12 experienced EOC relapse, with a median progression-free survival (PFS) of 11 months (95% CI, 10.04–28.38). In contrast, among the 6 ctDNA-negative patients, only 1 relapsed within 15 months, and the median PFS for this group was not reached. Kaplan-Meier analysis demonstrated significant differences in PFS between ctDNA-positive and ctDNA-negative patients at the end of chemotherapy (log-rank p = 0.017, Hazard ratio = 8.24; 95% CI [1.06–64.4]; Fig. 5 A). None of the ctDNA-negative patients at V6 or V7 died during follow-up, while 37.5% of ctDNA-positive patients (6 out of 16) succumbed during the surveillance period. Kaplan-Meier analysis also revealed significant differences in overall survival (OS) between ctDNA-positive and ctDNA-negative groups (log-rank p = 0.041; Fig. 5 B). Overall, our tumor-type informed SVM model effectively predicted EOC relapses using reduced sequencing data compared to tumor-informed strategies. 4 Discussion Highly sensitive methods are required for the detection and quantification of ctDNA in MRD assessment. In this study, we compared two approaches: a tumor-informed approach based on somatic SNVs identified from the patients tumor samples and a tumor-type informed approach utilizing common DMRs in EOC samples. These strategies were applied to 47 patients from the Biovaire cohort, which benefited from an extended follow-up period, with some patients monitored for up to 100 months after the completion of first-line therapy. In the tumor-informed approach, we performed WES on tumor samples, using constitutional DNA from PBMCs as a reference to identify tumor-specific SNVs. In EOC samples from TCGA ( 27 – 32 ), the median number of SNVs detected by WES is 62.8 (IQR: 34–77), close to that we observed for the 12 patients tested in our series by tumor informed approach (median 67, IQR: 43–71). Most patients harbored mutations in commonly altered EOC genes, such as TP53 , TTN , and CSMD3 ( 33 , 34 ). However, this relatively small number of SNVs limits ctDNA detection when plasma volume is restricted. On the 12 baseline plasma samples tested by the tumor-informed approach we extracted an average of 15.3 ng cfDNA, representing 4500 genome equivalents (GE). According to Zviran et al ., the probability of detecting ctDNA at a 10⁻⁴ fraction under these conditions is between 80%-100% ( 8 ). Consistent with this, 11 out of 12 samples were ctDNA-positive. The negative sample had one of the lowest cfDNA input for library preparation (6.1 ng). For the follow-up plasma samples, with an average cfDNA input of 4.2 ng, 8 out of 16 samples were ctDNA-positive. Sensitivity might be improved by increasing cfDNA input (which would require larger blood volumes) or by targeting variants more resistant to sequencing errors, such as indels or phased variants. However, this approach would significantly increase sequencing costs, particularly due to the need for WGS of the tumor to identify these specific variants ( 35 , 36 ). Optimizing the number of markers is thus critical to overcoming these limitations. To address this purpose, we evaluated whether cfDNA methylation analysis in a tumor-type informed context could enhance ctDNA detection while remaining cost-effective. Publicly available datasets, primarily derived from Illumina’s 27K methylation chip, have limited potential for biomarkers discovery. To overcome this, we generated a comprehensive dataset using a methylation panel covering ~ 14% of the CpGs in the human genome. To enhance robustness, we supplemented this dataset with two ENCODE WGBS files from healthy ovarian tissues. In total, we identified 6,286 DMRs, which were used for the design of a single panel for tumor-type informed cfDNA sequencing. Several genes associated with these DMRs, such as CELF2, FAIM2, SIM2, CDO1 , and C2CD4D , are already described as potential EOC biomarkers. Others, such as JAK3 and SOX2 , are known tumor driver genes ( 37 – 39 ). By identifying more than 6000 markers, this extensive panel increased the theoretical limit of ctDNA detection to 10⁻⁵ for 2000 GE of cfDNA input. When comparing the two strategies using baseline plasma samples, the proportion of ctDNA-positive samples was the same: 91.7%. However, in follow-up plasma samples, the tumor-type-informed method proved superior sensivity, detecting ctDNA in 11 out of 16 samples, whereas the tumor-informed approach detected only 8 ctDNA-positive samples. This innovative approach improves sensitivity, enabling detection of ctDNA at fractions as low as 0.05%, while reducing turnaround time, eliminating the need for tumor analysis, and significantly lowering costs by requiring a single methylation panel and eightfold fewer sequencing reads than the tumor-informed method. Horizontal methylation analysis further strengthened ctDNA detection by reducing dependence on unmethylated cytosine conversion rates and sequencing errors ( 22 ). By analyzing multiple CpGs per cfDNA fragment and integrating an SVM model, we achieved 100% specificity in ctDNA detection. Furthermore, the tumor-type informed method allows the identification of MRD of patients with a sensitivity of 92.3%, 12 patients relapse when ctDNA was detected at the end of adjuvant treatment with a median time of relapse of 11 months. The specificity of the test is lower (55.6%) but 2 patients were lost to follow-up before reaching two years. This study has some limitations. First, the DMRs selection was based on a small cohort comprising 12 EOC tissues, 12 matched PBMC samples, and 9 healthy ovarian tissues. Although we studied a limited number of patient tumors to identify our DMRs, which does not fully protect against inter-individual variability in tumor profiles, the overall sensitivity of our method for detecting ctDNA at baseline is 97.9%. While we accounted for age, cfDNA input, and patient variability, the possibility of overfitting in the SVM cannot be excluded even if its likelihood is weak. Finally, the performance of our test to predict EOC relapse at the end of adjuvant treatment was validated in only 22 patients. Larger independent cohorts are required to confirm these findings before initiating clinical trials to assess its clinical utility. Compared to other studies, the tumor-type informed approach demonstrated similar sensitivity and specificity to the study by Hou et al ., which employed deep whole-exome sequencing (WES) to identify tumor-specific SNVs and track them using an amplicon-based method ( 40 ). In contrast, the tumor-naïve assay developed by Hoe et al ., which monitors ctDNA through nine frequently mutated genes, detected ctDNA in only 69.2% of baseline samples. While their method accurately predicted relapse in many cases, some patients classified as negative ultimately relapsed, underscoring a notable lack of sensitivity ( 41 ). Our study holds promises for improving EOC patient management. Stratifying patients as MRD-positive or -negative could enhance quality of life by sparing unnecessary maintenance treatments for some while intensifying therapeutic efforts for others with MRD evidence. The tumor-type informed approach is cost-effective, time-efficient, and eliminates the need for tumor biopsies. It is also adaptable to other cancer types and could be a valuable tool for evaluating treatment responses in clinical assays. 5 Conclusion By analyzing cfDNA methylation patterns in plasma from healthy donors and EOC patients, we achieved improved sensitivity for detecting MRD and predicting relapse compared to a traditional tumor-informed approach. Notably, this approach allows classification of plasma samples regarding ctDNA presence without requiring prior knowledge of the patient’s tumor SNVs. It achieves this purpose with eightfold less data than tumor-informed strategies that rely on WES of the patient’s tumor and mutation tracking in plasma. This cost-effective and non-invasive method holds significant potential to improve EOC patient care. For patients deemed free of MRD following first-line treatment, a de-escalation of maintenance therapy can be envisioned. Conversely, MRD-positive patients could benefit from closer monitoring and tailored interventions, potentially improving outcomes. Future studies should validate these findings in larger, multi-center cohorts, paving the way for broader clinical implementation. Declarations Ethics approval and consent to participate Study protocol was approved by the ethic committee (Comité de Protection des Personnes – Ile de France II », Research number: 2013-10-01 MS1, Biovaire Study n° ID-RCB: 2013-A01108-37. All patients recruited in the study signed an informed written consent . Consent for publication Not applicable Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Competing interests For PLP: President of Canceropole Ile de France; Director of the « Institut du cancer Paris Carpem; Speaker, Consultancy or Advisory role for Biocartis; Pierre Fabre; BMS; Servier; MSD; Founder and ownership interests in Methys Dx. For MBS, CM, SG and EM: employees at Integragen SA. For HA: Speaker, Consultancy or Advisory role for GlaxoSmithKline. VT : Founder, president and Chief scientific officer, and ownership interests in Methys Dx. Funding This study is partly funded by l’Association Nationale de la Recherche et de la Technologie and IntegraGen. Authors' contributions MBS have made substantial contributions to experimental design, library preparation for sequencing, data analysis, interpretation of results, and drafting of the manuscript. HA have made substantial contributions to the experimental design, the interpretation of data and revised the work. CM have made substantial contributions to the conception, design of the work, interpretation of data and revised the work and the manuscript. SG have made substantial contributions to the analysis of data. EM have made substantial contributions to the conception and design of the works JA have made substantial contributions to the conception and design of the works. LB have made substantial contributions to the analysis of clinical data. ADR have made substantial contributions to the analysis of methylation data. EL have made substantial contributions to interpretation of clinical data. CD have made substantial contributions to the analysis of data regarding Kaplan-Meier Curve. JM have made substantial contributions to the conception and design of the works. BB have made substantial contributions to the conception and design of the works. AB have made substantial contributions to the conception and design of the works. VT have made substantial contributions to the conception, design of the work, interpretation of data and substantively revised the work and the manuscript. PLP lead the project and have made substantial contributions to the conception, design of the work, interpretation of data and substantively revised the work and the manuscript. All authors read and approved the final manuscript. Acknowledgements We thank the participating patients to BIOVAIRE cohort, Institut du cancer PARIS CARPEM, and the SIRIC CARPEM for their support in the collection as well as EPIGENTEC Biobank. References Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209–49. Hennessy BT, Coleman RL, Markman M. Ovarian cancer. Lancet. 2009;374(9698):1371–82. Berek JS, Renz M, Kehoe S, Kumar L, Friedlander M. Cancer of the ovary, fallopian tube, and peritoneum: 2021 update. Int J Gynaecol Obstet. 2021;155(1):61–85. du Bois A, Reuss A, Pujade-Lauraine E, Harter P, Ray-Coquard I, Pfisterer J. Role of surgical outcome as prognostic factor in advanced epithelial ovarian cancer: a combined exploratory analysis of 3 prospectively randomized phase 3 multicenter trials: by the Arbeitsgemeinschaft Gynaekologische Onkologie Studiengruppe Ovarialkarzinom (AGO-OVAR) and the Groupe d’Investigateurs Nationaux Pour les Etudes des Cancers de l’Ovaire (GINECO). Cancer. 2009;115(6):1234–44. Torre LA, Trabert B, DeSantis CE, Miller KD, Samimi G, Runowicz CD, et al. Ovarian cancer statistics, 2018. CA Cancer J Clin. 2018;68(4):284–96. Zhang M, Cheng S, Jin Y, Zhao Y, Wang Y. Roles of CA125 in diagnosis, prediction, and oncogenesis of ovarian cancer. Biochim Biophys Acta Rev Cancer. 2021;1875(2):188503. Pantel K, Alix-Panabières C. Liquid biopsy and minimal residual disease - latest advances and implications for cure. Nat Rev Clin Oncol. 2019;16(7):409–24. Zviran A, Schulman RC, Shah M, Hill STK, Deochand S, Khamnei CC, et al. Genome-wide cell-free DNA mutational integration enables ultra-sensitive cancer monitoring. Nat Med. 2020;26(7):1114–24. Lo YMD, Han DSC, Jiang P, Chiu RWK. Epigenetics, fragmentomics, and topology of cell-free DNA in liquid biopsies. Science. 2021;372(6538):eaaw3616. Kim SH, Weigelt B. Circulating cell-free (cf)DNA analysis: Current technologies and applications in gynecologic cancer. Gynecol Oncol Rep. 2024;54:101431. Dong Q, Chen C, Hu Y, Zhang W, Yang X, Qi Y, et al. Clinical application of molecular residual disease detection by circulation tumor DNA in solid cancers and a comparison of technologies: review article. Cancer Biol Ther. 2023;24(1):2274123. Chalmers ZR, Connelly CF, Fabrizio D, Gay L, Ali SM, Ennis R, et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med. 2017;9:34. Luo H, Wei W, Ye Z, Zheng J, Xu RH. Liquid Biopsy of Methylation Biomarkers in Cell-Free DNA. Trends Mol Med. 2021;27(5):482–500. Liang L, Zhang Y, Li C, Liao Y, Wang G, Xu J, et al. Plasma cfDNA methylation markers for the detection and prognosis of ovarian cancer. eBioMedicine. 