Colonoscope-derived mucus as a novel high-fidelity reservoir for KRAS-mutated precancerous colorectal neoplasia detection Authors: Fang-Chin Hsu, Ta-Wei Pu, Jung-Chun Lin, Chao-Yang Chen Published: 2026 ABSTRACT Highlights•Colonoscope-derived mucus may serve as a localized cfDNA source for colorectal neoplasia detection.•Mucus cfDNA KRAS testing showed promising sensitivity for colorectal neoplasia in this pilot cohort.•Low-threshold KRAS detection identified adenocarcinoma and advanced precancerous lesions with high sensitivity, but reduced specificity.•Mucus-based KRAS analysis showed higher sensitivity than FIT and serum CEA in this exploratory comparison.•This approach may provide a colonoscopy-integrated molecular adjunct for post-procedure risk stratification and surveillance optimization. Introduction Colorectal cancer (CRC) remains a significant global health burden, ranking as the fourth most common malignancy and the third leading cause of cancer-related mortality worldwide. According to GLOBOCAN 2022 estimates, approximately 1.93 million new CRC cases and 930,000 deaths occurred in 2022. Projections indicate a 71% increase in new cases by 2040, driven by aging populations and evolving lifestyle factors, particularly in low- and middle-income countries [1]. In Asia, incidence rates are rising in high-income regions such as Japan and South Korea, while in Western nations like the United States, over 150,000 new cases are anticipated annually by 2025 [1,2]. Taiwan exemplifies this trend; CRC is established as the second most common cancer, with age-standardized incidence rates of 43.12 per 100,000 for men and 29.31 per 100,000 for women (2020 data). Projections for Taiwan estimate a 71% rise in incidence and an 85% increase in mortality by 2040 [3,4]. Notably, Taiwan reports one of the highest incidence rates globally among individuals aged 40–49 years, underscoring the critical need for earlier and more effective screening strategies [3]. Current guidelines from the US Preventive Services Task Force and American Cancer Society recommend initiating screening at age 45 years for average-risk individuals, typically utilizing annual fecal immunochemical testing (FIT) [5]. In Taiwan, aiming to reduce mortality through early detection, the national screening program expanded in 2025 to provide free biennial FIT for individuals aged 45–74 years and high-risk groups aged ≥ 40 years [6]. However, despite these advancements, FIT and traditional guaiac-based fecal occult blood testing (gFOBT) have notable limitations. While FIT demonstrates superior sensitivity (70%–80% for CRC) and specificity compared with gFOBT, it is associated with high false-positive rates, leading to unnecessary colonoscopies and increased health-care burdens [7,8]. Furthermore, the sensitivity of FIT for advanced adenomas is < 50%, potentially missing precancerous lesions and contributing to interval cancers [7,9]. Even colonoscopy, the gold standard, is not infallible; the 5-year cumulative incidence of interval CRC after a negative examination ranges from 0.25% to 0.8% [10,11]. This residual risk—attributed to missed lesions, rapid tumor progression, or incomplete bowel preparation—highlights the urgent need for adjunctive biomarkers to refine risk stratification and optimize surveillance intervals [10,12]. KRAS mutations, particularly at codon 12 (G12), represent a promising target. Occurring in approximately 40% of CRC cases, KRAS G12 variants drive tumorigenesis and confer resistance to anti-EGFR therapies [[13], [14], [15]]. While standard in metastatic settings for treatment prediction, detecting low-level KRAS G12 mutations offers distinct advantages for screening and surveillance pathways because it enables the identification of high-risk precancerous lesions (present in > 50% of advanced adenomas) and facilitates personalized surveillance [16,17]. Although cell-free DNA (cfDNA) analysis via liquid biopsy has revolutionized detection, source selection is critical. Tissue biopsy is invasive, while plasma-derived cfDNA often lacks sensitivity for early-stage disease as a result of dilution [[18], [19], [20]]. By contrast, colonoscopy-derived mucus emerges as a novel, localized cfDNA source. It captures tumor-derived DNA directly from the intestinal microenvironment, offering higher concentrations (up to tenfold enrichment over plasma) and improved sensitivity for precancerous lesions (70%–90%) [21,22]. Given this background, the present prospective pilot study aimed to evaluate the feasibility of colonoscope-derived mucus cfDNA for KRAS G12 mutation detection during routine colonoscopy in a Taiwanese cohort and to compare its performance against FIT and carcinoembryonic antigen (CEA) as a potential colonoscopy-integrated molecular adjunct. To our knowledge, this is one of the first studies to investigate mucus collected from the external surface of the colonoscope sheath as a localized, lesion-proximal cfDNA reservoir for detecting KRAS-mutated colorectal neoplasia. In addition, by assessing both >1% and >0.1% variant allele frequency (VAF) thresholds, this study explored whether low-level KRAS-mutated cfDNA in mucus could expand the detectable molecular range for early and precancerous colorectal lesions. Materials and methods Study design and ethics This prospective, single-center pilot study was conducted at Tri-Service General Hospital, Taiwan, from March 2024 to March 2025. The protocol was approved by our Institutional Review Board (approval No.: C202404001) and adhered to the Declaration of Helsinki. Written informed consent was obtained from all participants. Patient enrollment and sample collection Eligible participants were adults aged 30–74 years undergoing routine outpatient colonoscopy. Exclusion criteria included inadequate bowel preparation (Boston Bowel Preparation Scale [BBPS] < 6) or incomplete cecal intubation. A total of 43 participants were enrolled (Fig. 1). Following standard bowel preparation and a complete colonic examination, mucus samples were collected noninvasively from the external surface of the colonoscope sheath using a sterile cytology brush. Sampling was performed after completion of the examination and withdrawal of the colonoscope. Mucus was collected from the outer surface of the scope (rather than the working channel) to minimize potential contamination from instrument passage or biopsy-related bleeding. The collected mucus was eluted in 5 mL of phosphate-buffered saline and stored at –80 °C until further analysis.Fig. 1Participant enrollment and pilot study design.Fig 1 dummy alt text Participant enrollment and pilot study design. DNA extraction and mutation analysis cfDNA was extracted using the QIAamp Circulating Nucleic Acid Kit (Qiagen) adapted for viscous fluids. KRAS codon 12 mutations (G12D, G12V, G12C, G12A) were detected via a competitive PCR assay with wild-type blocking. Primers targeting KRAS codon 12 wild-type and mutant alleles were designed as follows: forward 5’-TCAAGGCACTCTTGCCTACG-3’; reverse 5’-AGGTACTGGTGGAGTATTTGATAGT-3’; and blocker 5’-CTCTTGCCTACGCCACCAGC-3’ with phosphorylation at 3’. Blocker exhibited a higher affinity to the KRAS codon 12 wild-type sequence, and the phosphate group at 3’ of blocker would therefore inhibit the amplification of the wild-type sequence. The amplification of KRAS codon 12 sequences was conducted using the VeritiPro™ Thermal Cycler (Applied Biosystems, Waltham, MA, USA). The 20 μL reaction contained 10 μL 2 × Turbo Mix (150 mM Tris, pH 9.0; 8 mM MgCl2; 40 mM (NH4)2SO4; 400 nM dNTP; 300 mM betaine; 0.2% Tween-20), 200 nM forward/reverse primer, 350 nM blocker, 0.2 μL Taq