Advancing Colorectal Cancer Diagnosis: Integrating Synthetic Data and Machine Learning for Microbiome Analysis

This work highlights the importance of the gut microbiota in colorectal health, especially during AP-to-CRC transition. Synthetic data augmentation enlarged and balanced a multidimensional OTU table for machine learning categorization. The OTU table refinement used SVM and LG for sample validation and several statistical tests to assure synthetic data realism. In addition, deep learning feature extraction with LRP identified 64 unique bacterial taxa that were assessed for their ability to distinguish AP and CRC samples in diverse datasets. Fusobacterium was important in LRP and SHAP studies, consistent with its connection with CRC. SHAP analysis using XGBoost discovered differentiated features like Parvimonas, Alistipes, and Ruminococcus in stool and biopsy datasets. Classifying the saliva dataset with 100% accuracy is noteworthy. © 2025 by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved.