The gut microbiota has been increasingly recognized as a promising non-invasive biomarker source for Colorectal Cancer (CRC) detection. In this study, we develop a multi-layer stacking ensemble learning framework that integrates multiple machine learning models to improve the classification accuracy of CRC based on gut microbiota profiles. Our framework is trained using publicly available microbiome datasets comprising 1129 samples from diverse geographical regions (Europe, America, and Asia), and independently evaluated on an external validation cohort collected from Peking Union Medical College Hospital (PUMCH) in China. Based on our experiments, feature extraction of gut microbiota at the genus and species levels is found to achieve the best performance. The framework integrates multiple base classifiers—including Light Gradient Boosting Machine (LightGBM), eXtreme Gradient Boosting (XGBoost), Random Forests (RF), and Support Vector Machines (SVM)—combined through a weighting layer to optimize final classifications. Analysis of the feature importance in the trained model reveals several microbial populations previously reported to be associated with CRC, such as Gemella morbillorum and Fusobacterium nucleatum. These findings support the microbiological interpretability of our proposed framework. Experimental results show that our ensemble model achieves an Area Under the Receiver Operating Characteristic Curve (namely ROC_AUC) of 77.04% when validated on a real-world clinical dataset from Peking Union Medical College Hospital, surpassing existing microbiome-based CRC classification approaches.
- Article type
- Year
- Co-author
Open Access
Issue
To evaluate the feasibility and effectiveness of colonoscopy, fecal immunochemical test(FIT) and risk-adapted screening strategies in population-based colorectal cancer screening.
Based on the randomized controlled trial of colorectal cancer screening(TARGET-C) which was carried out in 6 centers of 5 provinces in China from May 2018 to May 2021, the participants who met the inclusion criteria of the study were randomly assigned to one of the following screening intervention groups in a ratio of 1∶2∶2, which were colonoscopy group, FIT group and risk-adapted screening group(participants evaluated to be at high risk were recommended to undertake colonoscopy, those at low risk were recommended to undertake FIT and FIT positive people received colonoscopy). The 3 groups received different schemes of colorectal cancer screening, in which the colonoscopy group only received baseline screening, whereas the FIT group and the risk-adapted screening group received annual follow-up screening on the basis of baseline screening. The main outcome was the detection rate for advanced colorectal neoplasm(including colorectal cancer and advanced adenoma). The participation rate, detection rate for any neoplasm and resource load of colonoscopy among different screening groups were further analyzed.
A total of 19 373 participants meeting the inclusion and exclusion criteria were enrolled, including 8082 males(41.7%) and 11 291 females(58.3%), with an average age of (60.5±6.5) years. Among them, 3883 were in the colonoscopy group, 7793 in the FIT group, and 7697 in the risk-adapted screening group. After 1 or 3 rounds of screening(2 rounds of follow-up screening completed in FIT group and risk-adapted screening group), the overall participation was the highest for the FIT group(99.3%), followed by the risk-adapted screening group(89.2%) and the coloscopy group(42.3%). According to the intention-to-treat analysis, the detection rates of advanced neoplasm in the colonoscopy group was higher than that in the FIT group(2.76% vs. 2.17%, ORcolonoscopy vs FIT=1.30, 95% CI: 1.01-1.65, P=0.037); there were no statistically significant differences regarding the detection rates of advanced neoplasm between the colonoscopy group and the risk-adapted screening group(2.76% vs. 2.35%, ORcolonoscopy vs risk-adapted screening=1.19, 95% CI: 0.93-1.51, P=0.156), and between the risk-adapted screening group and the FIT group(2.35% vs. 2.17%, ORrisk-adapted screening vs FIT=1.09, 95% CI: 0.88-1.34, P=0.440). The number of colonoscopies needed to be performed to detect one advanced neoplasm was used as an indicator to evaluate the resource load of colonoscopy. The number was the highest for the coloscopy group(15.4), followed by the risk-adapted screening group(10.2) and the FIT group(7.8).
The risk-adapted screening strategy is feasible and effective in population-based colorectal cancer screening, and could serve as an effective supplement to the traditional colonoscopy and FIT-based colorectal cancer screening strategies.
京公网安备11010802044758号