polymerase (5 U/μL), 5 μL extracted cfDNA, and nuclease-free water. For the amplification of KRAS codon 12 plasmids with different variant allele frequencies (VAFs), 12,000 copies of plasmids were loaded in the reaction. After amplification (95 °C for 4 min; 35 cycles of 95 °C 30 s / 62 °C 30 s / 72 °C 45 s; 72 °C 7 min), the PCR products were analyzed via gel electrophoresis. VAFs were categorized as 0.1%–1%, 1%–5%, > 5%, or Not Detected. Thresholds for positivity were defined as amplification within 35 cycles and evaluated by comparing the intensity of desired product with that from different VAFs of standard KRAS G12D plasmids. Representative gel electrophoresis images were included to demonstrate assay performance, including negative controls, VAF-calibrated KRAS G12D plasmid standards, and representative clinical samples. Because this assay was based on end-point competitive PCR rather than real-time fluorescence PCR, amplification curves were not generated. Semi-quantitative VAF categories were assigned according to band intensity relative to calibrated plasmid standards. (Supplementary Fig. 1) Clinical and histological assessment Clinical data were collected prospectively. Lesions identified during colonoscopy were biopsied and classified as adenocarcinoma, advanced precancerous (high-grade dysplasia, villous components, or > 1 cm), non-advanced precancerous, hyperplastic/inflammatory, or negative. Histological findings served as the reference standard. Statistical analysis All data were analyzed using SPSS version 27.0. The sensitivity and specificity of mucus cfDNA KRAS G12 detection were compared against fecal occult blood testing (FOBT)/FIT and CEA using Fisher’s exact test. Missing data were handled conservatively via listwise deletion or imputation as negative for sensitivity analyses. Statistical significance was set at p < .05. Results Participant characteristics The participants’ demographic and clinical data are summarized in Table 1. The cohort (N = 43) had a mean age of 59.1 ± 11.4 years, with a male predominance (65.1%). A significant portion (32.6%) reported a history of neoplasia. Comorbidities included metabolic syndrome (type 2 diabetes, hypertension), cardiovascular disease, and other malignancies (e.g., lung, prostate, and breast cancer).Table 1Demographic characteristics of assessed participants.Table 1 dummy alt textCharacteristicsParticipants who completed colonoscopyN=43%Age, yMean (SD, Standard deviation)59.1 (11.4)37.5% (13.7–69.4%)Median (Range)62.0 (32-74)25.0% (7.1–59.1%)30-3937.0%40-49511.6%50-591227.9%60-691330.2%≥701023.3%GenderFemale1534.9%Male2865.1%Medical HistoryUnknown49.3%Neoplastic HistoryYes1432.6%Polyp49.3%Malignancy1023.3%No2558.1%Metabolic and Cardiovascular ComorbiditiesYes818.6%Diabetes Mellitus37.0%Cardiovascular risk factors/disease49.3%Both12.3%No3172.1%Chronic infectious/inflammatory conditionsYes37.0%No3683.7%Renal impairment or urological issuesYes12.3%No3888.4% Table 1Demographic characteristics of assessed participants.Table 1 dummy alt textCharacteristicsParticipants who completed colonoscopyN=43%Age, yMean (SD, Standard deviation)59.1 (11.4)37.5% (13.7–69.4%)Median (Range)62.0 (32-74)25.0% (7.1–59.1%)30-3937.0%40-49511.6%50-591227.9%60-691330.2%≥701023.3%GenderFemale1534.9%Male2865.1%Medical HistoryUnknown49.3%Neoplastic HistoryYes1432.6%Polyp49.3%Malignancy1023.3%No2558.1%Metabolic and Cardiovascular ComorbiditiesYes818.6%Diabetes Mellitus37.0%Cardiovascular risk factors/disease49.3%Both12.3%No3172.1%Chronic infectious/inflammatory conditionsYes37.0%No3683.7%Renal impairment or urological issuesYes12.3%No3888.4% Demographic characteristics of assessed participants. Colonoscopy and pathology findings Pathological evaluation of colonoscopy findings revealed a distribution consistent with routine screening cohorts. Negative