Enhancing Non-Invasive Colorectal Cancer Screening with Stool DNA Methylation Markers and LightGBM Machine Learning

Enhancing Non-Invasive Colorectal Cancer Screening with Stool DNA Methylation Markers and LightGBM Machine Learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Non-Invasive Colorectal Cancer Screening with Stool DNA Methylation Markers and LightGBM Machine Learning Yi Xiang, Na Yang, Yunlong Zhu, Gangfeng Zhu, Zenghong Lu, Shi Geng, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3857174/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective: This study evaluates the effectiveness of stool DNA methylation markers CNRIP1, SFRP2, and VIM, along with Fecal Occult Blood Testing (FOBT), in the non-invasive screening of colorectal cancer (CRC), further integrating these markers with the Light Gradient Boosting Machine (LightGBM) machine learning (ML) algorithm. Methods: The study analyzed 100 stool samples, comprising 50 CRC patients and 50 healthy controls, from the First Affiliated Hospital of Gannan Medical University. Methylation Specific PCR (MSP) was used for assessing the methylation status of CNRIP1, SFRP2, and VIM gene promoters. FOBT was performed in parallel. Diagnostic performance was assessed using Receiver Operating Characteristic (ROC) curve analysis, and a LightGBM-based ML model was developed, incorporating these methylation markers and FOBT results. Results: ROC analysis demonstrated that SFRP2 had the highest diagnostic accuracy with an AUC of 0.87 (95% CI: 0.794-0.946) and a sensitivity of 0.88. CNRIP1 and VIM also showed substantial screening effectiveness, with AUCs of 0.83 and 0.80, respectively. FOBT, in comparison, had a lower predictive value with an AUC of 0.67. The LightGBM-based ML model significantly outperformed individual markers, achieving a high AUC of 0.95 (95% CI: 0.916-0.991). However, the sensitivity of the ML model was 0.78, suggesting a need for improvement in correctly identifying all positive CRC cases. Conclusion: Stool DNA methylation markers CNRIP1, SFRP2, and VIM exhibit high sensitivity in non-invasive CRC screening. The integration of these biomarkers with the LightGBM ML algorithm enhances the diagnostic accuracy, offering a promising approach for early CRC detection. Colorectal Cancer Non-invasive Screening DNA Methylation LightGBM Machine Learning Multi-Target Stool DNA Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Colorectal cancer (CRC) poses a substantial global health challenge, being especially prevalent in countries like China, India, and Russia [ 1 ] . These nations, despite hosting approximately 40% of the world's population, encounter a disproportionate burden, witnessing almost half of the newly diagnosed cases and related deaths [ 2 ] . CRC is a leading cancer of the digestive tract, and its incidence and mortality rates are ranked third and second, respectively, on a global scale. The increased prevalence of CRC, marked by 1.88 million new cases and 910,000 deaths in 2020 alone, underscores the critical necessity for efficacious diagnostic strategies [ 3 ] . The crucial role of early and timely detection cannot be overstated in reducing CRC-associated mortality. While colonoscopy with biopsy is the established gold standard for CRC diagnosis [ 4 ] , its invasive nature, the potential discomfort involved, and the high costs often lead to reluctance among individuals, fostering a growing demand for non-invasive screening options for CRC [ 5 ] . Several recent studies have shed light on the feasibility of leveraging tumor-specific DNA alterations found in body fluids like stool, serum, and urine as markers for CRC [ 6 – 8 ] . Multi-Target stool DNA (MT-sDNA) tests, which exploit the periodic shedding of intestinal epithelial cells, have emerged as potential non-invasive CRC detection tools, showcasing efficacy, specificity, and cost-effectiveness [ 9 ] . Within the ambit of MT-sDNA tests, specific molecules have been spotlighted as playing key roles in the genesis and development of CRC. CNRIP1, associated with the cannabinoid receptor 1 (CB1) system, is significant in metabolic regulation [ 10 ] . The methylation detection of CNRIP1 in stool samples has been proven to be a feasible non-invasive screening method for both CRC and its precursor lesions [ 11 ] . Similarly, SFRP2 serves as a pivotal tumor suppressor, with its methylation, leading to downregulation, being linked with CRC progression by inhibiting the Wnt/β-catenin signaling pathway [ 12 ] . The combination of SFRP2 methylation detection with fecal immunochemical tests (FIT) has yielded promising results in enhancing the accuracy of CRC screening [ 13 ] . Furthermore, Vimentin (VIM) is central to various cellular functions, including the epithelial-mesenchymal transition, a fundamental process in tumor invasion and metastasis [ 14 ] . Given its methylation is observed in a substantial number of CRC cases, VIM emerges as a promising molecular marker for early CRC detection. Recent advancements in the field of bioinformatics have significantly leveraged the potential of machine learning (ML) algorithms in the domain of oncological diagnostics and prognostics [ 15 , 16 ] . Notably, the construction of DNA methylation models using ML techniques has opened new vistas in the accurate prediction and early detection of malignancies [ 17 ] . Among various ML algorithms, Light Gradient Boosting Machine (LightGBM) has emerged as a particularly promising tool due to its efficiency in handling large-scale data and its superior performance in terms of speed and accuracy [ 18 , 19 ] . LightGBM's unique gradient-based one-side sampling and exclusive feature bundling techniques enable the handling of extensive datasets with a higher degree of precision and lower computational resource demand [ 20 ] . In the context of CRC screening, LightGBM's application in analyzing complex methylation patterns offers a potential paradigm shift. In light of the aforementioned context, this study is designed to meticulously examine the methylation status of CNRIP1, SFRP2, and VIM gene promoters in stool samples from both CRC patients and controls using Methylation Specific PCR (MSP). The primary objective is to assess the potential of these genes as non-invasive CRC screening tools. Additionally, this research will incorporate the use of the LightGBM algorithm to develop a ML model for these three genes, with the aim of enhancing the precision and accuracy of non-invasive CRC screening methodologies. This novel approach is anticipated to significantly advance the field of non-invasive CRC diagnostics. Materials and Methods This prospective, single-center study was ethically conducted following the Declaration of Helsinki and approved by the Ethics Committee of the First Affiliated Hospital of Gannan Medical University (Ethical Approval No: LLSC-2023-203). Written informed consent was obtained from all subjects before the study. Participant Recruitment and Baseline Data Collection Participants were meticulously selected based on stringent inclusion and exclusion criteria. The participant cohort was constituted of 100 participants, bifurcated into 50 CRC patients and 50 healthy controls, all of whom were enlisted from the Departments of Gastroenterology, Gastrointestinal Surgery, and Oncology at the First Affiliated Hospital of Gannan Medical University, Jiangxi Province. Baseline demographic and clinical information, encompassing participant identity, gender, age, diagnosis, histopathological classification, tumor site, tumor dimensions, CRC TNM staging, and histological differentiation of CRC tissues, were meticulously recorded. The gold standard for CRC diagnosis remained colonoscopic observation corroborated by histopathological evaluation from biopsy specimens. The schematic representation of subject recruitment and the experimental workflow for stool sample analysis is illustrated in Fig. 1 . Inclusion and Exclusion Criteria Inclusion Criteria: 1. Patients histopathologically diagnosed with colorectal adenocarcinoma via colonoscopy. 2. Control participants: Individuals devoid of CRC, adenomas, familial adenomatous polyposis, or any familial malignancy lineage. Exclusion Criteria: Individuals with incomplete medical documentation, those undergoing menstruation or pregnancy, incapable of rendering sufficient stool samples, presenting diarrhea or hemorrhoids, those devoid of electronic colonoscopy and pathological diagnosis, individuals post-endoscopic treatment, surgery, radiotherapy, chemotherapy, immunotherapy, or those with CRC recurrence were systematically excluded. Stool Sample Procurement and DNA Extraction Prior to any surgical or endoscopic intervention, participants furnished approximately 10g of fecal matter, either spontaneously excreted or post-laxative induction, utilizing a specialized fecal collection apparatus. Immediate storage at -80°C was ensured within two hours post-collection [ 21 ] . The methylation profiles of CNRIP1, SFRP2, and VIM gene promoters in stool samples from both the CRC and control cohorts were delineated using the MSP methodology. The protocol involved: 1. DNA isolation from fecal specimens leveraging the QIAamp Fast DNA Stool Mini Kit. 2. Ascertainment of DNA concentration and purity via spectrophotometry. An OD260 value equating to 50µg/mL of double-stranded DNA and an OD260/OD280 purity ratio ranging between 1.8 and 2.0 were deemed acceptable. Validation of Genomic DNA Reference To identify whether the extracted DNA contained human genomic DNA, PCR amplification targeting the human β-actin gene was performed using Unmethylated human control DNA as a positive control, with water as the negative control. The β-actin primer sequences (Table 1) were synthesized by a biotech company. The amplification conditions for the β-actin gene were as follows: 95°C for 5 minutes for 1 cycle; 95°C for 30 seconds, 60°C for 30 seconds, and 72°C for 35 seconds for 35 cycles; and a final extension at 72°C for 5 minutes (Fig. 2 A). Post-PCR, the products were separated using agarose gel electrophoresis: 5µL of each amplified sample was mixed with 6× Loading Buffer and loaded into the gel wells; 5µL of DNA Marker solution was added as a reference. The gel was run at 100V for 30 minutes. The gel was subsequently visualized and imaged. Sodium Bisulfite Modification Following gel electrophoresis of the aforementioned PCR products and the identification of respective target bands, bisulfite conversion was carried out. The procedure was performed as per the EpiTect Bisulfite Kits manual. This process chemically modified un-methylated cytosines to uracil, while methylated cytosines remained unchanged. The resultant DNA solution was then set aside for the subsequent MSP reaction. Methylation-Specific PCR (MSP) Analysis DNA, post-bisulfite conversion, served as the template for MSP. The PCR conditions were as follows: an initial denaturation at 95°C for 5 minutes, followed by 40 cycles of denaturation at 95°C for 30 seconds, annealing at 54–62°C for 30 seconds, and extension at 72°C for 30 seconds, with a final extension at 72°C for 10 minutes. Methylated human control DNA (bisulfite converted) and Unmethylated human control DNA (bisulfite converted) were used as positive controls for methylated and unmethylated conditions, respectively, with water serving as the negative control. Amplified products were separated using 3% agarose gel electrophoresis and subsequently visualized. The primer sequences for CNRIP1, SFRP2, and VIM (Table 1) were synthesized by a biotech company. MSP Sensitivity Assessment To evaluate the sensitivity of the primers and PCR reaction conditions, a serial dilution assay was performed using Methylated human control DNA (bisulfite converted) and Unmethylated human control DNA (bisulfite converted) as standards representing 100% and 0% methylation, respectively. Mixed standards resulted in final methylation proportions of 100%, 99%, 90%, 50%, 10%, 1%, and 0%. Amplification conditions and subsequent gel electrophoresis were conducted as previously