2022;83:104222. Liu MC, Oxnard GR, Klein EA, Swanton C, Seiden MV, Liu MC, et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol. 2020;31(6):745–59. Garrigou S, Perkins G, Garlan F, Normand C, Didelot A, Le Corre D, et al. A Study of Hypermethylated Circulating Tumor DNA as a Universal Colorectal Cancer Biomarker. Clin Chem. 2016;62(8):1129–39. Beinse G, Borghese B, Métairie M, Just PA, Poulet G, Garinet S, et al. Highly Specific Droplet-Digital PCR Detection of Universally Methylated Circulating Tumor DNA in Endometrial Carcinoma. Clin Chem. 2022;68(6):782–93. Vaisvila R, Ponnaluri VKC, Sun Z, Langhorst BW, Saleh L, Guan S, et al. Enzymatic methyl sequencing detects DNA methylation at single-base resolution from picograms of DNA. Genome Res. 2021;31(7):1280–9. Feng H, Conneely KN, Wu H. A Bayesian hierarchical model to detect differentially methylated loci from single nucleotide resolution sequencing data. Nucleic Acids Res. 2014;42(8):e69. Akalin A, Kormaksson M, Li S, Garrett-Bakelman FE, Figueroa ME, Melnick A, et al. methylKit: a comprehensive R package for the analysis of genome-wide DNA methylation profiles. Genome Biol. 2012;13(10):R87. Korthauer K, Chakraborty S, Benjamini Y, Irizarry RA. Detection and accurate false discovery rate control of differentially methylated regions from whole genome bisulfite sequencing. Biostatistics. 2019;20(3):367–83. Li W, Li Q, Kang S, Same M, Zhou Y, Sun C, et al. CancerDetector: ultrasensitive and non-invasive cancer detection at the resolution of individual reads using cell-free DNA methylation sequencing data. Nucleic Acids Res. 2018;46(15):e89. Loyfer N, Rosenski J, Kaplan T. wgbstools: A computational suite for DNA methylation sequencing data representation, visualization, and analysis. 2024. Tarasov A, Vilella AJ, Cuppen E, Nijman IJ, Prins P. Sambamba: fast processing of NGS alignment formats. Bioinformatics. 2015;31(12):2032–4. Loyfer N, Magenheim J, Peretz A, Cann G, Bredno J, Klochendler A, et al. A DNA methylation atlas of normal human cell types. Nature. 2023;613(7943):355–64. Affinito O, Palumbo D, Fierro A, Cuomo M, De Riso G, Monticelli A, et al. Nucleotide distance influences co-methylation between nearby CpG sites. Genomics. 2020;112(1):144–50. Berger AC, Korkut A, Kanchi RS, Hegde AM, Lenoir W, Liu W, et al. A Comprehensive Pan-Cancer Molecular Study of Gynecologic and Breast Cancers. Cancer Cell. 2018;33(4):690–e7059. de Bruijn I, Kundra R, Mastrogiacomo B, Tran TN, Sikina L, Mazor T, et al. Analysis and Visualization of Longitudinal Genomic and Clinical Data from the AACR Project GENIE Biopharma Collaborative in cBioPortal. Cancer Res. 2023;83(23):3861–7. Elkin R, Oh JH, Liu YL, Selenica P, Weigelt B, Reis-Filho JS, et al. Geometric network analysis provides prognostic information in patients with high grade serous carcinoma of the ovary treated with immune checkpoint inhibitors. NPJ Genom Med. 2021;6(1):99. Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609–15. Gao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013;6(269):pl1. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2(5):401–4. Lu N, Liu J, Xu M, Liang J, Wang Y, Wu Z, et al. CSMD3 is Associated with Tumor Mutation Burden and Immune Infiltration in Ovarian Cancer Patients. Int J Gen Med. 2021;14:7647–57. Ma MC, Lavi ES, Altwerger G, Lin ZP, Ratner ES. Predictive modeling of gene mutations for the survival outcomes of epithelial ovarian cancer patients. PLoS ONE. 2024;19(7):e0305273. Fernandez SV, Tan YF, Rao S, Fittipaldi P, Sheriff F, Borghaei H, et al. Validation of a Molecular Diagnostic Test for Circulating Tumor DNA by Next-Gen Sequencing. Int J Mol Sci. 2023;24(21):15779. Kurtz DM, Soo J, Co Ting Keh L, Alig S, Chabon JJ, Sworder BJ, et al. Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA. Nat Biotechnol. 2021;39(12):1537–47. Marinelli LM, Kisiel JB, Slettedahl SW, Mahoney DW, Lemens MA, Shridhar V, et al. Methylated DNA markers for plasma detection of ovarian cancer: Discovery, validation, and clinical feasibility. Gynecol Oncol. 2022;165(3):568–76. Kallio HM, Savolainen K, Virtanen T, Ryyppö L, Selin H, Martikainen P, et al. Sensitive circulating tumor DNA-based residual disease detection in epithelial ovarian cancer. Life Sci Alliance. 2024;7(6):e202402658. Senturk Kirmizitas T, van den Berg C, Boers R, Helmijr J, Makrodimitris S, Dag HH, et al. Epigenetic and Genomic Hallmarks of PARP-Inhibitor Resistance in Ovarian Cancer Patients. Genes (Basel). 2024;15(6):750. Hou JY, Chapman JS, Kalashnikova E, Pierson W, Smith-McCune K, Pineda G, et al. Circulating tumor DNA monitoring for early recurrence detection in epithelial ovarian cancer. Gynecol Oncol. 2022;167(2):334–41. Heo J, Kim YN, Shin S, Lee K, Lee JH, Lee YJ, et al. Serial Circulating Tumor DNA Analysis with a Tumor-Naïve Next-Generation Sequencing Panel Detects Minimal Residual Disease and Predicts Outcome in Ovarian Cancer. Cancer Res. 2024;84(3):468–78. Additional Declarations Competing interest reported. For PLP: President of Canceropole Ile de France; Director of the « Institut du cancer Paris Carpem; Speaker, Consultancy or Advisory role for Biocartis; Pierre Fabre; BMS; Servier; MSD; Founder and ownership interests in Methys Dx. For MBS, CM, SG and EM: employees at Integragen SA. For HA: Speaker, Consultancy or Advisory role for GlaxoSmithKline. VT : Founder, president and Chief scientific officer, and ownership interests in Methys Dx. Supplementary Files SupplementalFigures.docx Cite Share Download PDF Status: Published Journal Publication published 12 Jun, 2025 Read the published version in Journal of Experimental & Clinical Cancer Research → Version 1 posted Editorial decision: Revision requested 03 Mar, 2025 Reviews received at journal 02 Mar, 2025 Reviews received at journal 28 Feb, 2025 Reviewers agreed at journal 28 Feb, 2025 Reviewers agreed at journal 17 Feb, 2025 Reviewers invited by journal 17 Feb, 2025 Editor assigned by journal 17 Feb, 2025 Submission checks completed at journal 17 Feb, 2025 First submitted to journal 14 Feb, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6031886","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":423167927,"identity":"2731873f-21e6-47c0-9987-3d13e9ce64a0","order_by":0,"name":"Mehdi Ben Sassi","email":"","orcid":"","institution":"Centre de Recherche des Cordeliers, INSERM UMRS1138, CNRS SNC 5096, Sorbonne University, University of Paris","correspondingAuthor":false,"prefix":"","firstName":"Mehdi","middleName":"Ben","lastName":"Sassi","suffix":""},{"id":423167928,"identity":"b45d4e0b-6193-4dbb-accd-bbb205a7b34d","order_by":1,"name":"Henri Azais","email":"","orcid":"","institution":"Centre de Recherche des Cordeliers, INSERM UMRS1138, CNRS SNC 5096, Sorbonne University, University of Paris","correspondingAuthor":false,"prefix":"","firstName":"Henri","middleName":"","lastName":"Azais","suffix":""},{"id":423167929,"identity":"7d623a48-839e-4eb4-8084-d5f2261b4a05","order_by":2,"name":"Charles Marcaillou","email":"","orcid":"","institution":"IntegraGen SA","correspondingAuthor":false,"prefix":"","firstName":"Charles","middleName":"","lastName":"Marcaillou","suffix":""},{"id":423167930,"identity":"61ff6e65-ba7c-4ae9-b75c-e57e776980fd","order_by":3,"name":"Sylvain Guibert","email":"","orcid":"","institution":"IntegraGen SA","correspondingAuthor":false,"prefix":"","firstName":"Sylvain","middleName":"","lastName":"Guibert","suffix":""},{"id":423167931,"identity":"eb14ffcf-1b9e-4b41-a2fb-20eacf0180ff","order_by":4,"name":"Emmanuel Martin","email":"","orcid":"","institution":"IntegraGen SA","correspondingAuthor":false,"prefix":"","firstName":"Emmanuel","middleName":"","lastName":"Martin","suffix":""},{"id":423167932,"identity":"922947be-700d-4972-898c-44e0ffe61913","order_by":5,"name":"Jérôme Alexandre","email":"","orcid":"","institution":"AP-HP, Department of Medical Oncology, Hôpital Cochin","correspondingAuthor":false,"prefix":"","firstName":"Jérôme","middleName":"","lastName":"Alexandre","suffix":""},{"id":423167933,"identity":"5bc9d245-bd45-407c-80e9-416f8214f045","order_by":6,"name":"Louise Benoit","email":"","orcid":"","institution":"AP-HP, Department of Gynaecological Oncology and Breast Surgery, Hôpital Européen Georges Pompidou","correspondingAuthor":false,"prefix":"","firstName":"Louise","middleName":"","lastName":"Benoit","suffix":""},{"id":423167934,"identity":"e830e1e6-3750-4e91-98a6-5619fb84a86a","order_by":7,"name":"Aurélien De Reynies","email":"","orcid":"","institution":"Centre de Recherche des Cordeliers, INSERM UMRS1138, CNRS SNC 5096, Sorbonne University, University of Paris","correspondingAuthor":false,"prefix":"","firstName":"Aurélien","middleName":"","lastName":"De Reynies","suffix":""},{"id":423167935,"identity":"82c1751f-49b1-413a-8131-6c05ed4fa5d1","order_by":8,"name":"Emilie Laude","email":"","orcid":"","institution":"AP-HP, Department of Gynaecological Oncology and Breast Surgery, Hôpital Européen Georges Pompidou","correspondingAuthor":false,"prefix":"","firstName":"Emilie","middleName":"","lastName":"Laude","suffix":""},{"id":423167936,"identity":"e8acd9d4-c67d-4cb4-93c8-863b68db8e82","order_by":9,"name":"Cam Duong","email":"","orcid":"","institution":"AP-HP, Department of Gynaecological Oncology and Breast Surgery, Hôpital Européen Georges Pompidou","correspondingAuthor":false,"prefix":"","firstName":"Cam","middleName":"","lastName":"Duong","suffix":""},{"id":423167937,"identity":"5bb7ab21-5c8b-4119-9ea3-1068a54f460e","order_by":10,"name":"Jacques Medioni","email":"","orcid":"","institution":"AP-HP-centre Université Paris Cité, Hôpital Européen Georges Pompidou","correspondingAuthor":false,"prefix":"","firstName":"Jacques","middleName":"","lastName":"Medioni","suffix":""},{"id":423167938,"identity":"f1d61efc-ed5a-47f8-a3a5-b5d667fe12ce","order_by":11,"name":"Bruno Borghese","email":"","orcid":"","institution":"AP-HP, Department of Gynaecological surgery, site Cochin","correspondingAuthor":false,"prefix":"","firstName":"Bruno","middleName":"","lastName":"Borghese","suffix":""},{"id":423167939,"identity":"7aabec15-ff52-430d-8129-c747c9520053","order_by":12,"name":"Anne-Sophie Bats","email":"","orcid":"","institution":"AP-HP, Department of Gynaecological Oncology and Breast Surgery, Hôpital Européen Georges Pompidou","correspondingAuthor":false,"prefix":"","firstName":"Anne-Sophie","middleName":"","lastName":"Bats","suffix":""},{"id":423167940,"identity":"04e36346-5578-4b47-8bff-1f93546b4805","order_by":13,"name":"Valérie Taly","email":"","orcid":"","institution":"Centre de Recherche des Cordeliers, INSERM UMRS1138, CNRS SNC 5096, Sorbonne University, University of Paris","correspondingAuthor":false,"prefix":"","firstName":"Valérie","middleName":"","lastName":"Taly","suffix":""},{"id":423167941,"identity":"55eca4fb-78bc-43f3-9c34-fcd6babc4732","order_by":14,"name":"Pierre Laurent-Puig","email":"data:image/png;base64,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","orcid":"","institution":"Centre de Recherche des Cordeliers, INSERM UMRS1138, CNRS SNC 5096, Sorbonne University, University of Paris","correspondingAuthor":true,"prefix":"","firstName":"Pierre","middleName":"","lastName":"Laurent-Puig","suffix":""}],"badges":[],"createdAt":"2025-02-14 15:38:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6031886/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6031886/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s13046-025-03433-4","type":"published","date":"2025-06-12T15:57:39+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":78246867,"identity":"1c417b4d-6276-42ec-978c-04ec688d5127","added_by":"auto","created_at":"2025-03-11 09:34:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":3102103,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. Schematic representation of the study design. \u003c/strong\u003eDNA was extracted from ovarian tumor samples, matched patient PBMC and normal ovarian tissues from independent donors. \u003cstrong\u003eUpper panel\u003c/strong\u003eillustrates the tumor informed approach. Whole-exome sequencing (WES) was performed on tumor-PBMC pairs to identify tumor-specific SNVs. SNVs from tumors of 3–5 patients were grouped to design custom probe panels, which were then used for library target-enrichment from plasma cfDNA. The prepared cfDNA libraries were subsequently sequenced. \u003cstrong\u003eThe lower panel \u003c/strong\u003eshows the tumor-type informed approach: same tumors and PBMCs samples are used. These samples, along with normal ovarian tissues, underwent sequencing following the enzymatic-targeted methylation protocol. This protocol targets a 123 Mb panel, covering 3.98 million CpGs. Common differentially methylated loci (DMLs) were identified, comparing ovarian cancer samples to normal ovarian tissues enriched with 2 whole genome bisulfite sequencing (WGBS) publicly available dataset on one hand, and ovarian cancer tissues to PBMCs on the other. These identified DMLs allowed the design of a unique custom methylation panel, which was then used to enrich libraries from healthy and patient plasma samples for sequencing. \u003cstrong\u003eB. Distribution of selected DMLs according to hypermethylated or hypomethylated status among the CpGs genomic features. \u003c/strong\u003e(i.e CpG island, shore, shelf, open sea).\u003cstrong\u003e \u003c/strong\u003eSelected\u003cstrong\u003e \u003c/strong\u003eDMLs are either hypomethylated (blue) or hypermethylated (red) in cancer samples in comparison with healthy ovarian tissues and PBMCs. \u003cstrong\u003eC. Unsupervised clustering of ovarian samples based on the 2000 most variable DMLs methylation proportion\u003c/strong\u003e. Upper annotations show: first, the type of sample (i.e normal ovary in cyan or ovarian tumor in purple), second, tumor cellularity based on copy number variation calculated from tumor WES and, third, ovarian epithelial cells component proportion for each sample deciphered from the methylation status of the 3.98M CpGs using UXM_deconv algorithm (16).\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/dbc2e83c9a42f000c7f2651e.png"},{"id":78246865,"identity":"d1287c53-77c4-4e99-ba76-06ad6eeed0cb","added_by":"auto","created_at":"2025-03-11 09:34:13","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1279181,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. Distribution of tumor-SNV variant allele frequencies (VAF) in baseline plasma using the tumor-informed approach. \u003c/strong\u003eObserved VAFs were compared to sequencing error rates, estimated for each nucleotide from the frequency of unspecific SNVs (i.e., genomic positions confirmed to be free of variants in tumor and PBMC samples). The probability of each variant arising from sequencing noise was assessed, and variants with a probability \u0026lt; 10⁻³ were considered reliable. CtDNA presence in sample was deemed significant when at least two tumor-SNVs met this criterion, with annotations as follows: ns (not significant), **** p \u0026lt; 10⁻⁶. \u003cstrong\u003eB\u003c/strong\u003e. \u003cstrong\u003eBoxplot illustrating the average VAF in baseline plasma samples for two distinct groups of patients. \u003c/strong\u003eOne group receiving neo-adjuvant chemotherapy as the first-line treatment (red) and the other undergoing tumor debulking as the initial intervention (blue). The mean allele frequency of tumor-SNVs was calculated for each sample, grouped by treatment type, and visualized to compare the average allele frequency between groups. \u003cstrong\u003eC. CtDNA follow-up relying on allele frequencies of tumor-SNVs by tumor-informed method. \u003c/strong\u003ePatients with multiple plasma samples available are shown. Error bars correspond to standard deviation of tumor-SNVs VAF. Red points indicate statistically significant values (\u0026lt; 10\u003csup\u003e⁻6\u003c/sup\u003e) for ctDNA detection compared with sequencing errors background noise, while black points are not significant.\u003cstrong\u003e \u003c/strong\u003eThe number of reliable tumor-SNVs found in the sample is displayed at the top of each panel.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/04f73eefebe4a1625a398c7c.png"},{"id":78246863,"identity":"3af6b6af-2c83-4144-8971-aaf0583aca4c","added_by":"auto","created_at":"2025-03-11 09:34:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2918097,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. Plasma samples workflow. \u003c/strong\u003eThe selection of plasma samples for support vector machine (SVM) training, validation against the tumor-informed approach, and clinical validation is depicted. \u003cstrong\u003eB.\u003c/strong\u003e \u003cstrong\u003eSchematic representation of the horizontal methylation analysis workflow. \u003c/strong\u003eExample of a differentially methylated region (DMR) identified as hypermethylated in ovarian tumor samples as compared to healthy ovarian tissues and PBMCs. CfDNA fragments are stacked and aligned based on their genomic coordinates, then grouped for their association to the DMR. Horizontal bars (red, and green) represent cfDNA fragments, visualized and classified using wgbstools (17). Methylation patterns across adjacent differentially methylated loci (DML) from individual fragments are analyzed. Red bars represent cfDNA fragments with a high density of methylated DML (\u0026gt;85%), indicating a fragment of a tumor-derived origine in this example, while green bars correspond to fragments with lower methylation levels (\u0026lt;85%) considering as non-tumor-derived origin. The beta value is defined as the ratio of cfDNA fragments associated with a tumor-derived origin to the total number of fragments within the DMR. \u003cstrong\u003eC. Strategy for DMRs refinement on plasma samples and machine learning analysis.\u003c/strong\u003e The initially selected DMRs by comparing ovarian tumor tissues to healthy ovarian tissus and PBMC (n = 11 260) were filtered by analyzing healthy plasma samples, excluding those with a beta values standard deviation above 0.05. The remaining DMR were used to assess the ability of beta values to distinguish between healthy and ovarian cancer plasma samples, by calculating for each DMR, the area under the curve (AUC). Only DMRs with an AUC \u0026gt; 0.65 are retained for Support Vector Machine (SVM) learning analysis.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/de75cc372cdb4e4c75443e1e.png"},{"id":78246858,"identity":"21201dc5-1621-44e2-9412-30c3b9bbad19","added_by":"auto","created_at":"2025-03-11 09:34:11","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":2039436,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. SVM model-predicted values for ctDNA presence of plasma samples. \u003c/strong\u003eResults for plasmas samples previously analyzed with the tumor-informed approach and 30 healthy donors are shown.\u003cstrong\u003e \u003c/strong\u003eThe\u003cstrong\u003e \u003c/strong\u003eSVM model predicts a value for each sample relying on the beta values of selected DMRs.\u003cstrong\u003e \u003c/strong\u003eThe red vertical line represents the classification threshold: samples with predicted value above 0 are classified as ctDNA-positive. Healthy control plasmas are represented as blue circles, while patient plasma samples are differentiated by colors and shapes according to the sampling time points (V1 to V7).\u003cstrong\u003e B. Correlation of the ctDNA quantification between tumor-type informed and tumor-informed approaches. C. ctDNA follow-up relying on tumor-type informed approach for patients with multiple plasma samples available, previously analyzed using the tumor-informed approach. \u003c/strong\u003eRed points indicate samples with positive predicted values for ctDNA presence using the support vector machine model, while black points represent negative predictions. Error bars represent the 95% confidence intervals (CI95) of the beta values from DMRs used for tumoral DNA quantification.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/05b46f897eb22d0bb8ab24a4.png"},{"id":78246868,"identity":"97489f34-073a-40be-bb6e-48670ae41d81","added_by":"auto","created_at":"2025-03-11 09:34:15","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":515646,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA. Progression-Free Survival regarding microscopic residual disease status defined with the support vector machine learning classifier. B. Overall survival regarding microscopic residual disease status defined with the support vector machine learning classifier.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/c50152ef64f0da5a6061c078.png"},{"id":84726489,"identity":"249e663a-a7ea-4398-8f41-8fe3d2cf9d3f","added_by":"auto","created_at":"2025-06-16 16:05:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11962957,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/64514fee-b4bb-46ba-9d83-67250093081d.pdf"},{"id":78246862,"identity":"56bd37f7-41c4-4d0e-911f-1b48bcffb636","added_by":"auto","created_at":"2025-03-11 09:34:11","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":509802,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFigures.docx","url":"https://assets-eu.researchsquare.com/files/rs-6031886/v1/27fa67c2710e0ff523278143.docx"}],"financialInterests":"Competing interest reported. For PLP: President of Canceropole Ile de France; Director of the « Institut du cancer Paris Carpem; Speaker, Consultancy or Advisory role for Biocartis; Pierre Fabre; BMS; Servier; MSD; Founder and ownership interests in Methys Dx.\nFor MBS, CM, SG and EM: employees at Integragen SA.\nFor HA: Speaker, Consultancy or Advisory role for GlaxoSmithKline.\nVT : Founder, president and Chief scientific officer, and ownership interests in Methys Dx.","formattedTitle":"Improved Tumor-Type informed compared to Tumor-Informed Mutation Tracking for ctDNA Detection and Microscopic Residual Disease Assessment in Epithelial Ovarian Cancer","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eEpithelial ovarian cancer (EOC) is the eighth most common cancer in women worldwide (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Due to unspecific symptoms, patients are often diagnosed at advanced stage (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). According to the International Federation of Gynecology and Obstetrics recommendations, the first-line treatment combines chemotherapy and cytoreductive surgery. Bevacizumab or Poly-ADP-ribose-polymerase inhibitors are commonly employed as maintenance treatments (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e). After surgery, the absence of macroscopic residual disease is the major prognostic factor for disease curing (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). Despite first-line treatment and maintenance therapy, the 5-year survival rate remains below 40% with frequent relapses (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). Current follow-up methods, such as imaging and CA125 blood monitoring lack of sensitivity, hampering early recurrence detection and therapeutic response evaluation (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). Indeed, assessing microscopic residual disease (MRD) after treatment could improve prognostication and therapeutic management of patients (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). A reliable, non-invasive and sensitive monitoring approach of MRD is critically needed.\u003c/p\u003e \u003cp\u003eCirculating tumor DNA (ctDNA) shed by tumor cells into the bloodstream, offers a promising approach for both tumor burden assessment and characterization of tumor genomic alterations (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). It could improve the disease monitoring, including screening, treatment response assessment, MRD evaluation after surgery and early detection of recurrence (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe ctDNA detection is achieved through either tumor-informed or agnostic strategies. The tumor-informed approach, first identifies patient-specific tumor alterations by analyzing the tumor tissue and then tracks these alterations in DNA extracted from plasma (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). This method offers high sensitivity but is resource intensive. In contrast, the tumor-agnostic approach bypasses the need for tumor tissue by leveraging knowledge of common cancer alterations to create a standardized, one-size-fits-all test (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). This method can employ broad genomic panels, such as whole-exome sequencing (WES) and whole-genome sequencing (WGS) or focus on specific genes using PCR based methods such as digital PCR. While this approach offers advantages in turnaround time and cost-effectiveness and has proven high sensitivity for therapeutic response analysis, it generally sacrifices sensitivity compared to the tumor-informed approach (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn 2020, Zviran et al. modeled the probability of detecting ctDNA based on the number of genetic targets, the cfDNA input and the ctDNA fraction (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Their study demonstrated a clear benefit of increasing the number of targets, with detection of ctDNA fractions as low as 10⁻\u003csup\u003e5\u003c/sup\u003e achievable using tens of nanograms of cfDNA input when up to 10,000 targets are analyzed. In EOC, the average tumor mutation burden is between 2 and 3 mutations per mega base, enabling the identification of around 9,000 mutations through WGS (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). However, WGS is cost-prohibitive for many applications. WES, which reduces the volume of sequencing data, directly correlated to cost, identifies only a hundred mutations, significantly limiting the sensitivity of tumor-informed strategies for ctDNA detection around a ctDNA fraction of 10⁻⁴.