results accounted for 46.5% of cases, characterized by no visible abnormalities and confirmation of normal mucosa on biopsy. Hyperplastic or inflammatory conditions, encompassing hyperplastic polyps and colitis-associated lesions, such as those linked to ulcerative colitis, were identified in 16.3%. Non-advanced precancerous lesions, primarily consisting of small tubular adenomas (≤ 1 cm) exhibiting low-grade dysplasia, comprised 14.0%. Advanced precancerous lesions, including adenomas larger than 1 cm, serrated adenomas, tubulovillous adenomas, or those with focal high-grade dysplasia, were present in 18.6%. Finally, adenocarcinoma was diagnosed in 4.7%, with two confirmed cases, one of which involved an ulcerative-type tumor in a patient with Crohn’s disease. Detection performance compared with conventional screening-associated tests Performance metrics against histological standards are detailed in Table 2. FOBT/FIT showed moderate sensitivity for adenocarcinoma (50.0%), but limited sensitivity for precancerous lesions (33.3%–37.5%). CEA demonstrated high sensitivity for advanced precancerous lesions (75.0%), but failed to detect adenocarcinoma in this cohort (0.0%) and exhibited poor specificity (37.0%). By contrast, mucus-derived cfDNA KRAS testing showed threshold-dependent performance. At a VAF > 1%, sensitivity was 50.0% for adenocarcinoma and advanced lesions, with improved specificity (63.0%) over CEA. Notably, lowering the threshold to a VAF > 0.1% achieved 100.0% sensitivity for adenocarcinoma and 75.0% for advanced lesions, significantly outperforming phenotypic tests, although with reduced specificity (29.6%).Table 2Comparison of detection capabilities for FOBT, CEA, and mucus cfDNA.Table 2 dummy alt textMethodologyAdenocarcinoma (N=2)Advanced Precancerous (N=8)Non-Advanced Precancerous (N=6)Negative Findings (N=27)% (95% CI)% (95% CI)% (95% CI)% (95% CI)FOBT / FITUnknown13212Positive50.0% (9.5–90.5%)37.5% (13.7–69.4%)50.0% (18.8–81.2%)37.0% (21.5–55.8%)Negative0.0% (0.0–65.8%)25.0% (7.1–59.1%)16.7% (3.0–56.4%)18.5% (8.2–36.7%)CEAUnknown06317Positive (≥5 ng/mL)0.0% (0.0–65.8%)12.5% (2.2–47.1%)16.7% (3.0–56.4%)0.0% (0.0–12.5%)Negative (<5 ng/mL)100.0% (34.2–100.0%)12.5% (2.2–47.1%)33.3% (9.7–70.0%)37.0% (21.5–55.8%)cfDNA KRAS G12 (VAF >1%)Positive50.0% (9.5–90.5%)50.0% (21.5–78.5%)16.7% (3.0–56.4%)37.0% (21.5–55.8%)Negative (Not detected)50.0% (9.5–90.5%)50.0% (21.5–78.5%)83.3% (43.6–97.0%)63.0% (44.2–78.5%)cfDNA KRAS G12 (VAF >0.1%)Positive100.0% (34.2–100.0%)75.0% (40.9–92.9%)50.0% (18.8–81.2%)70.4% (51.5–84.1%)Negative (Not detected)0.0% (0.0–65.8%)25.0% (7.1–59.1%)50.0% (18.8–81.2%)29.6% (15.9–48.5%)Notes: Percentages and 95% CIs calculated using the Wilson score method. Cases with missing or uninterpretable results (Unknown) are excluded from calculations. Negative Findings (N=27): no abnormality on colonoscopy (n=20) or hyperplastic/inflammatory polyps (n=7).Abbreviations: CI, confidence interval; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. Table 2Comparison of detection capabilities for FOBT, CEA, and mucus cfDNA.Table 2 dummy alt textMethodologyAdenocarcinoma (N=2)Advanced Precancerous (N=8)Non-Advanced Precancerous (N=6)Negative Findings (N=27)% (95% CI)% (95% CI)% (95% CI)% (95% CI)FOBT / FITUnknown13212Positive50.0% (9.5–90.5%)37.5% (13.7–69.4%)50.0% (18.8–81.2%)37.0% (21.5–55.8%)Negative0.0% (0.0–65.8%)25.0% (7.1–59.1%)16.7% (3.0–56.4%)18.5% (8.2–36.7%)CEAUnknown06317Positive (≥5 ng/mL)0.0% (0.0–65.8%)12.5% (2.2–47.1%)16.7% (3.0–56.4%)0.0% (0.0–12.5%)Negative (<5 ng/mL)100.0% (34.2–100.0%)12.5% (2.2–47.1%)33.3% (9.7–70.0%)37.0% (21.5–55.8%)cfDNA KRAS G12 (VAF >1%)Positive50.0% (9.5–90.5%)50.0% (21.5–78.5%)16.7% (3.0–56.4%)37.0% (21.5–55.8%)Negative (Not detected)50.0% (9.5–90.5%)50.0% (21.5–78.5%)83.3% (43.6–97.0%)63.0% (44.2–78.5%)cfDNA