described (Fig. 2 B-D). Criteria for Methylation Detection Criteria for determining gene methylation in samples were the appearance of only a methylated band or both methylated and unmethylated bands concurrently. A sample was considered non-methylated if only an unmethylated band appeared. For combined detection results, a positive result was defined by the presence of methylation in at least one gene. Immunochemical fecal occult blood test (FOBT) An immunochemical FOBT was performed using the Fecal Occult Blood Gold Gel Stripe (W.H.P.M., Beijing, China) in accordance with the manufacturer's instructions. Briefly, feces samples were taken from multiple spots. A homogenized suspension was tested using the testing strips. Samples in which the control line and reaction line appeared simultaneously and clearly in the test strip within 5 min were designated as positive. Samples in which only a control line appeared in the test strip within 5 min were considered negative. Any other result was an invalidation and the sample testing was repeated. Machine learning (ML) algorithm To evaluate the model's performance accurately, we split the dataset into training and test sets. The training set was used to train the model, while the test set remained untouched until the final evaluation. A standard split ratio of 4:1 was applied to ensure an unbiased assessment. LightGBM, a gradient boosting framework that uses tree-based learning algorithms, was chosen as the primary model for this prediction task [ 19 ] . LightGBM is known for its high training speed and low memory footprint. It utilizes a histogram-based algorithm, accelerating the model training process. LightGBM employs a tree-based model, iteratively generating weak classifiers and boosting the model's accuracy through gradient boosting [ 22 ] . Fine-tuning the hyperparameters of the LightGBM model was a critical step to achieve optimal performance. We utilized techniques such as grid search and random search to explore the hyperparameter space efficiently. The better hyperparameters are finally selected to build our model. The final step involved evaluating the model on the previously untouched test set. The Area under the ROC curve (AUC) as the primary measure of modeling. Additionally, performance metrics such as accuracy, sensitivity, specificity, PPV, NPV and F1-score were calculated to assess the model's effectiveness. Statistical Analysis All statistical analyses were conducted using SPSS 21.0 (IBM). Continuous variables with a normal distribution were expressed as mean ± standard deviation (SD), and intergroup age differences were assessed using independent samples t-tests. Categorical variables were presented as frequencies (percentages) and were assessed using the chi-squared (χ 2 ) test; Fisher's exact test was employed when the case number was below 40. The accuracy of the diagnostic tests was assessed using the Receiver Operating Characteristic Curve (ROC), with the AUC and its 95% confidence interval (CI) calculated. A P -value threshold of < 0.05 was considered statistically significant. Results A total of 100 stool samples were collected, comprising 50 CRC cases and 50 healthy controls, to validate the sensitivity and specificity of fecal gene methylation tests, compared to FOBT. Patient Demographics and Tumor Characteristics This study analyzed 100 samples from residents of the Gannan region. The mean age was 62.1 years (range: 40-81 years) in the CRC group and 60.3 years (range: 45-80 years) in the control group, with no significant age difference between the groups ( P = 0.338). Gender distribution was also similar, with 68.0% males in the CRC group and 64.0% in the control group ( P = 0.673, Table 2). Tumor location varied: 18.0% in the ascending colon, 2.0% in the transverse colon, 14.0% each in the descending and sigmoid colon, and 52.0% in the rectum. AJCC staging indicated 44.0% in stages I-II and 56.0% in stages III-IV. Tumor invasion was categorized as T1-T2 in 30.0% and T3-T4 in 70.0% of cases; lymph node metastasis was N0-N1 in 80.0% and N2-N3 in 20.0%; distant metastasis was M0 in 84.0% and M1 in 16.0%. Tumor differentiation was predominantly poorly differentiated (48.0%), followed by moderately (42.0%) and highly differentiated (10%). Assessment of Stool DNA Methylation and FOBT in CRC Screening The DNA methylation tests were performed on all 100 stool samples to validate the sensitivity of independent gene methylation tests. The positive detection rates of SFRP2, CNRIP1, and VIM gene promoters and FOBT in the CRC group were 88.0%, 84.0%, 78.0%, and 50.0%, respectively, significantly higher than in the control group (14.0%, 18.0%, 18.0%, and 16.0%, respectively; P < 0.001 for all comparisons, Figure 3). In the evaluation of stool DNA methylation for CRC screening, the ROC curve analysis revealed varied diagnostic efficacies (Figure4 and Table 3). The AUC for SFRP2 methylation was 0.87 (95% CI: 0.794-0.946), indicating a high predictive accuracy, with a sensitivity (true positive rate, TPR) of 0.88. Methylation of CNRIP1 also demonstrated substantial screening effectiveness, with an AUC of 0.83 (95% CI: 0.745-0.915) and a sensitivity of 0.84. VIM methylation showed a slightly lower, yet notable, diagnostic potential with an AUC of 0.80 (95% CI: 0.709-0.891) and a sensitivity of 0.78. In comparison, traditional FOBT presented a significantly lower predictive value, evidenced by its AUC of 0.67 (95% CI: 0.563-0.777) and a sensitivity of only 0.50. These findings underscore the superior sensitivity of fecal DNA methylation markers, specifically SFRP2, CNRIP1, and VIM, in the screening for CRC compared to traditional FOBT. Enhanced CRC Screening Performance by LightGBM-Based ML-Model The ROC curve analysis for the ML model, constructed using the LightGBM algorithm and based on the methylation of the three genes (CNRIP1, SFRP2, and VIM) combined with FOBT results, demonstrated a significant enhancement in the non-invasive screening efficacy for CRC (Figure4). The ML-model achieved a notably high AUC of 0.95 (95% CI: 0.916-0.991), indicating its superior predictive accuracy in identifying CRC cases. However, the TPR was recorded at 0.78 (Table 3), suggesting that while the model has a high overall accuracy, there is still room for improvement in its ability to correctly identify all positive CRC cases. This substantial improvement in the AUC value, as compared to the individual biomarkers, highlights the potential of integrating advanced ML techniques. Discussion This study marks a significant advancement in non-invasive CRC screening by combining stool DNA methylation markers (CNRIP1, SFRP2, and VIM) with the LightGBM ML algorithm. Our findings highlight the superior diagnostic accuracy of these methylation markers, particularly SFRP2, over traditional FOBT. The LightGBM-based model further enhanced this accuracy, demonstrating a high AUC in the ROC analysis. Our study represents a pioneering exploration in the realm of CRC screening by synergistically integrating MT-sDNA testing with advanced ML techniques, specifically the LightGBM. This novel approach marks a significant leap in the field, blending the sensitivity of molecular biomarkers with the analytical prowess of artificial intelligence (AI). Historically, the utilization of MT-sDNA in CRC screening has been limited by its reliance on individual biomarker analysis [ 23 , 24 ] . However, our study demonstrates that with the continual advancement in ML algorithms, there is a remarkable potential to cohesively amalgamate these disparate methylation signals into a more coherent and robust diagnostic tool. By harnessing the computational power of LightGBM [ 25 , 26 ] , our study successfully navigated the intricate landscape of methylation patterns, offering a more nuanced and comprehensive approach to CRC screening. The LightGBM algorithm represents a sophisticated form of AI that aids physicians in processing data and making informed decisions by uncovering hidden interactions within malignant tumor datasets [ 27 , 28 ] . Compared to traditional statistical techniques, this algorithm excels in handling the diversity of non-linear, high-dimensional, and complex data [ 29 , 30 ] . Utilizing machine learning algorithms such as LightGBM for predicting disease risk has emerged as a major area of research in the field of medical big data. In our research, the integration of the LightGBM ML algorithm with MT-sDNA testing has led to a notable improvement in the diagnostic efficiency for CRC screening. This combination has been particularly effective in enhancing the specificity and sensitivity of CRC detection. Our study showed that the application of LightGBM increased the AUC to 0.95, a significant improvement compared to the AUCs observed with traditional MT-sDNA tests alone. This higher AUC indicates a superior balance between sensitivity and specificity, crucial for reducing both false positives and false negatives in CRC screening. These results underscore the transformative potential of ML in refining the diagnostic capabilities of MT-sDNA tests. By accurately distinguishing between benign and malignant colorectal conditions [ 31 ] , LightGBM has the potential to decrease unnecessary diagnostic colonoscopies [ 32 ] , thereby reducing the burden on both patients and healthcare systems [ 33 ] . The application of LightGBM to other MT-sDNA tests like Cologuard [ 34 ] could revolutionize CRC screening, offering more accurate, non-invasive options for early cancer detection [ 35 ] . This innovative approach exemplifies the integration of AI in enhancing the precision of medical diagnostics. Considering the screening frequency and its implications on cost-effectiveness, patient compliance, and diagnostic sensitivity, the combined MT-sDNA and LightGBM approach could serve as an effective interim screening method, especially in identifying cases that might develop during the long interval between colonoscopies [ 9 , 36 ] . This is particularly pertinent considering the progressive nature of CRC, where the development of malignancies could occur within the decade-long interval of traditional endoscopic screenings [ 37 ] . The higher sensitivity and specificity of this combined method positions it as a reliable, non-invasive test that could be administered more frequently than colonoscopy, without subjecting patients to the associated risks and discomforts of invasive procedures. Future research may explore the integration of such methylation-based screening into routine healthcare practices [ 38 , 39 ] , evaluating its cost-effectiveness and operational feasibility on a larger scale. Despite the promising findings of our study, it is imperative to acknowledge its limitations and suggest avenues for future research. One notable limitation is the relatively small sample size and the study's confinement to a single center, which may limit the generalizability of our results. Future studies should aim to include a larger, more diverse population across multiple centers to validate our findings and ensure their applicability to broader demographic groups. Additionally, while our study primarily focused on CNRIP1, SFRP2, and VIM gene promoters, exploring additional methylation markers could further enhance the diagnostic accuracy and reliability of the CRC screening process. Another aspect warranting further investigation is the refinement and optimization of the LightGBM-based ML model. While our model demonstrated high accuracy, the true positive rate suggests room for improvement in its predictive capabilities, particularly in diverse clinical settings. Future research could explore the integration of more complex algorithms or the inclusion of additional clinical parameters to improve the model's sensitivity and specificity. Addressing these limitations will not only solidify the foundation laid by this study but also expand the horizons of non-invasive CRC diagnostics, potentially leading to more personalized and effective screening strategies. Conclusion This study illustrates the significant