\u003c/p\u003e \u003cp\u003eWe aimed to develop a method that focuses on common alterations specific to a particular cancer type. By incorporating a sufficient number of these alterations, this approach could achieve sensitivity comparable to tumor-informed methods based on WGS identification of alteration, while maintaining the versatility of a one-size-fits-all test for that specific cancer type. This can be completed by studying epigenetic marks specific to a tumor type. This method can be named as tumor-type informed and tumor-patient agnostic.\u003c/p\u003e \u003cp\u003eDNA methylation offers an avenue for ctDNA detection (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). Recent advances identified distinct methylation signatures across cancer types, endeavor to develop ctDNA detection strategies (\u003cspan additionalcitationids=\"CR15 CR16\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Tumor-type informed approach was applied in three key steps. First, we identified specific epigenetic markers of EOC tissue using enzymatic conversion of unmethylated cytosines (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), an enrichment panel, and next-generation sequencing (NGS). This approach enabled the identification of several thousand of differentially methylated CpGs within a reduced genomic window, distinguishing EOC tissue from healthy ovaries and peripheral blood mononuclear cells (PBMCs). Second, we evaluated the performance of these markers in a series of plasma samples from EOC patients and compared the results to a tumor-informed approach, which involved WES of solid tumors and mutation tracking in plasma samples. Finally, we assessed the potential of this method to detect MRD following the completion of first-line therapy.\u003c/p\u003e"},{"header":"2 Materials \u0026 Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSample Collection\u003c/h2\u003e \u003cp\u003eThe patients included in this study derived from the BIOVAIRE cohort (Paris, France). Of the 67 patients of the BIOVAIRE cohort, 47 were included in this study. Twelve patients were excluded due to a non-EOC diagnosis confirmed after tissue biopsy analysis, and 8 were excluded because of unavailable baseline plasma sample. All participants provided informed consent, and the study received ethical approval (CPP n\u0026deg;: 2013-10-01 and ID-RCB: 2013-A01108-37). Tumor tissues were stored in RNAlater at -80\u0026deg;C, and blood samples were collected in Streck tubes and were processed for plasma and PBMC isolation.\u003c/p\u003e \u003cp\u003eNormal ovarian tissues were obtained from OriGene (Herford, Germany) and classified as non-cancerous by two independent pathologists. Samples were flash-frozen in optimal cutting temperature compound (OCT) and stored at -80\u0026deg;C. Publicly available whole-genome bisulfite sequencing (WGBS) datasets of normal ovarian tissues were downloaded from the ENCODE database (Accession: ENCSR417YFD and ENCSR803SIO). Blood collected in Streck tubes from healthy women donors were provided by Biopredic International (UP, India) or l\u0026rsquo;\u0026Eacute;tablissement Fran\u0026ccedil;ais du Sang (EFS, Evry, France). DNA from ovarian tissues and PBMC was extracted using the Qiagen DNeasy Blood \u0026amp; Tissue Kit.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003e Enzymatic-targeted methylation protocol and DMR analysis\u003c/h3\u003e\n\u003cp\u003eLibraries were prepared with the NEBNext Enzymatic Methyl-seq kit with 100 ng of input DNA, followed by targeted hybrid-capture using the Twist Human Methylome Panel (Twist Bioscience). Sequencing was performed on an Illumina NovaSeq 6000 using 2x100 bp paired-end mode. Reads were processed using trim_galore (v. 0.6.6), BWAmeth (v. 0.2.7), Picard MarkDuplicates (v. 2.22.8), and methylation calling with MethylDackel (v. 0.6.0). CpG methylation status was compared between ovarian tumor and normal ovarian tissues, as well as between ovarian tumor and PBMCs from the same individuals. Differentially methylated loci (DMLs) were identified using DSS R package (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e) and MethylKit R package (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e), while differentially methylated regions (DMRs) were detected using DMRseq R package (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). A threshold of \u0026ge;\u0026thinsp;30% methylation difference was applied, with significance criteria set at q-values\u0026thinsp;\u0026lt;\u0026thinsp;0.001 for DMLs and \u0026lt;\u0026thinsp;0.01 for DMRs. CpG sites that were found differentially methylated in at least two of the three packages were selected for panel design. Subsequently, DMRs were redefined by clustering the selected differentially methylated CpGs, requiring a minimum of four CpGs per region, each separated by less than 100 bp. Regions within 50bp of each other were merged. For the panel design, DMRs overlapping repetitive or low-complexity regions were excluded.\u003c/p\u003e\n\u003ch3\u003eTargeted Methylation Sequencing of Plasma Samples\u003c/h3\u003e\n\u003cp\u003eCfDNA was extracted from 3 mL of plasma using the QIAamp Circulating Nucleic Acid Kit. For cfDNA quantification, 1 \u0026micro;L of the 30 \u0026micro;L elution volume was analyzed on a Fragment Analyzer (Agilent), using the NGS Fragment kit (1-6000 bp). Half of the remaining elution volume, corresponding to 1.1 to 25 ng (mean 8.8 ng, IQR 3.5\u0026ndash;10.6) was processed using the NEBNext Enzymatic-Methyl-seq kit, with UMI methylated adapters (Twist Bioscience). Libraries were hybrid-captured with the custom-panel targeting identified CpG sites, sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode and aligned to human genome (hg38) with BWAmeth (v. 0.2.7) after data normalization to 1.5 Gb per sample. Duplicates reads were managed using fgbio (v. 2.3.0) to distinguish real duplicates from collisions (i.e., distinct fragments sharing the same genomic coordinates) thanks to molecular barcodes. Methylation patterns at the resolution of individual reads (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e) were identified using wgbstools (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). A fragment is considered to originate from a tumor if at least 85% of the differentially methylated loci (DMLs) within the fragment share the same status as defined by the analysis of DNA extracted from EOCs. For each DMR, a beta value was computed, representing the proportion of plasma-derived fragments associated with a tumor-derived origin. The standard deviation of this beta value was then calculated for each DMR across a cohort of 24 healthy plasma donors. Only DMRs with a standard deviation\u0026thinsp;\u0026le;\u0026thinsp;0.05 were retained. Subsequently, the area under the receiver operating characteristic curve (AUC) was determined for each DMR by comparing the beta values between the 24 healthy plasma donors and 35 plasma samples from patients at baseline, which were presumed to contain ctDNA. DMRs with an AUC\u0026thinsp;\u0026ge;\u0026thinsp;0.65 were selected for further analysis.\u003c/p\u003e \u003cp\u003eThe Support Vector Machine classifier was trained using beta values from selected DMRs as features. The training dataset consisted of the same 24 plasma samples from healthy donors and 35 baseline plasma samples used for DMR selection. Once trained, the model was used to classify new plasma samples based on their predicted values. A sample was classified as ctDNA-positive if its predicted value exceeded 0, while a predicted value below 0 indicated a ctDNA-negative sample.\u003c/p\u003e\n\u003ch3\u003ectDNA proportion assessment with the tumor-type informed approach\u003c/h3\u003e\n\u003cp\u003eTo assess the ctDNA proportion we selected DMRs with a background noise level below 1 in 10,000 reads across the 54 healthy plasma donors, resulting in a final selection of 201 DMRs. Then we calculated the ratio of sequencing reads exhibiting the EOC-specific methylation pattern to the total sequencing reads overlapping the genomic coordinates of the 201 DMRs in samples classified as ctDNA-positive\u003c/p\u003e\n\u003ch3\u003eDNA mixes used for ctDNA quantification validation in the tumor-type informed approach\u003c/h3\u003e\n\u003cp\u003eTwelve DNA extracted from ovarian tumor were combined with DNA extracted from matched PBMC at a 1:99 ratio. The tumor cellularity in the 12 tumor samples used for the mixes was determined by copy number alteration (CNA) analysis. CNA analysis was performed on sequencing data generated for the identification of genetic alterations (see below) using GATK (v. 4.1.4.1). The expected tumor DNA fraction in the 12 mixes was calculated using the formula: tumor cellularity divided by 100. These samples were used to validate ctDNA quantification methods in the tumor-type informed approach. DNA mixes were processed using the NEBNext Enzymatic-Methyl-seq kit, with UMI methylated adapters (Twist Bioscience) and 20 ng of fragmented DNA input. Libraries were hybrid-captured with the custom-panel targeting identified CpG sites, sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode and followed the same analysis as described for plasma samples.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eIdentification of Genetic Alterations\u003c/h2\u003e \u003cp\u003eTumor and matched PBMC DNA from 12 patients underwent WES using the Twist Human Core Exome Panel (Twist Bioscience) with an input of 100 ng. Libraries were sequenced in paired-end 2x100 mode on an Illumina Novaseq 6000. Reads were aligned to the human genome (hg38) with BWA (v. 0.7.15), and duplicates removed using Sambamba (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e). Variant calling for constitutional DNA (PBMCs) used GATK HaplotypeCaller (v. 3.8.1), while MuTect2 (v. 2.0) was used for tumor DNA, followed by post-processing to filter artifacts and germline variants. Variants were annotated with VEP (v. 101).\u003c/p\u003e \u003cp\u003eSingle nucleotide variants (SNVs) for the ctDNA detection panel were selected based on somatic score, tumor variant allele frequency (VAF, \u0026ge;\u0026thinsp;5%), and low population frequency (\u0026le;\u0026thinsp;0.5%). Additionally, each SNVs was evaluated in each patient\u0026rsquo;s PBMC to exclude those potentially linked to clonal hematopoiesis.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eDetection of ctDNA in Plasma based on genetic alterations\u003c/h3\u003e\n\u003cp\u003eCell-free DNA from the remaining elution volume of plasma extractions, corresponding to a mean input of 9 ng (IQR 2.6\u0026ndash;9.7) was used for mutation analysis with the Twist Mechanical Fragmentation Kit (Twist Bioscience). To enhance the detection of low-frequency SNVs, we added molecular barcodes during library preparation, enabling the generation of consensus reads after sequencing. Libraries were target-enriched for tumor-specific SNVs genetic coordinates using designed panels with Twist MRD500 panels solution and sequenced on an Illumina NovaSeq X using 2x150 bp paired-end mode. Data processing involved molecular barcodes management with fgbio (v. 2.3.0), alignment with BWA (v. 2.2.1), and variant calling using samtools mpileup (v. 1.9). To assess the reliability of ctDNA detection, the frequency of reads carrying tumor-specific SNVs was compared to the frequency of reads carrying unspecific SNVs at genomic positions confirmed to be non-mutated in the tumor and PBMC of each patient. For each tumor-SNV position, a p-value was calculated using the binomial probability distribution \u0026#119875;(\u0026#119883;=nb.tumor-SNV ∣ Depth, Freq.unspecific-SNV) where:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003enb.tumor-SNV represents the number of reads carrying a tumor-specific SNV,\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eDepth is the total number of reads overlapping the genomic coordinate of the tumor-SNV\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eFreq.unspecific-SNV corresponds to the sequencing error rate, estimated for each base type and each library using genomic regions confirmed to be free of SNVs in both the tumor and PBMC of the patient (i.e. the distribution under null hypothesis \u003cem\u003eH\u003c/em\u003e\u003csub\u003e\u003cem\u003e0\u003c/em\u003e\u003c/sub\u003e).\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eTumor-SNV with p-value\u0026thinsp;\u0026ge;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e were considered non-relevant. A sample was classified as ctDNA positive when at least two tumor-SNVs were detected with p-values\u0026thinsp;\u0026le;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e, corresponding to a false positive probability\u0026thinsp;\u0026le;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;6\u003c/sup\u003e. This method enables a statistical evaluation of tumor-derived ctDNA detection while accounting for background sequencing noise.\u003c/p\u003e\n\u003ch3\u003ectDNA proportion assessment with the tumor-informed approach\u003c/h3\u003e\n\u003cp\u003eCtDNA quantification using the tumor-informed approach was performed by calculating the ratio of sequencing reads carrying tumor-SNVs to the total number of reads overlapping the genomic coordinates of all tumor-SNVs identified in the patient\u0026rsquo;s tumor tissue, for samples classified as ctDNA-positive.