KRAS G12 (VAF >0.1%)Positive100.0% (34.2–100.0%)75.0% (40.9–92.9%)50.0% (18.8–81.2%)70.4% (51.5–84.1%)Negative (Not detected)0.0% (0.0–65.8%)25.0% (7.1–59.1%)50.0% (18.8–81.2%)29.6% (15.9–48.5%)Notes: Percentages and 95% CIs calculated using the Wilson score method. Cases with missing or uninterpretable results (Unknown) are excluded from calculations. Negative Findings (N=27): no abnormality on colonoscopy (n=20) or hyperplastic/inflammatory polyps (n=7).Abbreviations: CI, confidence interval; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. Comparison of detection capabilities for FOBT, CEA, and mucus cfDNA. Notes: Percentages and 95% CIs calculated using the Wilson score method. Cases with missing or uninterpretable results (Unknown) are excluded from calculations. Negative Findings (N=27): no abnormality on colonoscopy (n=20) or hyperplastic/inflammatory polyps (n=7). Abbreviations: CI, confidence interval; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. The apparent discrepancy between high sensitivity and low specificity at the VAF >0.1% threshold reflects a threshold-dependent performance trade-off. Lowering the positivity threshold increased the detection of low-abundance KRAS-mutated cfDNA signals, thereby reducing false-negative findings. However, this also increased the number of positive molecular signals among histologically negative cases, resulting in reduced specificity. Therefore, the VAF >0.1% threshold should be interpreted as an exploratory sensitivity-prioritizing cutoff rather than a clinically validated diagnostic threshold. Overall detection rates Conservative analysis (Table 3) revealed that for positive findings (neoplasia), cfDNA KRAS (VAF > 0.1%) achieved an overall sensitivity of 80.0% (95% confidence interval: 49.0–94.3), significantly outperforming CEA (10.0%, p = 0.005) and trending higher than FOBT/FIT (40.0%). While CEA offered the highest specificity for negative findings (97.0%), the superior sensitivity of low-threshold cfDNA came with a trade-off in specificity.Table 3Evaluated the overall detection rate for the positive & negative findings using a conservative missing-data estimation across different methodologies.Table 3 dummy alt textMethodologyNumberSensitivity (95% CI)OR (vs. cfDNA KRAS G12 VAF>0.1%)p-Value (Fisher Exact Test)Positive Findings (N=10)FOBT / FIT (+)440.0% (16.8–68.7%)0.170.17CEA (≥5 ng/mL)110.0% (1.8–40.4%)0.030.005cfDNA KRAS G12 (VAF>1%)550.0% (23.7–76.3%)0.250.350cfDNA KRAS G12 (VAF>0.1%)880.0% (49.0–94.3%)1Negative Findings (N=33)FOBT / FIT (-)2060.6% (43.7–75.3%)3.080.048CEA (<5 ng/mL)3297.0% (84.7–99.5%)64.00<0.001cfDNA KRAS G12 (VAF<1%)2266.7% (49.6–80.2%)40.013cfDNA KRAS G12 (VAF<0.1%)1133.3% (19.8–50.3%)1Notes: Missing data handled by conservative estimation. Positive Findings (N=10): adenocarcinoma (n=2) + advanced precancerous lesions (n=8). Negative Findings (N=33): non-advanced precancerous lesions (n=6) + Negative Findings from Table 2 (n=27). p-values by Fisher's exact test.Abbreviations: CI, confidence interval; OR, odds ratio; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. Table 3Evaluated the overall detection rate for the positive & negative findings using a conservative missing-data estimation across different methodologies.Table 3 dummy alt textMethodologyNumberSensitivity (95% CI)OR (vs. cfDNA KRAS G12 VAF>0.1%)p-Value (Fisher Exact Test)Positive Findings (N=10)FOBT / FIT (+)440.0% (16.8–68.7%)0.170.17CEA (≥5 ng/mL)110.0% (1.8–40.4%)0.030.005cfDNA KRAS G12 (VAF>1%)550.0% (23.7–76.3%)0.250.350cfDNA KRAS G12 (VAF>0.1%)880.0% (49.0–94.3%)1Negative Findings (N=33)FOBT / FIT (-)2060.6% (43.7–75.3%)3.080.048CEA (<5 ng/mL)3297.0% (84.7–99.5%)64.00<0.001cfDNA KRAS G12 (VAF<1%)2266.7% (49.6–80.2%)40.013cfDNA KRAS G12 (VAF<0.1%)1133.3% (19.8–50.3%)1Notes: Missing data handled by