Background: Growing studies have suggested that personalized tumor-informed circulating tumor DNA (ctDNA) based minimal residual disease (MRD) detection can precisely predict recurrence in resectable patients (pts) with colorectal cancer (CRC). We developed an improved approach, a Tumor-Informed Patient-specific panel integrating with a tiny tumor-specific Core panel (TIPC) to monitor MRD by ultra-deep sequencing of plasma samples. The combination with a core panel confers TIPC overcoming the tumor temporal and spatial heterogeneity to some extent and providing more comprehensive information to guide individualized management. Here, we investigated the clinical utility of this assay and performed a head-to-head comparison with a Tumor-Informed Off-the-shelf panel approach (TIO). Methods: The retrospective study enrolled 34 I-IV CRC patients undergoing curative intent surgery from multiple centers and a median follow-up of 482 days. Tissue samples were collected at surgery. Peripheral blood samples collected postoperative and before relapse were analyzed. For all patients, at least one sample was available within 6 months before progression, defined as surveillance time point. TIPC is a whole exome sequencing (WES) based approach to identify somatic mutations from tissue samples, followed by customizing a patient-specific capture panel covering up to 50 top-ranked variants, and then combined with a CRC universal core panel containing 10 actionable or high prevalence cancer-related genes to conduct ultra-deep sequencing above 100,000X of follow-up blood samples. TIO assay was performed in serial plasma samples by using an off-the-shelf 338 gene panel at a sequencing depth of 30,000X. Results: As of August 31, 2022, 25 pts had confirmed recurrence. For the surveillance time point, 24/25 (96.0%) disease recurrence pts were detected MRD positive with TIPC assay, meanwhile, all disease-free pts were MRD negative (100%, 9/9). MRD detection preceded radiographic progression (median 5.6 months). TIO approach showed the same MRD detection specificity performance as TIPC, whereas the sensitivity was 88.0% (22/25). The ctDNA fraction of these two TIO undetectable pts was quantified using TIPC were 0.0168% and 0.0177%, while the positive mutations were beyond the 338 gene panel. The detectable postoperative ctDNA was associated with a poor prognosis for patients by either TIPC or TIO assay, with hazard ratios 20.8 (p＜0.0001) and 7.8 (p＜0.0001) respectively. To further verify TIPC may possess higher sensitivity, another cohort of 4 radiographic progressed CRC pts with TIO longitudinal undetectable MRD was analyzed, 3/4 (75.0%) was detectable by TIPC (ctDNA fraction range 0.0006%-0.0075%). Conclusions: WES-based patient-specific MRD detection assay demonstrated encouraging recurrence risk stratification value in CRC pts and may pave the way for the management of individual treatment.