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were performed in R (version 4.0.2) environments. Machine Learning model was developed in Python (version 3.7.2) with scikit-learn 1.0.2 package, numpy 1.21.6, pandas 1.3.5 and seaborn 0.12.2. Deconvolution of methylation for cellular origin calculations was performed using UXM_deconvolution (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). Progression-free survival (PFS) and overall survival (OS) were compared between patient groups based on ctDNA presence. PFS was defined as the time from the end of first-line therapy to tumor recurrence. OS were defined as the time from the initial diagnosis to the death from any cause. Kaplan-Meier curves were used to visualize differences in PFS and OS, and a log-rank test was performed to compare the two groups.\u003c/p\u003e \u003c/div\u003e"},{"header":"3 Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eCharacteristics of the study population\u003c/h2\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003ePatients\u003c/h2\u003e \u003cp\u003eBaseline characteristics of the 47 patients included in this study are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The median age at diagnosis was 66 (IQR 60\u0026ndash;71). Patients received standard-of-care treatment consisting of primary cytoreductive surgery followed by adjuvant chemotherapy (n\u0026thinsp;=\u0026thinsp;23), neoadjuvant chemotherapy before interval cytoreductive surgery followed by adjuvant chemotherapy (n\u0026thinsp;=\u0026thinsp;17), or chemotherapy alone for 7 patients. The median follow-up time for the 47 patients is 84 months (IQR [35.8-not reached]). Among included patients, 33 were analyzed at multiple time points, resulting in a total of 63 plasma samples collected while chemotherapy (V2-V7, Supplemental Fig.\u0026nbsp;1).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eHeathly subjects\u003c/h2\u003e \u003cp\u003e Plasmas obtained from Biopredic International were sourced from women with a median age of 64 years (IQR 53\u0026ndash;75) and without gynecological diseases. Additional plasma samples were provided by l\u0026rsquo;\u0026Eacute;tablissement Fran\u0026ccedil;ais du Sang (EFS), collected from women aged 40 to 50 years, with no reported diseases. Healthy ovarian tissue samples were acquired from Origene (n\u0026thinsp;=\u0026thinsp;7), with donors having a median age of 52 years (IQR 45\u0026ndash;59). To enhance the dataset, we included publicly available whole-genome bisulfite sequencing (WGBS) data from the ENCODE database, derived from two ovarian tissue donors aged 30 and 51 years.\u003c/p\u003e \u003cp\u003e \u003cb\u003eCharacterizing genetic and epigenetic alterations of ovarian tumors, PBMC and healthy ovaries for custom panels design\u003c/b\u003e \u003c/p\u003e \u003cp\u003eWe aimed to compare the capabilities of patient tumor-informed and tumor-type informed approaches in detecting and quantifying ctDNA. For the patient tumor-informed approach, we performed WES on DNA extracted from ovarian tumors and matched PBMCs from 12 patients from the BIOVAIRE cohort. We found between 16 and 204 SNVs per patient (mean\u0026thinsp;=\u0026thinsp;74). Eight out of the 12 patients showed a mutated allele of \u003cem\u003eTP53\u003c/em\u003e (67%), 4 had mutations in \u003cem\u003eTTN\u003c/em\u003e (33%), and 3 in \u003cem\u003eCSMD3\u003c/em\u003e (25%). However, no recurrent hotspot mutation was identified (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA, upper panel; see Methods for details).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCharacteristics of the 47 study patients.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eN\u0026thinsp;=\u0026thinsp;47\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge at diagnostic (median (IQR))\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e66 (60, 71)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI (median (IQR))\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e22.48 (20.8\u0026nbsp;; 27.66)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMenopausal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e41 (87%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBRCA mutation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e37 (78.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBRCA 1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBRCA 2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNon investigated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 (12.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eASA score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10 (21.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e21 (44.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7 (14.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo data available\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9 (19.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHistology\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClear cell adenocarcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEndometrioid carcinoma (grade I and II)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLow grade serous carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026nbsp;High grade serous carcinoma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e42 (89.4%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO stage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO II B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO III\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 (6.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO III B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (8.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO III C\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25 (53.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFIGO IV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14 (29.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSurgical resection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e44 (88%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCC0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e38 (80.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCC1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCC2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e30-day postoperative mortality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2 (4.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTherapeutic sequence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUpfront surgery followed by adjuvant chemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23 (49%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNeoadjuvant chemotherapy followed by interval surgery V3/V4/V5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16 (34%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNeoadjuvant chemotherapy followed by surgery after V6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChemotherapy alone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7 (15%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTotal number of chemotherapy cycles (median; IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 (6 ; 7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFollow-up in months (median; IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e84 (35.75\u0026nbsp;; not reached)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRecurrence rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26 (55.3%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime from last chemotherapy to recurrence (median; IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11.7 (9.43\u0026nbsp;; 16.1)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMortality rate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17 (36%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLost to follow-up\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTime from death to last chemotherapy (median; IQR)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.2 (17.6\u0026nbsp;; 46.6)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003csup\u003e\u003cem\u003e1\u003c/em\u003e\u003c/sup\u003e\u0026nbsp;\u003cem\u003eMedian (iqr); n (%)\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eIQR: interquartile range; BMI: body mass index; FIGO: international federation of gynecology and obstetrics; BRCA: breast cancer gene; CC: cytoreduction score.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFor the tumor-type informed approach, we endeavored to identify specific methylation patterns of EOC using enzymatic conversion and library enrichment on a panel targeting 3,98\u0026nbsp;million CpGs by comparing tumor samples (n\u0026thinsp;=\u0026thinsp;12) with healthy ovarian tissues (n\u0026thinsp;=\u0026thinsp;7) and PBMCs (n\u0026thinsp;=\u0026thinsp;12). PBMC samples were analyzed because the majority of cfDNA present in plasma samples originates from these cells. The two WGBS datasets were added to the healthy ovarian tissues group. Based on this approach, we identified 52,173 CpGs that were consistently differentially methylated in the same direction (either hypomethylated or hypermethylated) when comparing ovarian tumors to both healthy ovarian tissues and PBMCs (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA, bottom panel). The distribution of DMLs across genomic regions revealed distinct patterns of hypermethylation and hypomethylation (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). In CpG islands, 5,179 CpGs were hypermethylated, compared to 1,574 that were hypomethylated, highlighting a pronounced bias toward hypermethylation in these regions. In contrast, open sea regions exhibited a predominant pattern of hypomethylation, with 23,239 CpGs hypomethylated versus 6,730 hypermethylated. In the shores and shelves of CpG islands, the distributions of hypermethylated and hypomethylated CpGs were balanced. Specifically, CpG island shores contained 4,756 hypermethylated and 4,751 hypomethylated CpGs, while shelves showed a slight bias toward hypomethylation, with 3,042 hypomethylated and 2,902 hypermethylated CpGs. These findings underscore the region-specific nature of DNA methylation alterations in ovarian tumors, with hypermethylation predominantly occurring in CpG islands and hypomethylation prevailing in open sea regions.\u003c/p\u003e \u003cp\u003eTo assess whether the identified DMLs could differentiate ovarian tumor samples from healthy ovarian tissues, we selected the 2,000 DMLs with the highest standard deviation of methylation value across samples to generate unsupervised clustering. This clustering revealed three distinct clusters of samples, with healthy ovarian tissues clustering together. The distance between healthy samples and tumor samples increases with tumor cellularity and tumor percentage of ovarian epithelium component (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eC, Supplemental Fig.\u0026nbsp;2). Thus, the selected DMLs enabled a clear classification of ovarian tissues and were selected for the design of a unique custom panel of 2.90 Mb.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eTumor-informed approach for ctDNA characterization\u003c/h2\u003e \u003cp\u003eAfter identifying tumor-SNVs and designing the custom panels, we analyzed plasma samples (n\u0026thinsp;=\u0026thinsp;28) from the twelve patients at different time-points: baseline (V1, n\u0026thinsp;=\u0026thinsp;12), during treatment (V2-V5, n\u0026thinsp;=\u0026thinsp;11), and at the end of treatment (V6-V7, n\u0026thinsp;=\u0026thinsp;5). We generated an average of 11.66 Gb of data per sample (range: 7.65\u0026ndash;15.56 Gb). The presence of ctDNA in plasma samples was assessed by comparing the frequency of reads carrying tumor-specific SNVs to the background sequencing error rate for each nucleotide, which was determined at genomic positions without variations in both tumor and PBMC samples from the corresponding patients (Supplemental Fig.\u0026nbsp;3). This method provided a robust framework for evaluating the reliability of ctDNA detection in each sample. Among the 28 plasma samples analyzed, 19 were ctDNA positive with at least 2 tumor-SNVs with a p-value\u0026thinsp;\u0026lt;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eFigure 2A illustrates the baseline distribution of tumor-SNVs VAF in plasma for each patient. Among the 12 patients, 11 were classified as ctDNA-positive, with a mean VAF of tumor-SNVs of 1.30% (range: 0.03\u0026ndash;3.39%). Patients undergoing neoadjuvant chemotherapy exhibited a higher mean VAF of 2.70% (95% CI [1.45\u0026ndash;3.95]) compared to those undergoing primary cytoreductive surgery, who had a mean VAF of 0.46% (95% CI [0.01\u0026ndash;0.91]) (Wilcoxon test, p-value = 9 \u0026times; 10⁻⁴, Fig. 2B). Among plasma samples collected at different time points (n = 16) from 8 of the 12 patients, 8 samples were ctDNA-positive. Importantly, all patients showed a significant decrease in ctDNA levels during chemotherapy (Fig. 2C). Two out of 5 patients remained ctDNA-positive at the end-of-treatment time points (V6\u0026ndash;V7).