Identification of Key Microbial Signatures Associated With Breast Cancer: Machine Learning-Based Approaches Using Gut...

Breast cancer (BC) is one of the most common types of cancer incidence and mortality rates among women all over the world. Limitations in existing diagnostic methods, especially in low- and middle-income countries, require novel, noninvasive biomarkers. A complex and modifiable ecosystem, the gut microbiome has become a potential origin of such biomarkers because it has a systemic influence on host immunity, metabolism, and regulation of hormones. This study explores the gut microbiome alterations associated with BC and evaluates machine learning (ML) models to predict disease status based on 16S rRNA gene sequencing data of Ghanaian cohort (520 BC patients, 442 healthy controls).

Following high-quality sequence processing with QIIME2 and DADA2, alpha and beta diversity analyses showed a lower microbial richness and different community structures in BC patients (PERMANOVA, p = 0.001). Random Forest-based feature selection was used to find the top 20 discriminative genera, which were used to train 10 supervised ML algorithms. Ensemble models, particularly Random Forest, achieved the highest performance in predictive models (mean across 100 folds: AUC = 0.78, accuracy = 0.70), indicating strong discriminative potential. The results of SHAP analysis provided clear pictures of how specific microbes influence BC risk. We identified several risk-associated genera, notably Bacteroides, Lachnoclostridium, Parasutterella, and Tannerellaceae, whereas Coprococcus, Romboutsia, Ruminococcus, and Prevotella were associated with lower BC risk.

This study reveals a clear shift in the gut microbiome community structure among BC patients. We identified specific bacterial genera that act as either elevated or decreased disease risk. These findings not only highlight the power of machine learning models for noninvasive risk prediction but also provide a foundation for novel microbiome-based diagnostic strategies.