A reliable method for colorectal cancer prediction based on feature selection and support vector machine

Area of Science:

• Biomedical Informatics
• Computational Biology
• Oncology

Background:

• Colorectal cancer (CRC) is a leading cause of cancer death globally, necessitating improved early detection methods.
• Accurate and timely prediction of CRC remains a significant clinical challenge.
• Identifying key risk factors is crucial for developing effective diagnostic strategies.

Purpose of the Study:

• To develop and validate an integrated computational model for classifying colorectal cancer.
• To identify significant independent risk factors associated with CRC development.
• To optimize a machine learning model for enhanced CRC prediction accuracy.

Main Methods:

• An integrated model combining logistic regression (LR) and support vector machine (SVM) was constructed.
• Logistic regression was employed to select significant features (p < 0.05) including gut bacteria (Firmicutes, Bacteroidetes), BMI, and age.
• A grid-search SVM model with various kernel types (Linear, RBF, Sigmoid, Polynomial) was optimized for classification.

Main Results:

• Logistic regression identified Firmicutes (AUC 0.918), Bacteroidetes (AUC 0.856), BMI (AUC 0.777), and age (AUC 0.710) as significant predictors.
• The combined factors achieved an AUC of 0.942 for CRC detection.
• The RBF kernel SVM model demonstrated the highest accuracy, reaching 91.2% with k=10.

Conclusions:

• The integrated LR-SVM model offers a novel and effective approach for colorectal cancer prediction.
• Specific gut microbiome compositions and host factors like BMI and age are critical indicators for CRC.
• The study highlights the potential of machine learning in improving early detection and risk assessment for colorectal cancer.