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eTumor-type informed approach for ctDNA characterization\u003c/h2\u003e \u003cp\u003eTumor-type informed approach is based on the training and validation of a support vector machine (SVM) classifier. We defined a training group of 35 baseline plasmas from EOC patients and 24 plasmas from healthy subjects, used for DMR refinement and SVM model training. The validation group consisted of 28 plasma samples, from the twelve same patients analyzed with the tumor-informed approach and 30 new plasma samples from different healthy subjects. This validation group was used both to assess the sensitivity and specificity of the SVM classifier and to compare the tumor informed and tumor-type informed strategies (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003eUsing DMLs from the tumor-type informed custom panel, we defined DMRs based on the principle of methylation coherence among closely spaced CpGs (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e). Briefly, we extended genomic regions by 100bp in each direction from DMLs, and merged the regions within 50bp of each other. A DMR was considered only if it contained at least 4 CpGs. Using this procedure 11,260 DMRs were defined, spanning a genomic region of 3,4 Mb.\u003c/p\u003e \u003cp\u003eAfter the targeted-methylation enrichment and sequencing of cfDNA extracted from the plasma samples, we performed horizontal analysis of cell-free DNA, focusing on the individual fragment methylation patterns rather than the average methylation across all fragments at each DML (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). The tumor-derived cfDNA is identified based on the similarity of the methylation status of the different DMLs it contains. A fragment is attributed to a tumor-derived origin if at least 85% of the present DMLs share the same status as that defined by the analysis of EOCs. The beta value represents the proportion of tumor-derived fragments in the plasma sample for each DMR. In the first step, we analyzed the plasma samples from healthy subjects in the training group to calculate the beta value standard deviation for the selected DMRs. We then excluded 1,528 DMRs with a beta value standard deviation greater than 0.05. In the second step, we evaluated the ability of each DMR's beta value to differentiate between healthy and cancer plasma samples by calculating the area under the curve (AUC).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eA total of 6,286 DMRs, 1,470 hypermethylated DMRs and 4,816 hypomethylated DMRs with an AUC\u0026thinsp;\u0026ge;\u0026thinsp;0.65 were retained (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC). On these selected DMRs, beta values were increased for baseline patient samples compared with healthy controls (Supplemental Fig.\u0026nbsp;4). These DMRs were used to train a SVM model for samples classification between healthy and cancer patients which is considered as a surrogate marker of the absence or the presence of ctDNA, respectively. To quantify ctDNA by the tumor-type informed method, we used healthy samples from the training group to identify the background noise of beta value and selected a reduced set of DMRs (n\u0026thinsp;=\u0026thinsp;201). We validated the method using synthetic samples (n\u0026thinsp;=\u0026thinsp;12) with known tumor DNA quantities (Supplemental Fig.\u0026nbsp;5, see Methods). The correlation between the expected and measured tumor DNA fractions yielded an R of 0.85 (p\u0026thinsp;\u0026lt;\u0026thinsp;10⁻\u0026sup3;).\u003c/p\u003e \u003cp\u003eThe SVM model was then tested on the validation group. All healthy subjects were classified as ctDNA negative whereas all but one cancer baseline plasma (V1) patients were considered as ctDNA positive, leading to a theorical sensitivity of 91.7%, a specificity of 100% and an accuracy of 97.6% (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). The mean ctDNA fraction of positive samples was 1.27% (range: 0.04\u0026ndash;3.77%). Interestingly, the patient with a tumor-informed ctDNA negative status at baseline was considered as ctDNA positive with the tumor-type informed method, with ctDNA fractions of 0.08. The negative one by this latter method was considered as ctDNA positive by the tumor-informed approach with a tumor-SNV VAF of 0.15%. The correlation of ctDNA fraction obtained with tumor-type informed approach with the tumor-informed approach is shown on Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB, with an R of 0.53 (Spearman correlation, p\u0026thinsp;=\u0026thinsp;\u003cem\u003e3.4\u003c/em\u003e x \u003cem\u003e10\u003c/em\u003e\u003csup\u003e\u003cem\u003e\u0026minus;\u0026thinsp;3\u003c/em\u003e\u003c/sup\u003e).\u003c/p\u003e \u003cp\u003eWhen we considered the follow-up plasmas of this cohort of validation collected at different time points during chemotherapy (V2\u0026ndash;V5, n\u0026thinsp;=\u0026thinsp;11), 7 samples were classified as ctDNA-positive by the SVM classifier. Four out of 5 patients remained ctDNA-positive at the final time point (V6\u0026ndash;V7).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eLike the tumor-informed approach, all patients exhibited a significant decrease in ctDNA levels during chemotherapy (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC). Of the patient plasma samples, 67.8% were concordant for ctDNA status between the tumor-informed and tumor-type informed strategies (n\u0026thinsp;=\u0026thinsp;19), 21.4% were positive only with the tumor-type informed approach (n\u0026thinsp;=\u0026thinsp;6), and three were positive only with the tumor-informed approach (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003eContingency table comparing results of tumor-informed and tumor-type informed approaches for ctDNA detection.\u003c/b\u003e\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePositives samples tumor-informed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNegative samples tumor-informed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTotal\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePositive samples tumor-type informed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNegative samples tumor-type informed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eTotal\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003e19\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e9\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e28\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn the tumor-informed approach, samples are classified as ctDNA-positive when at least 2 tumor-SNVs with a p-value\u0026thinsp;\u0026lt;\u0026thinsp;10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003e are detected. For the tumor-type informed approach, samples are classified as ctDNA-positive if the predicted value from the SVM classifier is positive.\u003c/p\u003e \u003cp\u003eThese results highlighted the comparable sensitivity of the developed tumor-type informed approach with the tumor-informed one (McNemar test, p\u0026thinsp;=\u0026thinsp;0.51), with a tendency to a better detection of ctDNA at the last time points with the tumor-type informed approach. We illustrated that using horizontal analysis on specific DMRs, we were able to detect ctDNA with a similar sensitivity than with the tumor-informed approach with a reduced quantity of data.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eOver-fitting assessment of the Support Vector Machine classifier\u003c/h2\u003e \u003cp\u003eWe evaluated our machine learning model for overfitting using 47 unused follow-up plasma samples from 25 patients. These samples came from the same patients whose baseline plasma samples were used to train the SVM model. We showed that 70.2% of the samples (33 out of 47) were classified as ctDNA positive, comparable to that obtained on the follow-up plasma from the validation group (68.8%; 11 out of the 16 samples). Next, combining all samples except those used in the training of the SVM model, we examined whether cfDNA input in library preparation could influence the predicted value of the model for a sample and found no significant correlation (Spearman\u0026rsquo;s test \u003cem\u003eR = -5.9\u003c/em\u003e\u003cb\u003e\u0026times;\u003c/b\u003e\u003cem\u003e10\u003c/em\u003e\u003csup\u003e\u003cem\u003e\u0026minus;\u0026thinsp;2\u003c/em\u003e\u003c/sup\u003e, \u003cem\u003ep\u0026thinsp;=\u0026thinsp;0.95\u003c/em\u003e). The model tended to separate healthy plasma samples from plasma patients with better accuracy when cfDNA input increases (Supplemental Fig.\u0026nbsp;6). Finally, on the 54 healthy plasma samples, we assessed for each DMR the association between beta values and age of healthy subjects and found no significant correlation (Supplemental Fig.\u0026nbsp;7).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eClinical validation of MRD assessment using the SVM classifier\u003c/h2\u003e \u003cp\u003eOn the 47 supplementals follow-up plasmas samples from 25 patients, (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA), we assessed the capacity of the model to detect MRD. Results of ctDNA detection and quantification from this group of plasma not analyzed with the tumor-informed approach are presented in Supplemental Fig.\u0026nbsp;8. Most patients showed a decrease in ctDNA fraction along treatment (21 out of the 25 patients tested).\u003c/p\u003e \u003cp\u003eWe assessed the clinical relevance of our tumor-type informed approach for predicting EOC outcomes by analyzing plasma samples collected at the end of treatment. Among 22 patients tested for MRD detection at V6 (n\u0026thinsp;=\u0026thinsp;7) or V7 (n\u0026thinsp;=\u0026thinsp;15), 16 (72.7%) were classified as ctDNA-positive, with a mean ctDNA fraction of 0.17%. Of these positive patients, 12 experienced EOC relapse, with a median progression-free survival (PFS) of 11 months (95% CI, 10.04\u0026ndash;28.38). In contrast, among the 6 ctDNA-negative patients, only 1 relapsed within 15 months, and the median PFS for this group was not reached. Kaplan-Meier analysis demonstrated significant differences in PFS between ctDNA-positive and ctDNA-negative patients at the end of chemotherapy (log-rank p\u0026thinsp;=\u0026thinsp;0.017, Hazard ratio\u0026thinsp;=\u0026thinsp;8.24; 95% CI [1.06\u0026ndash;64.4]; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003eNone of the ctDNA-negative patients at V6 or V7 died during follow-up, while 37.5% of ctDNA-positive patients (6 out of 16) succumbed during the surveillance period. Kaplan-Meier analysis also revealed significant differences in overall survival (OS) between ctDNA-positive and ctDNA-negative groups (log-rank p\u0026thinsp;=\u0026thinsp;0.041; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOverall, our tumor-type informed SVM model effectively predicted EOC relapses using reduced sequencing data compared to tumor-informed strategies.\u003c/p\u003e \u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eHighly sensitive methods are required for the detection and quantification of ctDNA in MRD assessment. In this study, we compared two approaches: a tumor-informed approach based on somatic SNVs identified from the patients tumor samples and a tumor-type informed approach utilizing common DMRs in EOC samples. These strategies were applied to 47 patients from the Biovaire cohort, which benefited from an extended follow-up period, with some patients monitored for up to 100 months after the completion of first-line therapy.\u003c/p\u003e \u003cp\u003eIn the tumor-informed approach, we performed WES on tumor samples, using constitutional DNA from PBMCs as a reference to identify tumor-specific SNVs. In EOC samples from TCGA (\u003cspan additionalcitationids=\"CR28 CR29 CR30 CR31\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e), the median number of SNVs detected by WES is 62.8 (IQR: 34\u0026ndash;77), close to that we observed for the 12 patients tested in our series by tumor informed approach (median 67, IQR: 43\u0026ndash;71). Most patients harbored mutations in commonly altered EOC genes, such as \u003cem\u003eTP53\u003c/em\u003e, \u003cem\u003eTTN\u003c/em\u003e, and \u003cem\u003eCSMD3\u003c/em\u003e (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). However, this relatively small number of SNVs limits ctDNA detection when plasma volume is restricted. On the 12 baseline plasma samples tested by the tumor-informed approach we extracted an average of 15.3 ng cfDNA, representing 4500 genome equivalents (GE). According to Zviran \u003cem\u003eet al\u003c/em\u003e., the probability of detecting ctDNA at a 10⁻⁴ fraction under these conditions is between 80%-100% (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). Consistent with this, 11 out of 12 samples were ctDNA-positive. The negative sample had one of the lowest cfDNA input for library preparation (6.1 ng). For the follow-up plasma samples, with an average cfDNA input of 4.2 ng, 8 out of 16 samples were ctDNA-positive. Sensitivity might be improved by increasing cfDNA input (which would require larger blood volumes) or by targeting variants more resistant to sequencing errors, such as indels or phased variants. However, this approach would significantly increase sequencing costs, particularly due to the need for WGS of the tumor to identify these specific variants (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). Optimizing the number of markers is thus critical to overcoming these limitations.