conservative estimation. Positive Findings (N=10): adenocarcinoma (n=2) + advanced precancerous lesions (n=8). Negative Findings (N=33): non-advanced precancerous lesions (n=6) + Negative Findings from Table 2 (n=27). p-values by Fisher's exact test.Abbreviations: CI, confidence interval; OR, odds ratio; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. Evaluated the overall detection rate for the positive & negative findings using a conservative missing-data estimation across different methodologies. Notes: Missing data handled by conservative estimation. Positive Findings (N=10): adenocarcinoma (n=2) + advanced precancerous lesions (n=8). Negative Findings (N=33): non-advanced precancerous lesions (n=6) + Negative Findings from Table 2 (n=27). p-values by Fisher's exact test. Abbreviations: CI, confidence interval; OR, odds ratio; FOBT, fecal occult blood test; FIT, fecal immunochemical test; CEA, carcinoembryonic antigen; cfDNA, cell-free DNA; VAF, variant allele frequency. Discussion The detection of KRAS codon 12 mutations in colonoscope-derived mucus cfDNA demonstrated promising sensitivity, particularly at a lower VAF threshold of > 0.1%, achieving 80% overall for positive neoplastic findings compared with 40% for FOBT/FIT and 10% for CEA. This enhanced performance across adenocarcinoma (100%) and advanced precancerous lesions (75%) underscores the potential of mucus cfDNA as an adjunct to colonoscopy, capturing localized tumor-derived DNA with minimal dilution from systemic sources [21]. However, the 100% sensitivity observed for adenocarcinoma should be interpreted cautiously because the adenocarcinoma subgroup included only two cases. The clinical urgency for improved precancerous detection is underscored by the recent pivotal validation of a blood-based circulating tumor DNA test (PREEMPT CRC study), published in 2025 [23]. While the blood-based assay demonstrated a commendable 79.2% sensitivity for established CRC and high specificity (91.5%), its performance for advanced precancerous lesions was strikingly low, at 12.5%, limiting its utility for cancer prevention [23]. By sharp contrast, our pilot data using mucus-derived cfDNA achieved 75% sensitivity for advanced precancerous lesions (at VAF > 0.1%). This profound discrepancy highlights the dilution effect inherent in systemic circulation, which masks the faint signals of early neoplasia. Thus, sampling directly from the localized mucin layer effectively bridged the diagnostic gap where systemic liquid biopsies currently fail. A potential question that arises is why molecular analysis of mucus is necessary when colonoscopy already allows direct visualization and biopsy of lesions. Importantly, mucus-based testing is not intended to replace tissue biopsy, which remains the diagnostic gold standard. Rather, it may serve as a complementary molecular layer of information. Colonoscopy is primarily a morphological tool, and histopathology is dependent on targeted sampling. Early molecular alterations, field cancerization, or subclinical clonal expansions may precede visible lesions or fail to be captured by limited biopsy specimens. In this context, mucus-derived cfDNA provides a broader, lesion-proximal sampling of the colonic surface, potentially identifying molecular risk signals that are not yet morphologically apparent. Such information may be particularly relevant for refining post-procedure surveillance strategies. Notably, KRAS variants were detected in a subset of cases with histology-negative or visually inconspicuous mucosa. Such findings may reflect true biological signals, including field cancerization or subclinical clonal expansions within the colonic epithelium that precede overt morphological transformation. However, reduced specificity at lower VAF thresholds also highlights potential biological