potential of integrating stool DNA methylation markers CNRIP1, SFRP2, and VIM with the ML algorithm - LightGBM in enhancing the accuracy of non-invasive CRC screening. Our findings demonstrate that this innovative approach surpasses traditional screening methods in diagnostic efficacy, offering a promising direction for early and more personalized CRC detection in clinical practice. Abbreviations CRC: Colorectal Cancer; CNRIP1: Cannabinoid Receptor Interacting Protein 1; SFRP2: Secreted Frizzled-Related Protein 2; VIM: Vimentin; qPCR: Quantitative Polymerase Chain Reaction; MSP: Methylation-Specific PCR; FOBT: Fecal Occult Blood Testing; MT-sDNA: Multi-Target Stool DNA; ML: Machine Learning; AI: Artificial Intelligence; AJCC: American Joint Committee on Cancer; TNM: Tumor, Node, Metastasis; OD: Optical Density; TPR: True Positive Rate; TNR: True Negative Rate; FPR: False Positive Rate; PPV, Positive Predictive Value; NPV: Negative Predictive Value; FIT: Fecal Immunochemical Test; AUC: Area Under the ROC Curve; ROC: Receiver Operating Characteristic Curve; CI: Confidence Interval; SD: Standard Deviation. Declarations Authors’ contributions Yi Xiang: Performed statistical analyses; Na Yang: Took the lead in writing the manuscript; Yunlong Zhu: Managed the collection and arrangement of the data; Gangfeng Zhu and Liangjian Zheng: Reviewed and made significant revisions to the initial draft; Shi Geng and Xiaofei Feng: Engaged in the clinical practices associated with this study; Zenghong Lu and Rui Zhu: Provided technical support throughout the research process; Xueming Xu and Xiangcai Wang: Were responsible for primary data collection; Tianlei Zheng and Li Huang: Oversaw the overall direction and planning of the project, and designed the research topic. All authors read and approved the final manuscript. Conflict of interest The authors have declared that no competing interest exists. Data availability The data included in the study are available from the corresponding author upon reasonable request. Acknowledgements The authors acknowledge all the clinical and research staff from the research centers. Financial support The Opening Project of Jiangsu Key Laboratory of Xuzhou Medical University (XZSYSKF2021030). Project supported by the Affiliated Hospital of Xuzhou Medical University (2022ZL26). Ethics Statement The study was approved by the Ethics Review Board of The First Affiliated Hospital of Gannan Medical University (Ganzhou, China). Written informed consent was obtained from all subjects before the study. References Kocarnik JM, Compton K, Dean FE, et al. Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life Years for 29 Cancer Groups From 2010 to 2019: A Systematic Analysis for the Global Burden of Disease Study 2019. JAMA Oncol 2022; 8 (3): 420-44. Goss PE, Strasser-Weippl K, Lee-Bychkovsky BL, et al. Challenges to effective cancer control in China, India, and Russia. Lancet Oncol 2014; 15 (5): 489-538. Morgan E, Arnold M, Gini A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN. Gut 2023; 72 (2): 338-44. Iovanescu D, Frandes M, Lungeanu D, Burlea A, Miutescu BP, Miutescu E. Diagnosis reliability of combined flexible sigmoidoscopy and fecal-immunochemical test in colorectal neoplasia screening. Onco Targets Ther 2016; 9 : 6819-28. Yu J, Feng Q, Wong SH, et al. Metagenomic analysis of faecal microbiome as a tool towards targeted non-invasive biomarkers for colorectal cancer. Gut 2017; 66 (1): 70-8. Marcuello M, Vymetalkova V, Neves RPL, et al. Circulating biomarkers for early detection and clinical management of colorectal cancer. Mol Aspects Med 2019; 69 : 107-22. Bach S, Paulis I, Sluiter NR, et al. Detection of colorectal cancer in urine using DNA methylation analysis. Sci Rep 2021; 11 (1): 2363. Liang JQ, Li T, Nakatsu G, et al. A novel faecal Lachnoclostridium marker for the non-invasive diagnosis of colorectal adenoma and cancer. Gut 2020; 69 (7): 1248-57. Ebner DW, Eckmann JD, Burger KN, et al. Detection of Postcolonoscopy Colorectal Neoplasia by Multi-target Stool DNA. Clin Transl Gastroenterol 2021; 12 (6): e00375. Booth WT, Walker NB, Lowther WT, Howlett AC. Cannabinoid Receptor Interacting Protein 1a (CRIP1a): Function and Structure. Molecules 2019; 24 (20). Lind GE, Danielsen SA, Ahlquist T, et al. Identification of an epigenetic biomarker panel with high sensitivity and specificity for colorectal cancer and adenomas. Mol Cancer 2011; 10 : 85. Yang Q, Huang T, Ye G, Wang B, Zhang X. Methylation of SFRP2 gene as a promising noninvasive biomarker using feces in colorectal cancer diagnosis: a systematic meta-analysis. Sci Rep 2016; 6 : 33339. Xiao Z, Li B, Wang G, et al. Validation of methylation-sensitive high-resolution melting (MS-HRM) for the detection of stool DNA methylation in colorectal neoplasms. Clin Chim Acta 2014; 431 : 154-63. Niknami Z, Eslamifar A, Emamirazavi A, Ebrahimi A, Shirkoohi R. The association of vimentin and fibronectin gene expression with epithelial-mesenchymal transition and tumor malignancy in colorectal carcinoma. Excli j 2017; 16 : 1009-17. Thagaard J, Broeckx G, Page DB, et al. Pitfalls in machine learning-based assessment of tumor-infiltrating lymphocytes in breast cancer: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer. J Pathol 2023; 260 (5): 498-513. Primakov SP, Ibrahim A, van Timmeren JE, et al. Automated detection and segmentation of non-small cell lung cancer computed tomography images. Nat Commun 2022; 13 (1): 3423. Liang N, Li B, Jia Z, et al. Ultrasensitive detection of circulating tumour DNA via deep methylation sequencing aided by machine learning. Nat Biomed Eng 2021; 5 (6): 586-99. Hu X, Yin S, Zhang X, et al. Blood pressure stratification using photoplethysmography and light gradient boosting machine. Front Physiol 2023; 14 : 1072273. Rufo DD, Debelee TG, Ibenthal A, Negera WG. Diagnosis of Diabetes Mellitus Using Gradient Boosting Machine (LightGBM). Diagnostics (Basel) 2021; 11 (9). Zeng X. Length of Stay Prediction Model of Indoor Patients Based on Light Gradient Boosting Machine. Comput Intell Neurosci 2022; 2022 : 9517029. Yang F, Sun J, Luo H, et al. Assessment of fecal DNA extraction protocols for metagenomic studies. Gigascience 2020; 9 (7). Seto H, Oyama A, Kitora S, et al. Gradient boosting decision tree becomes more reliable than logistic regression in predicting probability for diabetes with big data. Sci Rep 2022; 12 (1): 15889. Carethers JM. Fecal DNA Testing for Colorectal Cancer Screening. Annu Rev Med 2020; 71 : 59-69. Redwood DG, Dinh TA, Kisiel JB, et al. Cost-Effectiveness of Multitarget Stool DNA Testing vs Colonoscopy or Fecal Immunochemical Testing for Colorectal Cancer Screening in Alaska Native People. Mayo Clin Proc 2021; 96 (5): 1203-17. Park DJ, Park MW, Lee H, Kim YJ, Kim Y, Park YH. Development of machine learning model for diagnostic disease prediction based on laboratory tests. Sci Rep 2021; 11 (1): 7567. Zhang Y, Jiang Z, Chen C, Wei Q, Gu H, Yu B. DeepStack-DTIs: Predicting Drug-Target Interactions Using LightGBM Feature Selection and Deep-Stacked Ensemble Classifier. Interdiscip Sci 2022; 14 (2): 311-30. Park YM, Lee BJ. Machine learning-based prediction model using clinico-pathologic factors for papillary thyroid carcinoma recurrence. Sci Rep 2021; 11 (1): 4948. Ren J, Zhou X, Guo W, Feng K, Huang T, Cai YD. Identification of Methylation Signatures and Rules for Sarcoma Subtypes by Machine Learning Methods. Biomed Res Int 2022; 2022 : 5297235. Gardner W, Cutts SM, Phillips DR, Pigram PJ. Understanding mass spectrometry images: complexity to clarity with machine learning. Biopolymers 2021; 112 (4): e23400. Liu L, Meng Q, Weng C, Lu Q, Wang T, Wen Y. Explainable deep transfer learning model for disease risk prediction using high-dimensional genomic data. PLoS Comput Biol 2022; 18 (7): e1010328. Grosu S, Wesp P, Graser A, et al. Machine Learning-based Differentiation of Benign and Premalignant Colorectal Polyps Detected with CT Colonography in an Asymptomatic Screening Population: A Proof-of-Concept Study. Radiology 2021; 299 (2): 326-35. Vahdat V, Alagoz O, Chen JV, Saoud L, Borah BJ, Limburg PJ. Calibration and Validation of the Colorectal Cancer and Adenoma Incidence and Mortality (CRC-AIM) Microsimulation Model Using Deep Neural Networks. Med Decis Making 2023; 43 (6): 719-36. Nemlander E, Ewing M, Abedi E, et al. A machine learning tool for identifying non-metastatic colorectal cancer in primary care. Eur J Cancer 2023; 182 : 100-6. A stool DNA test (Cologuard) for colorectal cancer screening. Jama 2014; 312 (23): 2566. Zhao F, Bai P, Xu J, et al. Efficacy of cell-free DNA methylation-based blood test for colorectal cancer screening in high-risk population: a prospective cohort study. Mol Cancer 2023; 22 (1): 157. Pickhardt PJ, Graffy PM, Weigman B, Deiss-Yehiely N, Hassan C, Weiss JM. Diagnostic Performance of Multitarget Stool DNA and CT Colonography for Noninvasive Colorectal Cancer Screening. Radiology 2020; 297 (1): 120-9. Nakao SK, Fassler S, Sucandy I, Kim S, Zebley DM. Colorectal cancer following negative colonoscopy: is 5-year screening the correct interval to recommend? Surg Endosc 2013; 27 (3): 768-73. Morgacheva D, Ryzhova M, Zheludkova O, Belogurova M, Dinikina Y. DNA methylation-based diagnosis confirmation in a pediatric patient with low-grade glioma: a case report. Front Pediatr 2023; 11 : 1256876. Pickles JC, Fairchild AR, Stone TJ, et al. DNA methylation-based profiling for paediatric CNS tumour diagnosis and treatment: a population-based study. Lancet Child Adolesc Health 2020; 4 (2): 121-30. Tables Table 1 Primer sequences and PCR product size obtained used for MSP assays Gene Primer Sequence (5’-3’) PCR product size (bp) CNRIP1 M F: 5′-GTGGAATTGCGTTTTTTTTC-3′ 163 R: 5'-TACTACTCGCGAATCCGAC-3' U F: 5'-GTGGAATTGTGTTTTTTTTT-3' 163 R: 5'-TACTACTCACAAATCCAAC-3' SFRP2 M F: 5′-GAGTAGTAGTGTTTATTTTA-3′ 138 R: 5'-CCGCTCTCTTCGCTAAATACGACTCG-3' U F: 5'-TTTTGGGTTGGAGTTTTTTGGAGTTGTGT-3' 145 R: 5'-AACCCACTCTCTTCACTAAATACAACTCA-3' VIM M F: 5′-TCGTTTCGAGGTTTTCGCGTTAGAGAC-3′ 216 R: 5'-CGACTAAAACTCGACCGACTCGCGA-3' U F: 5′-TTGGTGGATTTTTTGTTGGTTGATG-3′ 183 R: 5'-CACAACTTACCTTAACCCTTAAACTACTCA-3' PCR, polymerase chain reaction; MSP, methylation‑specific; bp, base pair; SFRP2, secreted frizzled‑related protein 2; VIM, vimentin; M, methylated; U, unmethylated; F, forward; R, reverse. Table 2 Baseline characteristics and DNA methylation status of CRC patients and healthy controls Characteristics CRC patients Healthy controls χ 2 / t P -value ( n =50) ( n =50) Age (years, ±s) 62.1±9.59 60.3±9.30 0.963 0.338 Gender [% ( n )] 0.178 0.673 Male 68.0% (34) 64.0% (32) Tumor location [% ( n )] Ascending colon Transverse colon Descending colon Sigmoid colon Rectum TNM classification [% ( n )] II III-IV Tumor invasion [% ( n )] T1-T2 T3-T4 Lymph node metastasis [% ( n )] N0-N1 N2-N3 Distant metastasis [% ( n )] M0 M1 Tumor differentiation [% ( n )] Poor Moderate Well 18.0% (9) 2.0% (1) 14.0% (7) 14.0% (7) 52.0% (26) 44.0% (22) 56.0% (28) 30.0% (15) 70.0% (35) 80.0% (40) 20.0% (10) 84.0% (42) 16.0% (8) 48.0% (24) 42.0% (21) 10% (5) Methylation of SFRP2 [% ( n )] 54.782 <0.001 Positive 88.0% (44) 14.0% (7) Methylation of CNRIP1 [% ( n )] 43.577 <0.001 Positive 84.0% (42) 18.0% (9) Methylation of VIM [% ( n )] 36.058 <0.001 Positive 78.0% (39) 18.0% (9) FOBT detection [% ( n )] 13.071 <0.001 Positive 50.0% (25) 16.0% (8) Abbreviations: CRC, Colorectal Cancer; FOBT, Fecal Occult Blood Testing. Table 3 Performance of stool DNA methylation, FOBT and the ML-model for CRC screening Variables AUC 95% CI TPR (Sensitivity) TNR (Specificity) FPR PPV NPV SFRP2 0.87 0.794-0.946 0.88 0.86 0.14 0.863 0.878 CNRIP2 0.83 0.745-0.915 0.84 0.82 0.18 0.824 0.837 VIM 0.80 0.709-0.891 0.78 0.82 0.18 0.813 0.788 FOBT 0.67 0.563-0.777 0.50 0.84 0.16 0.758 0.627 lightGBM model 0.95 0.916-0.991 0.78 0.91 0.01 0.875 0.833 Abbreviations: FOBT, Fecal Occult Blood Testing; CRC, Colorectal Cancer; TPR, True Positive Rate; TNR, True Negative Rate; PPV, Positive Predictive Value; NPV, Negative Predictive Value. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3857174","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":267263991,"identity":"c1ce38b7-4fc4-450e-a737-be83824b4354","order_by":0,"name":"Yi Xiang","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yi","middleName":"","lastName":"Xiang","suffix":""},{"id":267263992,"identity":"26e0000f-e86b-4357-a669-ed34f2932517","order_by":1,"name":"Na Yang","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Na","middleName":"","lastName":"Yang","suffix":""},{"id":267263993,"identity":"a6fee9c4-5fb3-4a25-a7fd-9aa9ee3ea681","order_by":2,"name":"Yunlong Zhu","email":"","orcid":"","institution":"Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yunlong","middleName":"","lastName":"Zhu","suffix":""},{"id":267263994,"identity":"66d8fe99-ab6c-484d-b4b8-af3f997a38cd","order_by":3,"name":"Gangfeng Zhu","email":"","orcid":"","institution":"Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Gangfeng","middleName":"","lastName":"Zhu","suffix":""},{"id":267263995,"identity":"733ba129-3538-4c56-8e06-f38208c5d388","order_by":4,"name":"Zenghong Lu","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zenghong","middleName":"","lastName":"Lu","suffix":""},{"id":267263996,"identity":"f437dd32-8b9e-45b8-9817-3eab5d018dbd","order_by":5,"name":"Shi Geng","email":"","orcid":"","institution":"Affiliated Hospital of Xuzhou Medical College","correspondingAuthor":false,"prefix":"","firstName":"Shi","middleName":"","lastName":"Geng","suffix":""},{"id":267263997,"identity":"91df864f-0003-4301-a6df-ab4a560c9194","order_by":6,"name":"Liangjian Zheng","email":"","orcid":"","institution":"Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liangjian","middleName":"","lastName":"Zheng","suffix":""},{"id":267263998,"identity":"d5a3792b-05cf-47ff-9926-efa932339b06","order_by":7,"name":"Xiaofei Feng","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaofei","middleName":"","lastName":"Feng","suffix":""},{"id":267263999,"identity":"a688a144-2bdc-45ee-97c7-b4bb8736d8ac","order_by":8,"name":"Rui Zhu","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Rui","middleName":"","lastName":"Zhu","suffix":""},{"id":267264000,"identity":"9c8a3779-015e-4feb-8fcb-7dc43be9d9eb","order_by":9,"name":"Xueming Xu","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xueming","middleName":"","lastName":"Xu","suffix":""},{"id":267264001,"identity":"b2e20a9b-7947-4677-a1d8-3ae8cab3461f","order_by":10,"name":"Xiangcai Wang","email":"","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiangcai","middleName":"","lastName":"Wang","suffix":""},{"id":267264002,"identity":"6da3adad-71dc-4629-93b4-49a4fdc085bd","order_by":11,"name":"Tianlei Zheng","email":"","orcid":"","institution":"Affiliated Hospital of Xuzhou Medical College","correspondingAuthor":false,"prefix":"","firstName":"Tianlei","middleName":"","lastName":"Zheng","suffix":""},{"id":267264003,"identity":"94e482c5-ba1c-40d7-b08d-8ee50473ce2f","order_by":12,"name":"Li Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYBACA3YGxgMPGBjkIFw2YrQwMzAcSGBgMCZdS2ID0VrMmZkPHEiouJM+P+yMAcOHssMM/LMb8GuxbGZLOJBw5lnuxts5Bowzzh1mkLhzgIDDDvMYHEhsO5y7cXaOATNv22EGA4kEQlr4PxxI/Hc43RCk5S9xWngYDiQ2HE6QlwZqYSROC5vBgYRjhw03SKcVHOw5l84jcYOQluPNDx98qDksLz87eeODH2XWcvwzCGhB6D0AjCAgzUOkeiCQbyBe7SgYBaNgFIwwAAAoNUg5Tp7wHgAAAABJRU5ErkJggg==","orcid":"","institution":"First Affiliated Hospital of Gannan Medical University","correspondingAuthor":true,"prefix":"","firstName":"Li","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2024-01-12 13:59:34","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3857174/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3857174/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":49713325,"identity":"fbe31d2f-6f1a-45a1-b61e-fdb2efe0a760","added_by":"auto","created_at":"2024-01-16 20:38:34","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":759710,"visible":true,"origin":"","legend":"\u003cp\u003eProcess of the subject recruitment and the framework of the study design.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-3857174/v1/58c5dd457c1b53839c25f1d5.png"},{"id":49713323,"identity":"6a4edb5d-54da-4e76-b457-6b323bdba553","added_by":"auto","created_at":"2024-01-16 20:38:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":369027,"visible":true,"origin":"","legend":"\u003cp\u003eMSP detects methylation status. (A) PCR products of M channel and U channel showed the presence of methylated and unmethylated genes, respectively. (B-D) Target gene methylated DNA,CNRIP1 (B), SFRP2 (C) and VIM (D) and human beta-actin genomic DNA were used as positive controls for methylation and non-methylation. Water as a negative control.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-3857174/v1/a6ac858376a5f0996119f37a.png"},{"id":49713322,"identity":"1fc0054d-a680-4c14-842a-85b32d6338b5","added_by":"auto","created_at":"2024-01-16 20:38:33","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":38202,"visible":true,"origin":"","legend":"\u003cp\u003eDNA\u003cstrong\u003e \u003c/strong\u003eMethylation ratio of CNRIP1, SFRP2 and VIM genes in CRC patients and healthy controls.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-3857174/v1/63245b8a2dcd07ddc6e3fdf7.png"},{"id":49713324,"identity":"636c2d85-f22c-4913-b60a-e94d5e67af79","added_by":"auto","created_at":"2024-01-16 20:38:34","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":69091,"visible":true,"origin":"","legend":"\u003cp\u003eROC curves of stool DNAmethylations, FOBT and the ML-model for CRC screening.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-3857174/v1/585aaa3f8e030820f23a77c4.png"},{"id":49777374,"identity":"a5028f84-b6d4-4f63-b1d4-47008632e764","added_by":"auto","created_at":"2024-01-17 21:23:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1070751,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3857174/v1/704eb081-af11-4c8f-ba3a-4be8eeafa073.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Non-Invasive Colorectal Cancer Screening with Stool DNA Methylation Markers and LightGBM Machine Learning","fulltext":[{"header":"Introduction","content":"\u003cp\u003eColorectal cancer (CRC) poses a substantial global health challenge, being especially prevalent in countries like China, India, and Russia \u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. These nations, despite hosting approximately 40% of the world's population, encounter a disproportionate burden, witnessing almost half of the newly diagnosed cases and related deaths \u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. CRC is a leading cancer of the digestive tract, and its incidence and mortality rates are ranked third and second, respectively, on a global scale. The increased prevalence of CRC, marked by 1.88\u0026nbsp;million new cases and 910,000 deaths in 2020 alone, underscores the critical necessity for efficacious diagnostic strategies \u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe crucial role of early and timely detection cannot be overstated in reducing CRC-associated mortality. While colonoscopy with biopsy is the established gold standard for CRC diagnosis \u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e, its invasive nature, the potential discomfort involved, and the high costs often lead to reluctance among individuals, fostering a growing demand for non-invasive screening options for CRC \u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. Several recent studies have shed light on the feasibility of leveraging tumor-specific DNA alterations found in body fluids like stool, serum, and urine as markers for CRC \u003csup\u003e[\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]\u003c/sup\u003e. Multi-Target stool DNA (MT-sDNA) tests, which exploit the periodic shedding of intestinal epithelial cells, have emerged as potential non-invasive CRC detection tools, showcasing efficacy, specificity, and cost-effectiveness \u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWithin the ambit of MT-sDNA tests, specific molecules have been spotlighted as playing key roles in the genesis and development of CRC. CNRIP1, associated with the cannabinoid receptor 1 (CB1) system, is significant in metabolic regulation \u003csup\u003e[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. The methylation detection of CNRIP1 in stool samples has been proven to be a feasible non-invasive screening method for both CRC and its precursor lesions \u003csup\u003e[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]\u003c/sup\u003e. Similarly, SFRP2 serves as a pivotal tumor suppressor, with its methylation, leading to downregulation, being linked with CRC progression by inhibiting the Wnt/β-catenin signaling pathway \u003csup\u003e[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]\u003c/sup\u003e. The combination of SFRP2 methylation detection with fecal immunochemical tests (FIT) has yielded promising results in enhancing the accuracy of CRC screening \u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Furthermore, Vimentin (VIM) is central to various cellular functions, including the epithelial-mesenchymal transition, a fundamental process in tumor invasion and metastasis \u003csup\u003e[\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/sup\u003e. Given its methylation is observed in a substantial number of CRC cases, VIM emerges as a promising molecular marker for early CRC detection.\u003c/p\u003e \u003cp\u003eRecent advancements in the field of bioinformatics have significantly leveraged the potential of machine learning (ML) algorithms in the domain of oncological diagnostics and prognostics \u003csup\u003e[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/sup\u003e. Notably, the construction of DNA methylation models using ML techniques has opened new vistas in the accurate prediction and early detection of malignancies \u003csup\u003e[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. Among various ML algorithms, Light Gradient Boosting Machine (LightGBM) has emerged as a particularly promising tool due to its efficiency in handling large-scale data and its superior performance in terms of speed and accuracy \u003csup\u003e[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. LightGBM's unique gradient-based one-side sampling and exclusive feature bundling techniques enable the handling of extensive datasets with a higher degree of precision and lower computational resource demand \u003csup\u003e[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/sup\u003e. In the context of CRC screening, LightGBM's application in analyzing complex methylation patterns offers a potential paradigm shift.\u003c/p\u003e \u003cp\u003eIn light of the aforementioned context, this study is designed to meticulously examine the methylation status of CNRIP1, SFRP2, and VIM gene promoters in stool samples from both CRC patients and controls using Methylation Specific PCR (MSP). The primary objective is to assess the potential of these genes as non-invasive CRC screening tools. Additionally, this research will incorporate the use of the LightGBM algorithm to develop a ML model for these three genes, with the aim of enhancing the precision and accuracy of non-invasive CRC screening methodologies. This novel approach is anticipated to significantly advance the field of non-invasive CRC diagnostics.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e This prospective, single-center study was ethically conducted following the Declaration of Helsinki and approved by the Ethics Committee of the First Affiliated Hospital of Gannan Medical University (Ethical Approval No: LLSC-2023-203). Written informed consent was obtained from all subjects before the study.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eParticipant Recruitment and Baseline Data Collection\u003c/h2\u003e \u003cp\u003eParticipants were meticulously selected based on stringent inclusion and exclusion criteria. The participant cohort was constituted of 100 participants, bifurcated into 50 CRC patients and 50 healthy controls, all of whom were enlisted from the Departments of Gastroenterology, Gastrointestinal Surgery, and Oncology at the First Affiliated Hospital of Gannan Medical University, Jiangxi Province.\u003c/p\u003e \u003cp\u003eBaseline demographic and clinical information, encompassing participant identity, gender, age, diagnosis, histopathological classification, tumor site, tumor dimensions, CRC TNM staging, and histological differentiation of CRC tissues, were meticulously recorded. The gold standard for CRC diagnosis remained colonoscopic observation corroborated by histopathological evaluation from biopsy specimens. The schematic representation of subject recruitment and the experimental workflow for stool sample analysis is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003eInclusion and Exclusion Criteria\u003c/h2\u003e \u003cp\u003eInclusion Criteria: 1. Patients histopathologically diagnosed with colorectal adenocarcinoma via colonoscopy. 2. Control participants: Individuals devoid of CRC, adenomas, familial adenomatous polyposis, or any familial malignancy lineage.\u003c/p\u003e \u003cp\u003eExclusion Criteria: Individuals with incomplete medical documentation, those undergoing menstruation or pregnancy, incapable of rendering sufficient stool samples, presenting diarrhea or hemorrhoids, those devoid of electronic colonoscopy and pathological diagnosis, individuals post-endoscopic treatment, surgery, radiotherapy, chemotherapy, immunotherapy, or those with CRC recurrence were systematically excluded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eStool Sample Procurement and DNA Extraction\u003c/h2\u003e \u003cp\u003e Prior to any surgical or endoscopic intervention, participants furnished approximately 10g of fecal matter, either spontaneously excreted or post-laxative induction, utilizing a specialized fecal collection apparatus. Immediate storage at -80\u0026deg;C was ensured within two hours post-collection \u003csup\u003e[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe methylation profiles of CNRIP1, SFRP2, and VIM gene promoters in stool samples from both the CRC and control cohorts were delineated using the MSP methodology. The protocol involved: 1. DNA isolation from fecal specimens leveraging the QIAamp Fast DNA Stool Mini Kit. 2. Ascertainment of DNA concentration and purity via spectrophotometry. An OD260 value equating to 50\u0026micro;g/mL of double-stranded DNA and an OD260/OD280 purity ratio ranging between 1.8 and 2.0 were deemed acceptable.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eValidation of Genomic DNA Reference\u003c/h2\u003e \u003cp\u003eTo identify whether the extracted DNA contained human genomic DNA, PCR amplification targeting the human β-actin gene was performed using Unmethylated human control DNA as a positive control, with water as the negative control. The β-actin primer sequences (Table\u0026nbsp;1) were synthesized by a biotech company. The amplification conditions for the β-actin gene were as follows: 95\u0026deg;C for 5 minutes for 1 cycle; 95\u0026deg;C for 30 seconds, 60\u0026deg;C for 30 seconds, and 72\u0026deg;C for 35 seconds for 35 cycles; and a final extension at 72\u0026deg;C for 5 minutes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003ePost-PCR, the products were separated using agarose gel electrophoresis: 5\u0026micro;L of each amplified sample was mixed with 6\u0026times; Loading Buffer and loaded into the gel wells; 5\u0026micro;L of DNA Marker solution was added as a reference. The gel was run at 100V for 30 minutes. The gel was subsequently visualized and imaged.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eSodium Bisulfite Modification\u003c/h2\u003e \u003cp\u003eFollowing gel electrophoresis of the aforementioned PCR products and the identification of respective target bands, bisulfite conversion was carried out. The procedure was performed as per the EpiTect Bisulfite Kits manual. This process chemically modified un-methylated cytosines to uracil, while methylated cytosines remained unchanged. The resultant DNA solution was then set aside for the subsequent MSP reaction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eMethylation-Specific PCR (MSP) Analysis\u003c/h2\u003e \u003cp\u003eDNA, post-bisulfite conversion, served as the template for MSP. The PCR conditions were as follows: an initial denaturation at 95\u0026deg;C for 5 minutes, followed by 40 cycles of denaturation at 95\u0026deg;C for 30 seconds, annealing at 54\u0026ndash;62\u0026deg;C for 30 seconds, and extension at 72\u0026deg;C for 30 seconds, with a final extension at 72\u0026deg;C for 10 minutes. Methylated human control DNA (bisulfite converted) and Unmethylated human control DNA (bisulfite converted) were used as positive controls for methylated and unmethylated conditions, respectively, with water serving as the negative control. Amplified products were separated using 3% agarose gel electrophoresis and subsequently visualized. The primer sequences for CNRIP1, SFRP2, and VIM (Table\u0026nbsp;1) were synthesized by a biotech company.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eMSP Sensitivity Assessment\u003c/h2\u003e \u003cp\u003eTo evaluate the sensitivity of the primers and PCR reaction conditions, a serial dilution assay was performed using Methylated human control DNA (bisulfite converted) and Unmethylated human control DNA (bisulfite converted) as standards representing 100% and 0% methylation, respectively. Mixed standards resulted in final methylation proportions of 100%, 99%, 90%, 50%, 10%, 1%, and 0%. Amplification conditions and subsequent gel electrophoresis were conducted as previously described (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB-D).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003eCriteria for Methylation Detection\u003c/h2\u003e \u003cp\u003eCriteria for determining gene methylation in samples were the appearance of only a methylated band or both methylated and unmethylated bands concurrently. A sample was considered non-methylated if only an unmethylated band appeared. For combined detection results, a positive result was defined by the presence of methylation in at least one gene.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eImmunochemical fecal occult blood test (FOBT)\u003c/h2\u003e \u003cp\u003eAn immunochemical FOBT was performed using the Fecal Occult Blood Gold Gel Stripe (W.H.P.M., Beijing, China) in accordance with the manufacturer's instructions. Briefly, feces samples were taken from multiple spots. A homogenized suspension was tested using the testing strips. Samples in which the control line and reaction line appeared simultaneously and clearly in the test strip within 5 min were designated as positive. Samples in which only a control line appeared in the test strip within 5 min were considered negative. Any other result was an invalidation and the sample testing was repeated.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eMachine learning (ML) algorithm\u003c/h2\u003e \u003cp\u003eTo evaluate the model's performance accurately, we split the dataset into training and test sets. The training set was used to train the model, while the test set remained untouched until the final evaluation. A standard split ratio of 4:1 was applied to ensure an unbiased assessment.\u003c/p\u003e \u003cp\u003eLightGBM, a gradient boosting framework that uses tree-based learning algorithms, was chosen as the primary model for this prediction task \u003csup\u003e[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/sup\u003e. LightGBM is known for its high training speed and low memory footprint. It utilizes a histogram-based algorithm, accelerating the model training process. LightGBM employs a tree-based model, iteratively generating weak classifiers and boosting the model's accuracy through gradient boosting \u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. Fine-tuning the hyperparameters of the LightGBM model was a critical step to achieve optimal performance. We utilized techniques such as grid search and random search to explore the hyperparameter space efficiently. The better hyperparameters are finally selected to build our model.\u003c/p\u003e \u003cp\u003eThe final step involved evaluating the model on the previously untouched test set. The Area under the ROC curve (AUC) as the primary measure of modeling. Additionally, performance metrics such as accuracy, sensitivity, specificity, PPV, NPV and F1-score were calculated to assess the model's effectiveness.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were conducted using SPSS 21.0 (IBM). Continuous variables with a normal distribution were expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), and intergroup age differences were assessed using independent samples t-tests. Categorical variables were presented as frequencies (percentages) and were assessed using the chi-squared (χ\u003csup\u003e2\u003c/sup\u003e) test; Fisher's exact test was employed when the case number was below 40. The accuracy of the diagnostic tests was assessed using the Receiver Operating Characteristic Curve (ROC), with the AUC and its 95% confidence interval (CI) calculated. A \u003cem\u003eP\u003c/em\u003e-value threshold of \u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eA total of 100 stool samples were collected, comprising 50 CRC cases and 50 healthy controls, to validate the sensitivity and specificity of fecal gene methylation tests, compared to FOBT.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePatient Demographics\u003c/strong\u003e \u003cstrong\u003eand Tumor Characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study analyzed 100 samples from residents of the Gannan region. The mean age was 62.1 years (range: 40-81 years) in the CRC group and 60.3 years (range: 45-80 years) in the control group, with no significant age difference between the groups (\u003cem\u003eP\u003c/em\u003e= 0.338). Gender distribution was also similar, with 68.0% males in the CRC group and 64.0% in the control group (\u003cem\u003eP\u003c/em\u003e= 0.673, Table 2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTumor location varied: 18.0% in the ascending colon, 2.0% in the transverse colon, 14.0% each in the descending and sigmoid colon, and 52.0% in the rectum. AJCC staging indicated 44.0% in stages I-II and 56.0% in stages III-IV. Tumor invasion was categorized as T1-T2 in 30.0% and T3-T4 in 70.0% of cases; lymph node metastasis was N0-N1 in 80.0% and N2-N3 in 20.0%; distant metastasis was M0 in 84.0% and M1 in 16.0%. Tumor differentiation was predominantly poorly differentiated (48.0%), followed by moderately (42.0%) and highly differentiated (10%).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAssessment of Stool DNA Methylation and FOBT in CRC Screening\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe DNA methylation tests were performed on all 100 stool samples to validate the sensitivity of independent gene methylation tests. The positive detection rates of SFRP2, CNRIP1, and VIM gene promoters and FOBT in the CRC group were 88.0%, 84.0%, 78.0%, and 50.0%, respectively, significantly higher than in the control group (14.0%, 18.0%, 18.0%, and 16.0%, respectively; \u003cem\u003eP\u003c/em\u003e\u0026lt; 0.001 for all comparisons, Figure 3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the evaluation of stool DNA methylation for CRC screening, the ROC curve analysis revealed varied diagnostic efficacies (Figure4 and Table 3). The AUC for SFRP2 methylation was 0.87 (95% CI: 0.794-0.946), indicating a high predictive accuracy, with a sensitivity (true positive rate, TPR) of 0.88. Methylation of CNRIP1 also demonstrated substantial screening effectiveness, with an AUC of 0.83 (95% CI: 0.745-0.915) and a sensitivity of 0.84. VIM methylation showed a slightly lower, yet notable, diagnostic potential with an AUC of 0.80 (95% CI: 0.709-0.891) and a sensitivity of 0.78. In comparison, traditional FOBT presented a significantly lower predictive value, evidenced by its AUC of 0.67 (95% CI: 0.563-0.777) and a sensitivity of only 0.50.