\u003c/p\u003e \u003cp\u003eTo address this purpose, we evaluated whether cfDNA methylation analysis in a tumor-type informed context could enhance ctDNA detection while remaining cost-effective. Publicly available datasets, primarily derived from Illumina\u0026rsquo;s 27K methylation chip, have limited potential for biomarkers discovery. To overcome this, we generated a comprehensive dataset using a methylation panel covering\u0026thinsp;~\u0026thinsp;14% of the CpGs in the human genome. To enhance robustness, we supplemented this dataset with two ENCODE WGBS files from healthy ovarian tissues. In total, we identified 6,286 DMRs, which were used for the design of a single panel for tumor-type informed cfDNA sequencing. Several genes associated with these DMRs, such as \u003cem\u003eCELF2, FAIM2, SIM2, CDO1\u003c/em\u003e, and \u003cem\u003eC2CD4D\u003c/em\u003e, are already described as potential EOC biomarkers. Others, such as \u003cem\u003eJAK3\u003c/em\u003e and \u003cem\u003eSOX2\u003c/em\u003e, are known tumor driver genes (\u003cspan additionalcitationids=\"CR38\" citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). By identifying more than 6000 markers, this extensive panel increased the theoretical limit of ctDNA detection to 10⁻⁵ for 2000 GE of cfDNA input.\u003c/p\u003e \u003cp\u003eWhen comparing the two strategies using baseline plasma samples, the proportion of ctDNA-positive samples was the same: 91.7%. However, in follow-up plasma samples, the tumor-type-informed method proved superior sensivity, detecting ctDNA in 11 out of 16 samples, whereas the tumor-informed approach detected only 8 ctDNA-positive samples. This innovative approach improves sensitivity, enabling detection of ctDNA at fractions as low as 0.05%, while reducing turnaround time, eliminating the need for tumor analysis, and significantly lowering costs by requiring a single methylation panel and eightfold fewer sequencing reads than the tumor-informed method.\u003c/p\u003e \u003cp\u003eHorizontal methylation analysis further strengthened ctDNA detection by reducing dependence on unmethylated cytosine conversion rates and sequencing errors (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e). By analyzing multiple CpGs per cfDNA fragment and integrating an SVM model, we achieved 100% specificity in ctDNA detection. Furthermore, the tumor-type informed method allows the identification of MRD of patients with a sensitivity of 92.3%, 12 patients relapse when ctDNA was detected at the end of adjuvant treatment with a median time of relapse of 11 months. The specificity of the test is lower (55.6%) but 2 patients were lost to follow-up before reaching two years.\u003c/p\u003e \u003cp\u003eThis study has some limitations. First, the DMRs selection was based on a small cohort comprising 12 EOC tissues, 12 matched PBMC samples, and 9 healthy ovarian tissues. Although we studied a limited number of patient tumors to identify our DMRs, which does not fully protect against inter-individual variability in tumor profiles, the overall sensitivity of our method for detecting ctDNA at baseline is 97.9%. While we accounted for age, cfDNA input, and patient variability, the possibility of overfitting in the SVM cannot be excluded even if its likelihood is weak. Finally, the performance of our test to predict EOC relapse at the end of adjuvant treatment was validated in only 22 patients. Larger independent cohorts are required to confirm these findings before initiating clinical trials to assess its clinical utility.\u003c/p\u003e \u003cp\u003eCompared to other studies, the tumor-type informed approach demonstrated similar sensitivity and specificity to the study by Hou \u003cem\u003eet al\u003c/em\u003e., which employed deep whole-exome sequencing (WES) to identify tumor-specific SNVs and track them using an amplicon-based method (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e). In contrast, the tumor-na\u0026iuml;ve assay developed by Hoe \u003cem\u003eet al\u003c/em\u003e., which monitors ctDNA through nine frequently mutated genes, detected ctDNA in only 69.2% of baseline samples. While their method accurately predicted relapse in many cases, some patients classified as negative ultimately relapsed, underscoring a notable lack of sensitivity (\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur study holds promises for improving EOC patient management. Stratifying patients as MRD-positive or -negative could enhance quality of life by sparing unnecessary maintenance treatments for some while intensifying therapeutic efforts for others with MRD evidence. The tumor-type informed approach is cost-effective, time-efficient, and eliminates the need for tumor biopsies. It is also adaptable to other cancer types and could be a valuable tool for evaluating treatment responses in clinical assays.\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eBy analyzing cfDNA methylation patterns in plasma from healthy donors and EOC patients, we achieved improved sensitivity for detecting MRD and predicting relapse compared to a traditional tumor-informed approach. Notably, this approach allows classification of plasma samples regarding ctDNA presence without requiring prior knowledge of the patient\u0026rsquo;s tumor SNVs. It achieves this purpose with eightfold less data than tumor-informed strategies that rely on WES of the patient\u0026rsquo;s tumor and mutation tracking in plasma. This cost-effective and non-invasive method holds significant potential to improve EOC patient care. For patients deemed free of MRD following first-line treatment, a de-escalation of maintenance therapy can be envisioned. Conversely, MRD-positive patients could benefit from closer monitoring and tailored interventions, potentially improving outcomes. Future studies should validate these findings in larger, multi-center cohorts, paving the way for broader clinical implementation.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eEthics approval and consent to participate\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStudy protocol was approved by the ethic committee (Comit\u0026eacute; de Protection des Personnes \u0026ndash; Ile de France II\u0026nbsp;\u0026raquo;, Research number: 2013-10-01 MS1, Biovaire Study n\u0026deg; ID-RCB: 2013-A01108-37. All patients recruited in the study signed an informed written consent\u003cstrong\u003e\u003cem\u003e.\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eConsent for publication\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAvailability of data and materials\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the\u003cins cite=\"mailto:justine%20azais\" datetime=\"2025-01-28T16:00\"\u003e\u0026nbsp;\u003c/ins\u003e\u003cdel cite=\"mailto:justine%20azais\" datetime=\"2025-01-28T16:00\"\u003e\u0026nbsp;\u003c/del\u003ecorresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eCompeting interests\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor PLP: President of Canceropole Ile de France; Director of the \u0026laquo;\u0026nbsp;Institut du cancer Paris Carpem; Speaker, Consultancy or Advisory role for Biocartis; Pierre Fabre; BMS; Servier; MSD; Founder and ownership interests in Methys Dx.\u003c/p\u003e\n\u003cp\u003eFor MBS, CM, SG and EM: employees at Integragen SA.\u003c/p\u003e\n\u003cp\u003eFor HA: Speaker, Consultancy or Advisory role for GlaxoSmithKline.\u003c/p\u003e\n\u003cp\u003eVT : Founder, president and Chief scientific officer, and ownership interests in Methys Dx.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eFunding\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study is partly funded by l\u0026rsquo;Association Nationale de la Recherche et de la Technologie and IntegraGen.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAuthors\u0026apos; contributions\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMBS\u003c/strong\u003e have made substantial contributions to experimental design, library preparation for sequencing, data analysis, interpretation of results, and drafting of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHA\u003c/strong\u003e have made substantial contributions to the experimental design, the interpretation of data and revised the work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCM\u003c/strong\u003e have made substantial contributions to the conception, design of the work, interpretation of data and revised the work and the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSG\u003c/strong\u003e have made substantial contributions to the analysis of data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEM\u003c/strong\u003e have made substantial contributions to the conception and design of the works\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJA\u003c/strong\u003e have made substantial contributions to the conception and design of the works.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLB\u003c/strong\u003e have made substantial contributions to the analysis of clinical data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eADR\u003c/strong\u003e have made substantial contributions to the analysis of methylation data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEL\u003c/strong\u003e have made substantial contributions to interpretation of clinical data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCD\u003c/strong\u003e have made substantial contributions to the analysis of data regarding Kaplan-Meier Curve.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJM\u003c/strong\u003e have made substantial contributions to the conception and design of the works.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eBB\u003c/strong\u003e have made substantial contributions to the conception and design of the works.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAB\u003c/strong\u003e have made substantial contributions to the conception and design of the works.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVT\u003c/strong\u003e have made substantial contributions to the conception, design of the work, interpretation of data and substantively revised the work and the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePLP\u003c/strong\u003e lead the project and have made substantial contributions to the conception, design of the work, interpretation of data and substantively revised the work and the manuscript.\u003c/p\u003e\n\u003cp\u003eAll authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAcknowledgements\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWe thank the participating patients to BIOVAIRE cohort, Institut du cancer PARIS \u0026nbsp;CARPEM, and the SIRIC CARPEM for their support in the collection as well as EPIGENTEC Biobank. \u0026nbsp;\u003c/em\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin. 2021;71(3):209\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHennessy BT, Coleman RL, Markman M. Ovarian cancer. Lancet. 2009;374(9698):1371\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerek JS, Renz M, Kehoe S, Kumar L, Friedlander M. Cancer of the ovary, fallopian tube, and peritoneum: 2021 update. Int J Gynaecol Obstet. 2021;155(1):61\u0026ndash;85.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003edu Bois A, Reuss A, Pujade-Lauraine E, Harter P, Ray-Coquard I, Pfisterer J. Role of surgical outcome as prognostic factor in advanced epithelial ovarian cancer: a combined exploratory analysis of 3 prospectively randomized phase 3 multicenter trials: by the Arbeitsgemeinschaft Gynaekologische Onkologie Studiengruppe Ovarialkarzinom (AGO-OVAR) and the Groupe d\u0026rsquo;Investigateurs Nationaux Pour les Etudes des Cancers de l\u0026rsquo;Ovaire (GINECO). Cancer. 2009;115(6):1234\u0026ndash;44.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTorre LA, Trabert B, DeSantis CE, Miller KD, Samimi G, Runowicz CD, et al. Ovarian cancer statistics, 2018. CA Cancer J Clin. 2018;68(4):284\u0026ndash;96.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang M, Cheng S, Jin Y, Zhao Y, Wang Y. Roles of CA125 in diagnosis, prediction, and oncogenesis of ovarian cancer. Biochim Biophys Acta Rev Cancer. 2021;1875(2):188503.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePantel K, Alix-Panabi\u0026egrave;res C. Liquid biopsy and minimal residual disease - latest advances and implications for cure. Nat Rev Clin Oncol. 2019;16(7):409\u0026ndash;24.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZviran A, Schulman RC, Shah M, Hill STK, Deochand S, Khamnei CC, et al. Genome-wide cell-free DNA mutational integration enables ultra-sensitive cancer monitoring. Nat Med. 2020;26(7):1114\u0026ndash;24.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLo YMD, Han DSC, Jiang P, Chiu RWK. Epigenetics, fragmentomics, and topology of cell-free DNA in liquid biopsies. Science. 2021;372(6538):eaaw3616.