and technical confounders. Intestinal inflammation—such as ulcerative colitis or Crohn’s disease—may increase cfDNA shedding from activated neutrophils and regenerating epithelium, generating low-level variants unrelated to neoplasia [[24], [25], [26]]. In addition, sampling heterogeneity and low-level background amplification may contribute to variant detection in the absence of histologically confirmed lesions. Accordingly, mutation calls at very low VAFs should be interpreted as hypothesis-generating molecular signals within this exploratory cohort rather than definitive evidence of neoplasia. These findings underscore the need for age- and inflammation-adjusted thresholds, as well as for future validation incorporating paired tissue genotyping and orthogonal molecular confirmation, to improve specificity without compromising early detection sensitivity. From a clinical decision-making perspective, this sensitivity–specificity trade-off is important. At the current exploratory VAF >0.1% threshold, mucus cfDNA KRAS testing should not be used as a stand-alone diagnostic assay or as a replacement for colonoscopy and histopathological assessment. Instead, low-threshold positivity should be interpreted as a molecular risk signal that requires integration with colonoscopic findings, histology, inflammatory status, and patient-level risk factors. If applied without threshold optimization, the low specificity could lead to unnecessary surveillance colonoscopy, increased healthcare burden, and patient anxiety. Future studies should therefore perform ROC-based threshold optimization and external validation to identify a clinically meaningful cutoff that balances early detection sensitivity with acceptable specificity. Collectively, these findings affirm the potential of mucus as a localized cfDNA reservoir within a colonoscopy-integrated screening pathway. As illustrated in Fig. 2, mucus cfDNA testing is not intended to replace FIT or serve as a stand-alone pre-colonoscopy screening test in the current study design. Rather, because sampling is performed after colonoscopy, its most immediate role may be to provide post-procedure molecular risk stratification. In future validated algorithms, mutation-positive patients with otherwise low-risk colonoscopic or histological findings could be considered for closer surveillance, potentially addressing the residual interval CRC risk that persists even after negative or low-risk colonoscopy [10]. This tailored approach may support precision surveillance in routine care.Fig. 2Proposed algorithms for integrating mucus cfDNA KRAS mutation testing as a colonoscopy-integrated adjunct within FIT-based colorectal cancer screening pathways: (A) Standard FIT-based pathway for adults aged 50–74 years; (B) Enhanced pathway with post-colonoscopy cfDNA adjunct for risk stratification and surveillance optimization.Fig 2 dummy alt text Proposed algorithms for integrating mucus cfDNA KRAS mutation testing as a colonoscopy-integrated adjunct within FIT-based colorectal cancer screening pathways: (A) Standard FIT-based pathway for adults aged 50–74 years; (B) Enhanced pathway with post-colonoscopy cfDNA adjunct for risk stratification and surveillance optimization. Expanding beyond KRAS to multi-gene panels, including BRAF V600E (prevalent in 10%–15% of CRC) and methylation markers, could substantially bolster diagnostic accuracy by capturing complementary oncogenic pathways [27,28]. Future iterations might evaluate how these molecular insights can be integrated with stool-based first-line screening pathways [8]. In the current post-colonoscopy sampling framework, mucus cfDNA is best positioned as a colonoscopy adjunct for molecular risk assessment and confirmatory monitoring, particularly in high-risk cohorts [21]. This single-center pilot study has some