\u003c/p\u003e\n\u003cp\u003eThese findings underscore the superior sensitivity of fecal DNA methylation markers, specifically SFRP2, CNRIP1, and VIM, in the screening for CRC compared to traditional FOBT.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnhanced CRC Screening Performance by LightGBM-Based ML-Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe ROC curve analysis for the ML model, constructed using the LightGBM algorithm and based on the methylation of the three genes (CNRIP1, SFRP2, and VIM) combined with FOBT results, demonstrated a significant enhancement in the non-invasive screening efficacy for CRC (Figure4).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe ML-model achieved a notably high AUC of 0.95 (95% CI: 0.916-0.991), indicating its superior predictive accuracy in identifying CRC cases. However, the TPR was recorded at 0.78 (Table 3), suggesting that while the model has a high overall accuracy, there is still room for improvement in its ability to correctly identify all positive CRC cases. This substantial improvement in the AUC value, as compared to the individual biomarkers, highlights the potential of integrating advanced ML techniques.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study marks a significant advancement in non-invasive CRC screening by combining stool DNA methylation markers (CNRIP1, SFRP2, and VIM) with the LightGBM ML algorithm. Our findings highlight the superior diagnostic accuracy of these methylation markers, particularly SFRP2, over traditional FOBT. The LightGBM-based model further enhanced this accuracy, demonstrating a high AUC in the ROC analysis.\u003c/p\u003e \u003cp\u003eOur study represents a pioneering exploration in the realm of CRC screening by synergistically integrating MT-sDNA testing with advanced ML techniques, specifically the LightGBM. This novel approach marks a significant leap in the field, blending the sensitivity of molecular biomarkers with the analytical prowess of artificial intelligence (AI). Historically, the utilization of MT-sDNA in CRC screening has been limited by its reliance on individual biomarker analysis \u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. However, our study demonstrates that with the continual advancement in ML algorithms, there is a remarkable potential to cohesively amalgamate these disparate methylation signals into a more coherent and robust diagnostic tool. By harnessing the computational power of LightGBM \u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e, our study successfully navigated the intricate landscape of methylation patterns, offering a more nuanced and comprehensive approach to CRC screening.\u003c/p\u003e \u003cp\u003eThe LightGBM algorithm represents a sophisticated form of AI that aids physicians in processing data and making informed decisions by uncovering hidden interactions within malignant tumor datasets \u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]\u003c/sup\u003e. Compared to traditional statistical techniques, this algorithm excels in handling the diversity of non-linear, high-dimensional, and complex data \u003csup\u003e[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. Utilizing machine learning algorithms such as LightGBM for predicting disease risk has emerged as a major area of research in the field of medical big data.\u003c/p\u003e \u003cp\u003eIn our research, the integration of the LightGBM ML algorithm with MT-sDNA testing has led to a notable improvement in the diagnostic efficiency for CRC screening. This combination has been particularly effective in enhancing the specificity and sensitivity of CRC detection. Our study showed that the application of LightGBM increased the AUC to 0.95, a significant improvement compared to the AUCs observed with traditional MT-sDNA tests alone. This higher AUC indicates a superior balance between sensitivity and specificity, crucial for reducing both false positives and false negatives in CRC screening. These results underscore the transformative potential of ML in refining the diagnostic capabilities of MT-sDNA tests. By accurately distinguishing between benign and malignant colorectal conditions \u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e, LightGBM has the potential to decrease unnecessary diagnostic colonoscopies \u003csup\u003e[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]\u003c/sup\u003e, thereby reducing the burden on both patients and healthcare systems \u003csup\u003e[\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]\u003c/sup\u003e. The application of LightGBM to other MT-sDNA tests like Cologuard \u003csup\u003e[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]\u003c/sup\u003e could revolutionize CRC screening, offering more accurate, non-invasive options for early cancer detection \u003csup\u003e[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. This innovative approach exemplifies the integration of AI in enhancing the precision of medical diagnostics.\u003c/p\u003e \u003cp\u003eConsidering the screening frequency and its implications on cost-effectiveness, patient compliance, and diagnostic sensitivity, the combined MT-sDNA and LightGBM approach could serve as an effective interim screening method, especially in identifying cases that might develop during the long interval between colonoscopies \u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e. This is particularly pertinent considering the progressive nature of CRC, where the development of malignancies could occur within the decade-long interval of traditional endoscopic screenings \u003csup\u003e[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]\u003c/sup\u003e. The higher sensitivity and specificity of this combined method positions it as a reliable, non-invasive test that could be administered more frequently than colonoscopy, without subjecting patients to the associated risks and discomforts of invasive procedures. Future research may explore the integration of such methylation-based screening into routine healthcare practices \u003csup\u003e[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]\u003c/sup\u003e, evaluating its cost-effectiveness and operational feasibility on a larger scale.\u003c/p\u003e \u003cp\u003eDespite the promising findings of our study, it is imperative to acknowledge its limitations and suggest avenues for future research. One notable limitation is the relatively small sample size and the study's confinement to a single center, which may limit the generalizability of our results. Future studies should aim to include a larger, more diverse population across multiple centers to validate our findings and ensure their applicability to broader demographic groups. Additionally, while our study primarily focused on CNRIP1, SFRP2, and VIM gene promoters, exploring additional methylation markers could further enhance the diagnostic accuracy and reliability of the CRC screening process. Another aspect warranting further investigation is the refinement and optimization of the LightGBM-based ML model. While our model demonstrated high accuracy, the true positive rate suggests room for improvement in its predictive capabilities, particularly in diverse clinical settings. Future research could explore the integration of more complex algorithms or the inclusion of additional clinical parameters to improve the model's sensitivity and specificity. Addressing these limitations will not only solidify the foundation laid by this study but also expand the horizons of non-invasive CRC diagnostics, potentially leading to more personalized and effective screening strategies.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study illustrates the significant potential of integrating stool DNA methylation markers CNRIP1, SFRP2, and VIM with the ML algorithm - LightGBM in enhancing the accuracy of non-invasive CRC screening. Our findings demonstrate that this innovative approach surpasses traditional screening methods in diagnostic efficacy, offering a promising direction for early and more personalized CRC detection in clinical practice.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eCRC: Colorectal Cancer; CNRIP1: Cannabinoid Receptor Interacting Protein 1; SFRP2: Secreted Frizzled-Related Protein 2; VIM: Vimentin; qPCR: Quantitative Polymerase Chain Reaction; MSP: Methylation-Specific PCR; FOBT: Fecal Occult Blood Testing; MT-sDNA: Multi-Target Stool DNA; ML: Machine Learning; AI: Artificial Intelligence; AJCC: American Joint Committee on Cancer; TNM: Tumor, Node, Metastasis; OD: Optical Density; TPR: True Positive Rate; TNR: True Negative Rate; FPR: False Positive Rate; PPV, Positive Predictive Value; NPV: Negative Predictive Value; FIT: Fecal Immunochemical Test; AUC: Area Under the ROC Curve; ROC: Receiver Operating Characteristic Curve; CI: Confidence Interval; SD: Standard Deviation.\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYi Xiang: Performed statistical analyses; Na Yang: Took the lead in writing the manuscript; Yunlong Zhu: Managed the collection and arrangement of the data; Gangfeng Zhu and Liangjian Zheng: Reviewed and made significant revisions to the initial draft; Shi Geng and Xiaofei Feng: Engaged in the clinical practices associated with this study; Zenghong Lu and Rui Zhu: Provided technical support throughout the research process; Xueming Xu and Xiangcai Wang: Were responsible for primary data collection; Tianlei Zheng and Li Huang: Oversaw the overall direction and planning of the project, and designed the research topic. All authors read and approved the final manuscript.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have declared that no competing interest exists.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data included in the study are available from the corresponding author upon reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge all the clinical and research staff from the research centers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinancial support\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Opening Project of Jiangsu Key Laboratory of Xuzhou Medical University (XZSYSKF2021030). Project supported by the Affiliated Hospital of Xuzhou Medical University (2022ZL26).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was approved by the Ethics Review Board of The First Affiliated Hospital of Gannan Medical University (Ganzhou, China). Written informed consent was obtained from all subjects before the study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eKocarnik JM, Compton K, Dean FE, et al. Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life Years for 29 Cancer Groups From 2010 to 2019: A Systematic Analysis for the Global Burden of Disease Study 2019. \u003cem\u003eJAMA Oncol\u003c/em\u003e 2022; \u003cstrong\u003e8\u003c/strong\u003e(3): 420-44.\u003c/li\u003e\n\u003cli\u003eGoss PE, Strasser-Weippl K, Lee-Bychkovsky BL, et al. Challenges to effective cancer control in China, India, and Russia. \u003cem\u003eLancet Oncol\u003c/em\u003e 2014; \u003cstrong\u003e15\u003c/strong\u003e(5): 489-538.\u003c/li\u003e\n\u003cli\u003eMorgan E, Arnold M, Gini A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN. \u003cem\u003eGut\u003c/em\u003e 2023; \u003cstrong\u003e72\u003c/strong\u003e(2): 338-44.\u003c/li\u003e\n\u003cli\u003eIovanescu D, Frandes M, Lungeanu D, Burlea A, Miutescu BP, Miutescu E. Diagnosis reliability of combined flexible sigmoidoscopy and fecal-immunochemical test in colorectal neoplasia screening. \u003cem\u003eOnco Targets Ther\u003c/em\u003e 2016; \u003cstrong\u003e9\u003c/strong\u003e: 6819-28.\u003c/li\u003e\n\u003cli\u003eYu J, Feng Q, Wong SH, et al. Metagenomic analysis of faecal microbiome as a tool towards targeted non-invasive biomarkers for colorectal cancer. \u003cem\u003eGut\u003c/em\u003e 2017; \u003cstrong\u003e66\u003c/strong\u003e(1): 70-8.\u003c/li\u003e\n\u003cli\u003eMarcuello M, Vymetalkova V, Neves RPL, et al. Circulating biomarkers for early detection and clinical management of colorectal cancer. \u003cem\u003eMol Aspects Med\u003c/em\u003e 2019; \u003cstrong\u003e69\u003c/strong\u003e: 107-22.