A systematic review of artificial intelligence and machine learning for gut microbiome-based CRC screening

Over the last decade, ample evidence has shown that gut microbiome dysbiosis is significantly associated with colorectal cancer (CRC). More recently, studies have shown that artificial intelligence (AI) and machine learning (ML) models paired with gut microbiome data can detect CRC. The objectives for this systematic review are: (I) to systematically evaluate the diagnostic performance of AI/ML models using gut microbiome data for CRC detection; (II) to compare AI/ML-based microbiome screening approaches; (III) to identify microbial genera consistently associated with CRC across included studies; (IV) to assess study quality and risk of bias using QUADAS-2 and use the GRADE approach to assess certainty.

A systematic review was conducted across PubMed, MEDLINE, Scopus, Embase, and The Cochrane Library including studies from January 1, 2023 to November 1, 2025. Studies applying AI/ML models to human gut microbiome data for CRC screening and reporting diagnostic performance metrics and significant genera were included. Risk of bias was assessed with QUADAS-2 and used the GRADE approach to assess certainty. Primary outcomes included area under the receiver operating characteristic curve (AUC). Secondary outcomes include balanced accuracy, sensitivity, specificity, examine AI/ML approaches used, and identify significant microbial genera associated with CRC detected by AI/ML models. No meta-analysis was done due to the heterogeneity, which came from various microbiome methods, AI/ML/microbiome models used, and validation strategies applied. Thereby the data was synthesized narratively.

Twelve studies met inclusion criteria. AI/ML models demonstrated moderate diagnostic performance, with internal validation sets AUC values ranging from 0.61 to 0.98 and external validation sets AUC ranging from 0.70-0.87. Random forest and XGBoost models showed consistent performance, with multi-omics approaches achieving the highest AUC of 0.98. Porphyromonas (e.g., P. asaccharolytica), Peptostreptococcus (e.g., P. stomatis), Fusobacterium (particularly F. nucleatum subspecies: animalis, vincentii, polymorphum, sensu stricto), Parvimonas micra, Gemella morbillorum, Bacteroides (B. fragilis), and Streptococcus species were significantly enriched in CRC as predicted the selected AI/ML/microbiome models.

AI/ML-based gut microbiome models demonstrate moderate AUC for CRC detection and may enhance noninvasive screening strategies pending prospective validation. Limitations include heterogeneity in AI/ML/microbiome models, microbiome methodologies, and model validation, with predominantly retrospective case-control studies and limited external validation.