limitations. First, it is limited by its modest sample size, its focus on a single gene target (KRAS), and the fact that KRAS mutations account for only approximately 40% of adenoma pathogenesis, which imposes an inherent ceiling on achievable sensitivity. Current molecular diagnostic strategies increasingly employ multi-marker approaches combining genetic and epigenetic alterations. In particular, APC alterations could complement KRAS for detecting early adenomatous changes, BRAF V600E may help capture lesions arising through the serrated pathway, TP53 may provide additional information regarding later-stage neoplastic progression, and epigenetic DNA methylation markers could further improve sensitivity and specificity by capturing complementary molecular pathways. Future incorporation of additional biomarkers based on this platform could improve diagnostic coverage and broaden clinical applicability. Second, the use of conservative imputation for missing data may underestimate the true diagnostic performance. Larger, multicenter validation studies are therefore warranted. In addition, bowel preparation quality and intraprocedural irrigation may influence mucus composition and cfDNA concentration. Although patients with inadequate bowel preparation (BBPS < 6) were excluded, variations in cleansing techniques and water irrigation during colonoscopy could theoretically dilute or redistribute surface mucus. Because sampling was performed after completion of the examination, the collected mucus likely represents a cumulative surface layer rather than site-specific secretion. Therefore, the current assay cannot establish direct spatial correlation between KRAS mutations detected in mucus cfDNA and the anatomical location of individual neoplastic lesions. In this pilot study, mucus cfDNA findings should be interpreted as colon-level molecular risk signals rather than lesion-localizing results. Future studies incorporating segmental mucus sampling, lesion-directed collection before biopsy or resection, matched tissue genotyping, standardized irrigation protocols, and quantitative cfDNA measurements are needed to clarify the anatomical origin, biological validity, and procedural factors affecting mucus cfDNA signals. Conclusion The findings of this prospective pilot study support colonoscope-derived mucus as an untapped, localized reservoir for detecting KRAS mutations through mucus cfDNA analysis. Rather than replacing conventional FIT, CEA, colonoscopy, or histopathological assessment, this approach may serve as a colonoscopy-integrated molecular adjunct for detecting precancerous colorectal neoplasia and refining post-procedure risk stratification. While refinement is needed to reduce biological noise and improve specificity, this technique may help transform colonoscopy from a purely morphological examination into a platform for precision molecular risk assessment. Ultimately, integrating mucus biomarkers into screening and surveillance pathways could support more personalized CRC prevention. Funding' statement This study was sponsored and funded by VtR INC. (Taipei, Taiwan). Laboratory testing and assay development were performed by HEALERA BIOSCIENCES INC. (Taipei, Taiwan), a wholly owned subsidiary of VtR INC., under the direction and supervision of the study investigators. CRediT authorship contribution statement Fang-Chin Hsu: Writing – original draft, Formal analysis, Data curation. Ta-Wei Pu: Validation, Methodology, Data curation, Conceptualization. Jung-Chun Lin: Supervision, Resources, Methodology, Investigation. Chao-Yang Chen: Writing – review & editing, Supervision, Formal analysis, Conceptualization. Declaration of competing interest The authors declare no competing financial or personal interests that could have appeared to influence the work reported in this paper.