3615 Background: Molecular residual disease (MRD) using circulating tumor DNA (ctDNA) may identify recurrent risk and guide adjuvant chemotherapy (ACT) in colorectal cancer (CRC). This study explored the feasibility of a tumor-informed personalized approach to predict recurrence, comparing with tumor-informed (TI) and tumor naïve (TN) fixed panel approaches. Methods: Patients with stage I-IV CRC eligible for R0 resection were recruited. Blood samples were collected preoperatively, at postoperative day 7 and 30 (POD7/30), and every 3-6 months thereafter, which were tested by brPROPHET assay, using up to 50 highly ranked variants identified by whole-exome sequencing of surgical tissue samples. A fixed 168-gene panel was applied for TI and TN fixed panel tests. Results: A total of 214 patients were enrolled. With a median follow-up of 13.2 months, 20 (9.3%) participants relapsed. The brPROPHET assay had a positivity rate of 96% in preoperative samples. ctDNA positivity indicated by brPROPHET predicted worse recurrence-free survival (RFS) at both POD7 (hazard ratio [HR]=7.35, 95% confidence interval [95% CI], 3.04-17.78, P<0.001) and POD30 (HR=5.18, 95% CI, 2.00-13.38, P<0.001). Higher mean tumor molecules (MTM) level >0.0316/mL was observed in surveillance of 15 relapsed patients, which was associated with worse RFS (HR=44.47, 95% CI, 12.83-154.12, P<0.001). In ctDNA positive patients at POD7, patients benefit from ACT (HR=0.23, 95% CI, 0.05-1.00, P=0.03), which was not observed in ctDNA negative patients (HR=1.80, 95% CI, 0.48-6.81, P=0.38). Among the patients with results of all three approaches (n=168), the brPROPHET assay outperformed TI and TN fixed-panel assays in predicting RFS at POD7 (HR, 7.52 vs 3.43 and 5.60). Of note, the brPROPHET assay identified 6 additional relapsed patients beyond the fixed panels, with more than 87.5% of the positive sites derived from personalized designs. Conclusions: The individualized tumor-informed brPROPHET assay effectively identified postoperative patients with high risk of recurrence and benefit from ACT. Using a personalized tumor-informed panel that covers sites outside fixed panels may improve the utility of MRD detection. [Table: see text]