Identification of bacterial key genera associated with breast cancer using machine learning techniques

Abstract

Breast cancer (BC) is a leading cause of cancer-related morbidity and mortality among women, with both genetic and lifestyle factors significantly contributing to BC risk. Emerging evidence suggests that the gut microbiota, which influences immune function, metabolism, and inflammation, may also play a critical role in BC risk. This study provides a comprehensive analysis of fecal microbiota composition in BC patients compared to healthy controls and identifies bacterial key genera (bKG) using machine learning approaches. The results reveal significant differences in bacterial diversity and the abundance of specific genera. Alpha diversity indices (Observed, Chao1, and Good’s coverage estimator) showed reduced diversity in BC patients, while beta diversity profiles, assessed using Bray-Curtis distance and principal coordinate analysis (PCoA), indicated distinct microbial communities. Taxonomic analysis identified a higher prevalence of genera such as Bacteroides and Blautia in BC patients, and reduced abundance of beneficial genera like Prevotella, Roseburia, Succinivibrio, and Ruminococcus. Machine learning techniques were employed to identify bKG associated with BC, with the eXtreme Gradient Boosting (XGB) model demonstrating the highest predictive accuracy (76.0 %). SHapley Additive exPlanations (SHAP) analysis further revealed the impact of specific bacterial taxa, including Clostridium saccharogumia, Megamonas, and Eubacterium_dolichum, on BC risk. These findings highlight the potential of fecal microbiota as non-invasive biomarkers for BC detection and prognosis. Future research should focus on the functional roles of these bacterial alterations and their implications in BC pathogenesis, aiming to develop microbiota-based diagnostic and therapeutic strategies.

Keywords

• BC
• Fecal microbiota
• Diversity analysis
• Taxonomic analysis
• Machine learning
• SHAP analysis