This project collected 197 patients from Beijing Electric Power Hospital, Shanghai Electric Power Hospital, and Shandong Electric Power Hospital from the start of the project until six months before the project’s conclusion (30 August 2022 to 31 July 2024). The sample size was determined by G-power analysis. All experiments were approved by medical ethics committee of Capital Medical University Electric Power Teaching Hospital (2022083010101). All enrolled participants must undergo a colonoscopy within one week after completing two qFIT tests. The inclusion criteria were (1) asymptomatic people aged 45–75 years who were enrolled in the program for physical examination, (2) FIT positive with a cut-off value of 20 μg hemoglobin per gram stool (g), (3) participants who were willing to undergo colonoscopy. The exclusion criteria were (1) use of antibiotics within the past 3 months, (2) on a vegetarian diet, (3) had an invasive medical intervention within the past 3 months and (4) had a history of other types of cancer. Detailed records of all enrolled patients’ basic clinical information will be maintained, including age, gender, smoking history, pathological examination results, complete blood count data, and other relevant pathological findings ( Table 1 ). The workflow was shown as follows ( Figure 1 ).

After obtaining the fecal samples, a portion of the frozen samples is homogenized and centrifuged at 14,000 g for 5 minutes at room temperature using the NucleoSpin Soil Kit (Machery-Nagel GmbH & Co., Düren, Germany). The supernatant is used as the source for mt-sDNA. Genomic DNA was isolated and extracted from the fecal samples using the QIAamp DNA Stool Mini Kit according to the manufacturer’s instructions (Qiagen, Hilden, Germany). DNA concentration was measured with a NANO DROP 2000, and samples with concentrations below 10 ng/μl or OD260/OD280 ratios outside the range of 1.8–2.2 are re-extracted. All extracts were stored at −20°C until PCR amplification.

FIT testing is performed using the Fujirebio quantitative fecal occult blood analyzer according to the manufacturer’s instructions. DNA from each fecal sample is sent in batches to separate laboratories for blinded analysis. The operators conducting the FIT tests have experience with at least 1,000 samples and remain blinded to other study-related results. Fecal samples with hemoglobin levels exceeding 100 ng/ml of buffer (or 20 μg hemoglobin per g of feces) are classified as positive. 100 ng/ml of buffer (or 20 μg hemoglobin per g of feces) were classified as positive. All enrolled samples undergo hemoglobin threshold classification, with thresholds set at 10 ng/ml, 50 ng/ml, 100 ng/ml, 200 ng/ml, and 400 ng/ml. 100 ng/ml was set as the final cutoff to determine a positive criterion.

The colonoscopy was performed by an experienced endoscopist, who examined the entire colon up to the cecum with an adequate withdrawal time. Both the pathologist and endoscopist were blinded to the qFIT results and study objectives. Adenomas were evaluated based on tumor number, size, and location: the right colon included the cecum, ascending colon, and proximal two-thirds of the transverse colon, while the left colon comprised the distal third of the transverse colon, descending colon, sigmoid colon, and rectum. Colonoscopy results with colitis, hyperplastic polyps, non-bleeding hemorrhoids, or diverticulosis were considered negative. Based on the colonoscopy findings, samples were classified as non-adenomatous polyps, other lesions, non-advanced adenomas, advanced adenomas, colorectal cancer, or advanced colorectal cancer.

Specifically, a KRAS mutation detection kit (PCR-capillary electrophoresis) was used to detect KRAS mutation using stool-based DNA, targeting common mutations such as Gly12Asp, Gly12Val, Gly12Ser, Gly12Cys, Gly12Ala, Gly12Arg, and Gly13Asp. The target gene was amplified via fluorescent quantitative PCR, and KRAS mutation status was determined using a capillary electrophoresis device (Yuewei). The PCR conditions were as follows: pre-denaturation at 95°C for 3 minutes (1 cycle), denaturation at 94°C for 15 seconds, followed by annealing and extension at 60°C for 45 seconds (45 cycles). The results were analyzed by the study’s biostatistician.

Fecal samples in patients were collected and stored in -20°C followed by DNA extraction. Then, specific primers targeting M3 were designed and used for qPCR amplification. The primer sequences were as followings: Forward, 5’-AATGGGAATGGAGCGGATTC-3’; Reverse, 5’-CCTGCACCAGCTTATCGTCAA-3’. After normalization with the ACTB gene, M3 expression pattern was trisected equally into three parts based on the ΔCT value (high expression: <15.22, moderate expression: 15.22 - 20.03, and low expression: >20.03). Relative mRNA levels of the target genes were calculated using the 2^−ΔΔCT method.

Stool-based DNA was used to detected methylation burden of SDC2 gene using methylation-specific PCR. DNA extraction and bisulfite conversion were performed according to the manufacturer’s protocol. PCR was then used to detect methylation-specific fragments, with the internal control gene ACTB utilized to assess the adequacy of DNA quantity in the samples. Methylation-specific real-time PCR amplification was conducted on an ABI 7500 real-time PCR system (Thermo Fisher Scientific, MA, USA). The PCR protocol consisted of an initial denaturation at 94°C for 20 minutes, followed by 45 cycles at 62°C for 5 seconds, 55.5°C for 35 seconds, 93°C for 30 seconds, and a final extension at 40°C for 5 seconds. Positive and negative controls were included in each reaction to quantify the methylation burden. Methylation results were interpreted strictly in accordance with the manufacturer’s guidelines.

A receiver operating characteristic (ROC) curve analysis was conducted, and the area under the curve (AUC) was calculated to evaluate diagnostic performance. A 95% confidence interval for the AUC was estimated using a nonparametric approach. All statistical analyses were carried out using IBM SPSS Statistics version 23. P<0.05 was defined as the statistically significant differences.

