Background: Despite recent advances in cancer diagnostics and treatments, nearly 75% ovarian cancer patients relapse within two years of primary therapy and become candidates for treatment of recurrent diseases. A novel approach to post-surgical risk stratification for molecular residual disease (MRD) detection and treatment monitoring is needed to improve patient outcomes. Methods: We analyzed a cohort of 22 stage Ⅰ-Ⅳ ovarian cancer patients which were followed up at median 6 (range,3-15) times between July 2017 and December 2020. Among the 22 patients, stage I, II, III and Ⅳ comprised 9%, 5%, 50% and 36%, respectively. Tumor tissue and blood samples of all subjects were collected at Ruijin Hospital, Shanghai Jiao Tong University.Somatic mutations were identified by whole exome sequencing and assessed from plasma collected during treatment using personalized tumor-informed MRDtectassays with a median of 25（range,17-30） amplicons/assay targeting tumor-specific variants unique to each patient, sequenced at mean 200,000X depth. Plasma samples with ≥2 variants detected were considered MRD-positive and ctDNA levels were reported in mean tumor molecules per mL (MTM/mL) of plasma. Sequencing data were analyzed using PiVAT secondary analysis software and MRD positive was concluded when at least two variants were detected in plasma. Results: CtDNA was preoperatively detectable in 18 (85.7%) patients, and the positive rate became 38.8% (7/18) after surgery. Serial ctDNA analysis during surveillance after definitive treatment of the 22 patients with longitudinal collected plasma samples identified relapse with 100% sensitivity and 100% specificity. The median lead time from ctDNA detection in plasma to relapse detection by standard-of-care computed tomography was 135 days (range,45-387days). Variants from a total of 526 genes were selected from the WES results of these 22 patients. Among these 526 genes, 96% (505 genes) were selected for only one patient, 3.4% (18 genes) were selected for two patients and 0.2% (one gene) were selected for three patients. Compared with tumor-agnostic assay, tumor-informed assay is a more accurate route that is more consistent with the heterogeneity of tumor and can maximize the ability to reflect the real situation of each patient. Conclusions: Sensitive methods for risk stratification, therapeutic efficacy monitoring, and early relapse detection may have a major impact on treatment decisions and patient management for ovarian cancer patients. MRD can be used as an effective biomarker to prompt prognosis and recurrence risk, providing powerful assistance for accurate diagnosis and treatment for patients.