\u003c/li\u003e\n\u003cli\u003eBach S, Paulis I, Sluiter NR, et al. Detection of colorectal cancer in urine using DNA methylation analysis. \u003cem\u003eSci Rep\u003c/em\u003e 2021; \u003cstrong\u003e11\u003c/strong\u003e(1): 2363.\u003c/li\u003e\n\u003cli\u003eLiang JQ, Li T, Nakatsu G, et al. A novel faecal Lachnoclostridium marker for the non-invasive diagnosis of colorectal adenoma and cancer. \u003cem\u003eGut\u003c/em\u003e 2020; \u003cstrong\u003e69\u003c/strong\u003e(7): 1248-57.\u003c/li\u003e\n\u003cli\u003eEbner DW, Eckmann JD, Burger KN, et al. Detection of Postcolonoscopy Colorectal Neoplasia by Multi-target Stool DNA. \u003cem\u003eClin Transl Gastroenterol\u003c/em\u003e 2021; \u003cstrong\u003e12\u003c/strong\u003e(6): e00375.\u003c/li\u003e\n\u003cli\u003eBooth WT, Walker NB, Lowther WT, Howlett AC. Cannabinoid Receptor Interacting Protein 1a (CRIP1a): Function and Structure. \u003cem\u003eMolecules\u003c/em\u003e 2019; \u003cstrong\u003e24\u003c/strong\u003e(20).\u003c/li\u003e\n\u003cli\u003eLind GE, Danielsen SA, Ahlquist T, et al. Identification of an epigenetic biomarker panel with high sensitivity and specificity for colorectal cancer and adenomas. \u003cem\u003eMol Cancer\u003c/em\u003e 2011; \u003cstrong\u003e10\u003c/strong\u003e: 85.\u003c/li\u003e\n\u003cli\u003eYang Q, Huang T, Ye G, Wang B, Zhang X. Methylation of SFRP2 gene as a promising noninvasive biomarker using feces in colorectal cancer diagnosis: a systematic meta-analysis. \u003cem\u003eSci Rep\u003c/em\u003e 2016; \u003cstrong\u003e6\u003c/strong\u003e: 33339.\u003c/li\u003e\n\u003cli\u003eXiao Z, Li B, Wang G, et al. Validation of methylation-sensitive high-resolution melting (MS-HRM) for the detection of stool DNA methylation in colorectal neoplasms. \u003cem\u003eClin Chim Acta\u003c/em\u003e 2014; \u003cstrong\u003e431\u003c/strong\u003e: 154-63.\u003c/li\u003e\n\u003cli\u003eNiknami Z, Eslamifar A, Emamirazavi A, Ebrahimi A, Shirkoohi R. The association of vimentin and fibronectin gene expression with epithelial-mesenchymal transition and tumor malignancy in colorectal carcinoma. \u003cem\u003eExcli j\u003c/em\u003e 2017; \u003cstrong\u003e16\u003c/strong\u003e: 1009-17.\u003c/li\u003e\n\u003cli\u003eThagaard J, Broeckx G, Page DB, et al. Pitfalls in machine learning-based assessment of tumor-infiltrating lymphocytes in breast cancer: A report of the International Immuno-Oncology Biomarker Working Group on Breast Cancer. \u003cem\u003eJ Pathol\u003c/em\u003e 2023; \u003cstrong\u003e260\u003c/strong\u003e(5): 498-513.\u003c/li\u003e\n\u003cli\u003ePrimakov SP, Ibrahim A, van Timmeren JE, et al. Automated detection and segmentation of non-small cell lung cancer computed tomography images. \u003cem\u003eNat Commun\u003c/em\u003e 2022; \u003cstrong\u003e13\u003c/strong\u003e(1): 3423.\u003c/li\u003e\n\u003cli\u003eLiang N, Li B, Jia Z, et al. Ultrasensitive detection of circulating tumour DNA via deep methylation sequencing aided by machine learning. \u003cem\u003eNat Biomed Eng\u003c/em\u003e 2021; \u003cstrong\u003e5\u003c/strong\u003e(6): 586-99.\u003c/li\u003e\n\u003cli\u003eHu X, Yin S, Zhang X, et al. Blood pressure stratification using photoplethysmography and light gradient boosting machine. \u003cem\u003eFront Physiol\u003c/em\u003e 2023; \u003cstrong\u003e14\u003c/strong\u003e: 1072273.\u003c/li\u003e\n\u003cli\u003eRufo DD, Debelee TG, Ibenthal A, Negera WG. Diagnosis of Diabetes Mellitus Using Gradient Boosting Machine (LightGBM). \u003cem\u003eDiagnostics (Basel)\u003c/em\u003e 2021; \u003cstrong\u003e11\u003c/strong\u003e(9).\u003c/li\u003e\n\u003cli\u003eZeng X. Length of Stay Prediction Model of Indoor Patients Based on Light Gradient Boosting Machine. \u003cem\u003eComput Intell Neurosci\u003c/em\u003e 2022; \u003cstrong\u003e2022\u003c/strong\u003e: 9517029.\u003c/li\u003e\n\u003cli\u003eYang F, Sun J, Luo H, et al. Assessment of fecal DNA extraction protocols for metagenomic studies. \u003cem\u003eGigascience\u003c/em\u003e 2020; \u003cstrong\u003e9\u003c/strong\u003e(7).\u003c/li\u003e\n\u003cli\u003eSeto H, Oyama A, Kitora S, et al. Gradient boosting decision tree becomes more reliable than logistic regression in predicting probability for diabetes with big data. \u003cem\u003eSci Rep\u003c/em\u003e 2022; \u003cstrong\u003e12\u003c/strong\u003e(1): 15889.\u003c/li\u003e\n\u003cli\u003eCarethers JM. Fecal DNA Testing for Colorectal Cancer Screening. \u003cem\u003eAnnu Rev Med\u003c/em\u003e 2020; \u003cstrong\u003e71\u003c/strong\u003e: 59-69.\u003c/li\u003e\n\u003cli\u003eRedwood DG, Dinh TA, Kisiel JB, et al. Cost-Effectiveness of Multitarget Stool DNA Testing vs Colonoscopy or Fecal Immunochemical Testing for Colorectal Cancer Screening in Alaska Native People. \u003cem\u003eMayo Clin Proc\u003c/em\u003e 2021; \u003cstrong\u003e96\u003c/strong\u003e(5): 1203-17.\u003c/li\u003e\n\u003cli\u003ePark DJ, Park MW, Lee H, Kim YJ, Kim Y, Park YH. Development of machine learning model for diagnostic disease prediction based on laboratory tests. \u003cem\u003eSci Rep\u003c/em\u003e 2021; \u003cstrong\u003e11\u003c/strong\u003e(1): 7567.\u003c/li\u003e\n\u003cli\u003eZhang Y, Jiang Z, Chen C, Wei Q, Gu H, Yu B. DeepStack-DTIs: Predicting Drug-Target Interactions Using LightGBM Feature Selection and Deep-Stacked Ensemble Classifier. \u003cem\u003eInterdiscip Sci\u003c/em\u003e 2022; \u003cstrong\u003e14\u003c/strong\u003e(2): 311-30.\u003c/li\u003e\n\u003cli\u003ePark YM, Lee BJ. Machine learning-based prediction model using clinico-pathologic factors for papillary thyroid carcinoma recurrence. \u003cem\u003eSci Rep\u003c/em\u003e 2021; \u003cstrong\u003e11\u003c/strong\u003e(1): 4948.\u003c/li\u003e\n\u003cli\u003eRen J, Zhou X, Guo W, Feng K, Huang T, Cai YD. Identification of Methylation Signatures and Rules for Sarcoma Subtypes by Machine Learning Methods. \u003cem\u003eBiomed Res Int\u003c/em\u003e 2022; \u003cstrong\u003e2022\u003c/strong\u003e: 5297235.\u003c/li\u003e\n\u003cli\u003eGardner W, Cutts SM, Phillips DR, Pigram PJ. Understanding mass spectrometry images: complexity to clarity with machine learning. \u003cem\u003eBiopolymers\u003c/em\u003e 2021; \u003cstrong\u003e112\u003c/strong\u003e(4): e23400.\u003c/li\u003e\n\u003cli\u003eLiu L, Meng Q, Weng C, Lu Q, Wang T, Wen Y. Explainable deep transfer learning model for disease risk prediction using high-dimensional genomic data. \u003cem\u003ePLoS Comput Biol\u003c/em\u003e 2022; \u003cstrong\u003e18\u003c/strong\u003e(7): e1010328.\u003c/li\u003e\n\u003cli\u003eGrosu S, Wesp P, Graser A, et al. Machine Learning-based Differentiation of Benign and Premalignant Colorectal Polyps Detected with CT Colonography in an Asymptomatic Screening Population: A Proof-of-Concept Study. \u003cem\u003eRadiology\u003c/em\u003e 2021; \u003cstrong\u003e299\u003c/strong\u003e(2): 326-35.\u003c/li\u003e\n\u003cli\u003eVahdat V, Alagoz O, Chen JV, Saoud L, Borah BJ, Limburg PJ. Calibration and Validation of the Colorectal Cancer and Adenoma Incidence and Mortality (CRC-AIM) Microsimulation Model Using Deep Neural Networks. \u003cem\u003eMed Decis Making\u003c/em\u003e 2023; \u003cstrong\u003e43\u003c/strong\u003e(6): 719-36.\u003c/li\u003e\n\u003cli\u003eNemlander E, Ewing M, Abedi E, et al. A machine learning tool for identifying non-metastatic colorectal cancer in primary care. \u003cem\u003eEur J Cancer\u003c/em\u003e 2023; \u003cstrong\u003e182\u003c/strong\u003e: 100-6.\u003c/li\u003e\n\u003cli\u003eA stool DNA test (Cologuard) for colorectal cancer screening. \u003cem\u003eJama\u003c/em\u003e 2014; \u003cstrong\u003e312\u003c/strong\u003e(23): 2566.\u003c/li\u003e\n\u003cli\u003eZhao F, Bai P, Xu J, et al. Efficacy of cell-free DNA methylation-based blood test for colorectal cancer screening in high-risk population: a prospective cohort study. \u003cem\u003eMol Cancer\u003c/em\u003e 2023; \u003cstrong\u003e22\u003c/strong\u003e(1): 157.\u003c/li\u003e\n\u003cli\u003ePickhardt PJ, Graffy PM, Weigman B, Deiss-Yehiely N, Hassan C, Weiss JM. Diagnostic Performance of Multitarget Stool DNA and CT Colonography for Noninvasive Colorectal Cancer Screening. \u003cem\u003eRadiology\u003c/em\u003e 2020; \u003cstrong\u003e297\u003c/strong\u003e(1): 120-9.\u003c/li\u003e\n\u003cli\u003eNakao SK, Fassler S, Sucandy I, Kim S, Zebley DM. Colorectal cancer following negative colonoscopy: is 5-year screening the correct interval to recommend? \u003cem\u003eSurg Endosc\u003c/em\u003e 2013; \u003cstrong\u003e27\u003c/strong\u003e(3): 768-73.\u003c/li\u003e\n\u003cli\u003eMorgacheva D, Ryzhova M, Zheludkova O, Belogurova M, Dinikina Y. DNA methylation-based diagnosis confirmation in a pediatric patient with low-grade glioma: a case report. \u003cem\u003eFront Pediatr\u003c/em\u003e 2023; \u003cstrong\u003e11\u003c/strong\u003e: 1256876.\u003c/li\u003e\n\u003cli\u003ePickles JC, Fairchild AR, Stone TJ, et al. DNA methylation-based profiling for paediatric CNS tumour diagnosis and treatment: a population-based study. \u003cem\u003eLancet Child Adolesc Health\u003c/em\u003e 2020; \u003cstrong\u003e4\u003c/strong\u003e(2): 121-30.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"717\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"80.8926080892608%\" colspan=\"4\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable 1 Primer sequences and PCR product size obtained used for MSP assays\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.107391910739192%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003e \u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.715481171548117%\"\u003e\n \u003cp\u003eGene\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.320781032078104%\"\u003e\n \u003cp\u003ePrimer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"4.463040446304045%\"\u003e\n \u003cp\u003e \u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"54.39330543933055%\"\u003e\n \u003cp\u003eSequence (5\u0026rsquo;-3\u0026rsquo;)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.107391910739192%\"\u003e\n \u003cp\u003ePCR\u003cbr\u003e\u0026nbsp;product size (bp)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.715481171548117%\" rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003eCNRIP1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.320781032078104%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"4.463040446304045%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"54.39330543933055%\"\u003e\n \u003cp\u003e5\u0026prime;-GTGGAATTGCGTTTTTTTTC-3\u0026prime;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.107391910739192%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e163\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-TACTACTCGCGAATCCGAC-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.690363349131122%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.055292259083728%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"61.611374407582936%\"\u003e\n \u003cp\u003e5\u0026apos;-GTGGAATTGTGTTTTTTTTT-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.64296998420221%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e163\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-TACTACTCACAAATCCAAC-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.715481171548117%\" rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003eSFRP2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.320781032078104%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"4.463040446304045%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"54.39330543933055%\"\u003e\n \u003cp\u003e5\u0026prime;-GAGTAGTAGTGTTTATTTTA-3\u0026prime;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.107391910739192%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e138\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-CCGCTCTCTTCGCTAAATACGACTCG-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.690363349131122%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.055292259083728%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"61.611374407582936%\"\u003e\n \u003cp\u003e5\u0026apos;-TTTTGGGTTGGAGTTTTTTGGAGTTGTGT-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.64296998420221%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e145\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-AACCCACTCTCTTCACTAAATACAACTCA-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.715481171548117%\" rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003eVIM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"10.320781032078104%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"4.463040446304045%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"54.39330543933055%\"\u003e\n \u003cp\u003e5\u0026prime;-TCGTTTCGAGGTTTTCGCGTTAGAGAC-3\u0026prime;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"19.107391910739192%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e216\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-CGACTAAAACTCGACCGACTCGCGA-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"11.690363349131122%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003eU\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"5.055292259083728%\"\u003e\n \u003cp\u003eF:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"61.611374407582936%\"\u003e\n \u003cp\u003e5\u0026prime;-TTGGTGGATTTTTTGTTGGTTGATG-3\u0026prime;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"21.64296998420221%\" rowspan=\"2\" valign=\"top\"\u003e\n \u003cp\u003e183\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"7.5829383886255926%\"\u003e\n \u003cp\u003eR:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"92.41706161137441%\"\u003e\n \u003cp\u003e5\u0026apos;-CACAACTTACCTTAACCCTTAAACTACTCA-3\u0026apos;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"100%\" colspan=\"5\" valign=\"top\"\u003e\n \u003cp\u003ePCR, polymerase chain reaction; MSP, methylation‑specific; bp, base pair; SFRP2, secreted frizzled‑related protein 2; VIM, vimentin; M, methylated; U, unmethylated; F, forward; R, reverse.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" style=\"width: 41.7324%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable 2 Baseline characteristics and DNA methylation status of CRC patients and healthy controls\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eCharacteristics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003eCRC patients\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003eHealthy controls\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u003cem\u003e\u0026chi;\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e/\u003cem\u003et\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e-value\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;(\u003cem\u003en\u003c/em\u003e=50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e(\u003cem\u003en\u003c/em\u003e=50)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eAge (years,\u0026nbsp;\u0026plusmn;s)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e62.1\u0026plusmn;9.59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e60.3\u0026plusmn;9.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e0.963\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e0.338\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eGender [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e0.178\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e0.673\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e68.0% (34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e64.0% (32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eTumor location [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003cp\u003eAscending colon\u003c/p\u003e\n \u003cp\u003eTransverse colon\u003c/p\u003e\n \u003cp\u003eDescending colon\u003c/p\u003e\n \u003cp\u003eSigmoid colon\u003c/p\u003e\n \u003cp\u003eRectum\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eTNM classification [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003col\u003e\n \u003cli\u003eII\u003c/li\u003e\n \u003c/ol\u003e\n \u003cp\u003eIII-IV\u003c/p\u003e\n \u003cp\u003eTumor invasion [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003cp\u003eT1-T2\u003c/p\u003e\n \u003cp\u003eT3-T4\u003c/p\u003e\n \u003cp\u003eLymph node metastasis [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; N0-N1\u003c/p\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; N2-N3\u003c/p\u003e\n \u003cp\u003eDistant metastasis [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003cp\u003eM0\u003c/p\u003e\n \u003cp\u003eM1\u003c/p\u003e\n \u003cp\u003eTumor differentiation [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003cp\u003ePoor\u003c/p\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003cp\u003eWell\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e18.0% (9)\u003c/p\u003e\n \u003cp\u003e2.0% (1)\u003c/p\u003e\n \u003cp\u003e14.0% (7)\u003c/p\u003e\n \u003cp\u003e14.0% (7)\u003c/p\u003e\n \u003cp\u003e52.0% (26)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e44.0% (22)\u003c/p\u003e\n \u003cp\u003e56.0% (28)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e30.0% (15)\u003c/p\u003e\n \u003cp\u003e70.0% (35)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e80.0% (40)\u003c/p\u003e\n \u003cp\u003e20.0% (10)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e84.0% (42)\u003c/p\u003e\n \u003cp\u003e16.0% (8)\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e48.0% (24)\u003c/p\u003e\n \u003cp\u003e42.0% (21)\u003c/p\u003e\n \u003cp\u003e10% (5)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eMethylation of SFRP2 [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e54.782\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e88.0% (44)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e14.0% (7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eMethylation of CNRIP1 [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e43.577\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e84.0% (42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e18.0% (9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eMethylation of VIM [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e36.058\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e78.0% (39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e18.0% (9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 22.3515%;\"\u003e\n \u003cp\u003eFOBT detection [% (\u003cem\u003en\u003c/em\u003e)]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e13.071\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026lt;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 2.3265%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20.108%;\"\u003e\n \u003cp\u003ePositive\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 9.3893%;\"\u003e\n \u003cp\u003e50.0% (25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.6327%;\"\u003e\n \u003cp\u003e16.0% (8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 4.9024%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 5.484%;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"6\" style=\"width: 55.671%;\"\u003e\n \u003cp\u003eAbbreviations: CRC, Colorectal Cancer; FOBT, Fecal Occult Blood Testing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"689\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable 3 Performance of stool DNA methylation,\u003c/strong\u003e \u003cstrong\u003eFOBT and the ML-model for CRC screening\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVariables\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e95% CI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTPR (Sensitivity)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eTNR (Specificity)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eFPR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSFRP2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.794-0.946\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.863\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.878\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eCNRIP2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.745-0.915\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.824\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.837\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVIM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.709-0.891\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.813\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.788\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eFOBT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.67\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.563-0.777\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.758\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.627\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003elightGBM model\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.916-0.991 \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.875\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.833\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\"\u003e\n \u003cp\u003eAbbreviations: FOBT, Fecal Occult Blood Testing; CRC, Colorectal Cancer; TPR, True Positive Rate; TNR, True Negative Rate; PPV, Positive Predictive Value; NPV, Negative Predictive Value.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Colorectal Cancer, Non-invasive Screening, DNA Methylation, LightGBM, Machine Learning, Multi-Target Stool DNA","lastPublishedDoi":"10.21203/rs.3.rs-3857174/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3857174/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective: \u003c/strong\u003eThis study evaluates the effectiveness of stool DNA methylation markers CNRIP1, SFRP2, and VIM, along with Fecal Occult Blood Testing (FOBT), in the non-invasive screening of colorectal cancer (CRC), further integrating these markers with the Light Gradient Boosting Machine (LightGBM) machine learning (ML) algorithm.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eThe study analyzed 100 stool samples, comprising 50 CRC patients and 50 healthy controls, from the First Affiliated Hospital of Gannan Medical University. Methylation Specific PCR (MSP) was used for assessing the methylation status of CNRIP1, SFRP2, and VIM gene promoters. FOBT was performed in parallel. Diagnostic performance was assessed using Receiver Operating Characteristic (ROC) curve analysis, and a LightGBM-based ML model was developed, incorporating these methylation markers and FOBT results.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eROC analysis demonstrated that SFRP2 had the highest diagnostic accuracy with an AUC of 0.87 (95% CI: 0.794-0.946) and a sensitivity of 0.88. CNRIP1 and VIM also showed substantial screening effectiveness, with AUCs of 0.83 and 0.80, respectively. FOBT, in comparison, had a lower predictive value with an AUC of 0.67. The LightGBM-based ML model significantly outperformed individual markers, achieving a high AUC of 0.95 (95% CI: 0.916-0.991). However, the sensitivity of the ML model was 0.78, suggesting a need for improvement in correctly identifying all positive CRC cases.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion: \u003c/strong\u003eStool DNA methylation markers CNRIP1, SFRP2, and VIM exhibit high sensitivity in non-invasive CRC screening. The integration of these biomarkers with the LightGBM ML algorithm enhances the diagnostic accuracy, offering a promising approach for early CRC detection.\u003c/p\u003e","manuscriptTitle":"Enhancing Non-Invasive Colorectal Cancer Screening with Stool DNA Methylation Markers and LightGBM Machine Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-16 20:38:29","doi":"10.21203/rs.3.rs-3857174/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"03f6dfbb-f47b-4016-b55c-8bac408f6647","owner":[],"postedDate":"January 16th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-01-17T21:15:27+00:00","versionOfRecord":[],"versionCreatedAt":"2024-01-16 20:38:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3857174","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3857174","identity":"rs-3857174","version":["v1"]},"buildId":"_2-kVJe1T_tPrBINL-cwx","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}