This study enrolled a total of 197 patients, consisting of 121 tumor cases and 76 normal cases. Four biomarkers were evaluated and analyzed: FIT, KRAS mutation, m3 and Methylation, the most recognized standard in clinical CRC diagnosis. In the results, there were 40 FIT positive patients, accounting for 20.3%. The diagnostic performance was demonstrated for FIT with AUC value 0.59 in all patients, 0.579 in patients with early CRC (Stage I, no significant) and 0.621 in patients with advanced CRC (Stage II/III) ( Figures 2A–C ). The incidence of KRAS mutation in CRC patients was also detected and 63 cases accounting for 32% were positive, including G12C⁺, G12V⁺ and G13D⁺ as the most common. AUC value with ROC curves for KRAS was 0.643 in all patients, 0.649 in patients with early CRC (Stage I) and 0.625 in patients with advanced CRC (Stage II/III) ( Figures 2D–F ). Moreover, m3 bacterial genes were also evaluated in all these patients. The results showed that high levels of m3 were observed in 64 cases with △ct values of amplified m3 gene lower than 15.22, 54 cases interposed between 15.22 and 20.03 and △ct values of m3 gene in 79 cases were above 20.3, indicating the lower expression of m3 gene. The diagnosis performance for m3 was 0.824 in all patients, 0.823 in patients with early CRC (Stage I) and 0.825 in patients with advanced CRC (Stage II/III) ( Figures 2G–I ). Furthermore, 83 cases were detected with DNA methylation, and the AUC value for methylation was 0.778 in all patients, 0.779 in patients with early CRC (Stage I) and 0.776 in patients with advanced CRC (Stage II/III) ( Figures 2J–L ). From evidence above, it was found that m3 showed the highest diagnosed performance among these four indicators while FIT exhibited the worst performance with AUC value lower than 0.6.

Next, the diagnosis performance of random combination of two variables were evaluated. As shown in Figure 3 , it could be found that two indicators combination mostly showed better predictive performance than a single biomarker. The combination of FIT and KRAS demonstrated the lowest AUC value 0.699 in all patients, which was 0.696 in patients with early CRC and 0.719 in patients with advanced CRC ( Figures 3A–C ). Thus, the combination of FIT and KRAS were more suitable for patients with advanced CRC. Notably, the combination of FIT and m3 made no enhancement on the diagnosis performance with AUC value 0.824 equal to single m3 diagnosis ( Figure 3D versus Figure 2G ), and the diagnosis performance was 0.823 in patients with early CRC and 0.825 in patients with advanced CRC ( Figures 3E, F ). Once combined KRAS with m3, the AUC values elevated from 0.824 to 0.866 in all patients, which was 0.878 in patients with early CRC and 0.825 in patients with advanced CRC ( Figures 3G–I ), indicating that the combination of m3 and KRAS were more suitable for patients with early CRC. When combined methylation with FIT, KRAS and m3, respectively, the AUC values were 0.803, 0.824 and 0.903 in all patients ( Figures 3J, M, P ), respectively; For stage I CRC, the values were 0.779, 0.829 and 0.909 ( Figures 3K, N, Q ), which were 0.814, 0.811 and 0.886 in advanced CRC once combined methylation with FIT, KRAS and m3 ( Figures 3L, O, R ). Taken together, the combination of Methylation and m3 showed the highest AUC value for early CRC patients and advanced CRC patients, combinational marker for diagnosis showed better performance than single marker.

Then, ROC analysis was performed with random integration of three variables. As observed, AUC value of the model of FIT, KRAS and M3 integration was 0.864 in all patients, exhibiting no improvement over the AUC value compared to KRAS and M3 combination; AUC value of the model of FIT, KRAS and M3 integration was 0.878 in early CRC and 0.825 in advanced CRC ( Figures 4A–C ). Besides, it was shown that FIT, KRAS and methylation combination showed slight enhancement on the diagnosis performance with AUC value 0.841 compared to the combination of any two variables between them, which was 0.829 in early CRC and 0.842 in advanced CRC ( Figures 4D–F ). Interestingly, the combination of FIT, M3 and Methylation combination showed a higher AUC value (0.901) than the groups of FIT plus M3 or FIT plus methylation, while lower than the combination of M3 plus methylation ( Figures 3 , 4G ); the combination of FIT, M3 and Methylation combination showed the AUC value 0.909 in early CRC and 0.886 in advanced CRC ( Figures 4H, I ). Noteworthy, KRAS, M3 and Methylation combination possessed the most predictive ability with AUC value elevated to 0.920 (Confidence Interval: 0.881-0.960), which were better than any pairwise combinations ( Figure 4J ); and this combination harbored the highest AUC value 0.931 in early CRC and 0.886 in advanced CRC ( Figures 4K, L ). It could be found from the results that the model of KRAS, M3 and methylation integration showed the most potential predictive ability for early CRC diagnosis.

Subsequently, the four indicators were combined to evaluate AUC value from the ROC curve, and the result showed AUC value was 0.920 (Confidence Interval: 0.880-0.960), making no evident alteration comparing to the model established with KRAS, M3 and Methylation combination, indicating that FIT might make no significant contribution to diagnosis performance when KRAS, M3 and Methylation were used to predict CRC ( Figure 5A ); The combination of four indicators showed the highest AUC value 0.931 in early CRC and 0.886 in advanced CRC ( Figures 5B, C ). Taking into account that a smaller AUC confidence interval indicated more credible of the AUC value, the combination of KRAS, M3 and Methylation presented a better predictive efficiency than the combination of four factors. In addition to that, the combination of three indicators was suitable for CRC diagnosis in view of the clinical timeliness.