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim SH, Weigelt B. Circulating cell-free (cf)DNA analysis: Current technologies and applications in gynecologic cancer. Gynecol Oncol Rep. 2024;54:101431.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDong Q, Chen C, Hu Y, Zhang W, Yang X, Qi Y, et al. Clinical application of molecular residual disease detection by circulation tumor DNA in solid cancers and a comparison of technologies: review article. Cancer Biol Ther. 2023;24(1):2274123.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChalmers ZR, Connelly CF, Fabrizio D, Gay L, Ali SM, Ennis R, et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med. 2017;9:34.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLuo H, Wei W, Ye Z, Zheng J, Xu RH. Liquid Biopsy of Methylation Biomarkers in Cell-Free DNA. Trends Mol Med. 2021;27(5):482\u0026ndash;500.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiang L, Zhang Y, Li C, Liao Y, Wang G, Xu J, et al. Plasma cfDNA methylation markers for the detection and prognosis of ovarian cancer. eBioMedicine. 2022;83:104222.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu MC, Oxnard GR, Klein EA, Swanton C, Seiden MV, Liu MC, et al. Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA. Ann Oncol. 2020;31(6):745\u0026ndash;59.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarrigou S, Perkins G, Garlan F, Normand C, Didelot A, Le Corre D, et al. A Study of Hypermethylated Circulating Tumor DNA as a Universal Colorectal Cancer Biomarker. Clin Chem. 2016;62(8):1129\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeinse G, Borghese B, M\u0026eacute;tairie M, Just PA, Poulet G, Garinet S, et al. Highly Specific Droplet-Digital PCR Detection of Universally Methylated Circulating Tumor DNA in Endometrial Carcinoma. Clin Chem. 2022;68(6):782\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVaisvila R, Ponnaluri VKC, Sun Z, Langhorst BW, Saleh L, Guan S, et al. Enzymatic methyl sequencing detects DNA methylation at single-base resolution from picograms of DNA. Genome Res. 2021;31(7):1280\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeng H, Conneely KN, Wu H. A Bayesian hierarchical model to detect differentially methylated loci from single nucleotide resolution sequencing data. Nucleic Acids Res. 2014;42(8):e69.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAkalin A, Kormaksson M, Li S, Garrett-Bakelman FE, Figueroa ME, Melnick A, et al. methylKit: a comprehensive R package for the analysis of genome-wide DNA methylation profiles. Genome Biol. 2012;13(10):R87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKorthauer K, Chakraborty S, Benjamini Y, Irizarry RA. Detection and accurate false discovery rate control of differentially methylated regions from whole genome bisulfite sequencing. Biostatistics. 2019;20(3):367\u0026ndash;83.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi W, Li Q, Kang S, Same M, Zhou Y, Sun C, et al. CancerDetector: ultrasensitive and non-invasive cancer detection at the resolution of individual reads using cell-free DNA methylation sequencing data. Nucleic Acids Res. 2018;46(15):e89.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoyfer N, Rosenski J, Kaplan T. wgbstools: A computational suite for DNA methylation sequencing data representation, visualization, and analysis. 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTarasov A, Vilella AJ, Cuppen E, Nijman IJ, Prins P. Sambamba: fast processing of NGS alignment formats. Bioinformatics. 2015;31(12):2032\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoyfer N, Magenheim J, Peretz A, Cann G, Bredno J, Klochendler A, et al. A DNA methylation atlas of normal human cell types. Nature. 2023;613(7943):355\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAffinito O, Palumbo D, Fierro A, Cuomo M, De Riso G, Monticelli A, et al. Nucleotide distance influences co-methylation between nearby CpG sites. Genomics. 2020;112(1):144\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerger AC, Korkut A, Kanchi RS, Hegde AM, Lenoir W, Liu W, et al. A Comprehensive Pan-Cancer Molecular Study of Gynecologic and Breast Cancers. Cancer Cell. 2018;33(4):690\u0026ndash;e7059.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Bruijn I, Kundra R, Mastrogiacomo B, Tran TN, Sikina L, Mazor T, et al. Analysis and Visualization of Longitudinal Genomic and Clinical Data from the AACR Project GENIE Biopharma Collaborative in cBioPortal. Cancer Res. 2023;83(23):3861\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElkin R, Oh JH, Liu YL, Selenica P, Weigelt B, Reis-Filho JS, et al. Geometric network analysis provides prognostic information in patients with high grade serous carcinoma of the ovary treated with immune checkpoint inhibitors. NPJ Genom Med. 2021;6(1):99.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature. 2011;474(7353):609\u0026ndash;15.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao J, Aksoy BA, Dogrusoz U, Dresdner G, Gross B, Sumer SO, et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Sci Signal. 2013;6(269):pl1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data. Cancer Discov. 2012;2(5):401\u0026ndash;4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu N, Liu J, Xu M, Liang J, Wang Y, Wu Z, et al. CSMD3 is Associated with Tumor Mutation Burden and Immune Infiltration in Ovarian Cancer Patients. Int J Gen Med. 2021;14:7647\u0026ndash;57.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa MC, Lavi ES, Altwerger G, Lin ZP, Ratner ES. Predictive modeling of gene mutations for the survival outcomes of epithelial ovarian cancer patients. PLoS ONE. 2024;19(7):e0305273.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFernandez SV, Tan YF, Rao S, Fittipaldi P, Sheriff F, Borghaei H, et al. Validation of a Molecular Diagnostic Test for Circulating Tumor DNA by Next-Gen Sequencing. Int J Mol Sci. 2023;24(21):15779.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKurtz DM, Soo J, Co Ting Keh L, Alig S, Chabon JJ, Sworder BJ, et al. Enhanced detection of minimal residual disease by targeted sequencing of phased variants in circulating tumor DNA. Nat Biotechnol. 2021;39(12):1537\u0026ndash;47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarinelli LM, Kisiel JB, Slettedahl SW, Mahoney DW, Lemens MA, Shridhar V, et al. Methylated DNA markers for plasma detection of ovarian cancer: Discovery, validation, and clinical feasibility. Gynecol Oncol. 2022;165(3):568\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKallio HM, Savolainen K, Virtanen T, Ryypp\u0026ouml; L, Selin H, Martikainen P, et al. Sensitive circulating tumor DNA-based residual disease detection in epithelial ovarian cancer. Life Sci Alliance. 2024;7(6):e202402658.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSenturk Kirmizitas T, van den Berg C, Boers R, Helmijr J, Makrodimitris S, Dag HH, et al. Epigenetic and Genomic Hallmarks of PARP-Inhibitor Resistance in Ovarian Cancer Patients. Genes (Basel). 2024;15(6):750.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHou JY, Chapman JS, Kalashnikova E, Pierson W, Smith-McCune K, Pineda G, et al. Circulating tumor DNA monitoring for early recurrence detection in epithelial ovarian cancer. Gynecol Oncol. 2022;167(2):334\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeo J, Kim YN, Shin S, Lee K, Lee JH, Lee YJ, et al. Serial Circulating Tumor DNA Analysis with a Tumor-Na\u0026iuml;ve Next-Generation Sequencing Panel Detects Minimal Residual Disease and Predicts Outcome in Ovarian Cancer. Cancer Res. 2024;84(3):468\u0026ndash;78.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-experimental-and-clinical-cancer-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jecc","sideBox":"Learn more about [Journal of Experimental \u0026 Clinical Cancer Research](http://jeccr.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/jecc/default.aspx","title":"Journal of Experimental \u0026 Clinical Cancer Research","twitterHandle":"@OncoBioMed","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6031886/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6031886/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eEpithelial ovarian cancer (EOC) is a leading cause of cancer mortality in women, often diagnosed at advanced stages. While first-line treatments improve survival, relapses remain common, with 5-year survival rates below 40%. Circulating tumor DNA (ctDNA) is a promising biomarker for non-invasive EOC detection and monitoring. It may help assess treatment response, notably microscopic residual diseases. Our objective was to compare two ctDNA characterization strategies in EOC for assessing tumor burden during first-line treatment: a tumor-informed approach based on somatic mutations and a tumor-type informed approach utilizing DNA methylation patterns.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eIn the tumor-informed approach, whole exome sequencing (WES) was performed on EOC tumor DNA and matched PBMCs from 12 patients aiming to identify tumor-specific mutations. Custom panels were designed to target patient\u0026rsquo;s specific tumor mutations, which were then tracked in cfDNA from plasma samples. In the tumor-type informed approach, differentially methylated loci (DMLs) were identified by comparing EOC samples, healthy ovarian tissues, and PBMCs. A unique custom methylation panel was designed, and a support vector machine classifier was trained to distinguish between healthy and cancerous plasma samples. Plasma samples were collected from 47 advanced-stage EOC patients during neoadjuvant chemotherapy, alongside plasma from healthy subjects.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eFor the tumor-informed approach, WES identified an average of 74 somatic mutations per patient. CtDNA was detected in 11 out of 12 patients at baseline (mean VAF: 1.29%). For the tumor-type informed approach, 52,173 DMLs were identified. The classifier trained on these DMLs detected ctDNA in baseline plasma samples for 11 out of the 12 patients demonstrating equivalent sensitivity (mean VAF: 1.17%. In end-of-treatment samples, the tumor-type-informed approach detected ctDNA twice as often as the tumor-informed method. Detection using this more sensitive approach correlated with relapse and shorter progression-free survival (log-rank p\u0026thinsp;=\u0026thinsp;0.017, Hazard ratio\u0026thinsp;=\u0026thinsp;8.24; 95% CI [1.06\u0026ndash;64.4]) and was associated with poorer overall survival (log-rank p\u0026thinsp;=\u0026thinsp;0.036).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe tumor-type informed classifier demonstrated sensitivity and specificity for ctDNA detection, outperforming the tumor-informed approach in monitoring EOC progression. Requiring fewer sequencing data, it offers a practical, efficient solution for clinical management of EOC.\u003c/p\u003e","manuscriptTitle":"Improved Tumor-Type informed compared to Tumor-Informed Mutation Tracking for ctDNA Detection and Microscopic Residual Disease Assessment in Epithelial Ovarian Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-11 09:34:06","doi":"10.21203/rs.3.rs-6031886/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-03-03T07:04:58+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-03-03T00:54:13+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-02-28T09:28:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"113892353215474697852892709480916393201","date":"2025-02-28T09:18:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"127424126030112297319817042818345793326","date":"2025-02-18T00:57:33+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-02-17T14:50:04+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-02-17T14:41:21+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-02-17T12:20:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Experimental \u0026 Clinical Cancer Research","date":"2025-02-14T15:34:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-experimental-and-clinical-cancer-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jecc","sideBox":"Learn more about [Journal of Experimental \u0026 Clinical Cancer Research](http://jeccr.biomedcentral.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/jecc/default.aspx","title":"Journal of Experimental \u0026 Clinical Cancer Research","twitterHandle":"@OncoBioMed","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"817035a7-0c83-432f-ba90-b40b3684859a","owner":[],"postedDate":"March 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-06-16T16:01:01+00:00","versionOfRecord":{"articleIdentity":"rs-6031886","link":"https://doi.org/10.1186/s13046-025-03433-4","journal":{"identity":"journal-of-experimental-and-clinical-cancer-research","isVorOnly":false,"title":"Journal of Experimental \u0026 Clinical Cancer Research"},"publishedOn":"2025-06-12 15:57:39","publishedOnDateReadable":"June 12th, 2025"},"versionCreatedAt":"2025-03-11 09:34:06","video":"","vorDoi":"10.1186/s13046-025-03433-4","vorDoiUrl":"https://doi.org/10.1186/s13046-025-03433-4","workflowStages":[]},"version":"v1","identity":"rs-6031886","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6031886","identity":"rs-6031886","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}