Leveraging explainable AI for gut microbiome-based colorectal cancer classification

Area of Science:

• Microbiology
• Oncology
• Bioinformatics

Background:

• Colorectal cancer (CRC) is linked to gut microbiome alterations.
• Current machine learning methods identify general CRC biomarkers but miss individual variations.
• Personalized biomarker identification is crucial for effective CRC diagnosis and treatment.

Purpose of the Study:

• To investigate Shapley Additive Explanations (SHAP) for personalized colorectal cancer (CRC) biomarker discovery.
• To assess SHAP's ability to identify individual-specific bacterial biomarkers associated with CRC.
• To explore the potential of SHAP in stratifying CRC patients based on distinct microbiome profiles and CRC risk.

Main Methods:

• Application of Shapley Additive Explanations (SHAP) machine learning technique.
• Analysis of gut microbiome composition data from five independent colorectal cancer (CRC) datasets.
• Utilizing SHAP for inferring personalized bacterial biomarkers linked to CRC.

Main Results:

• SHAP successfully identified personalized bacterial biomarkers for colorectal cancer (CRC).
• The method demonstrated the ability to differentiate CRC subjects into distinct subgroups.
• Subgroups exhibited unique CRC probabilities and associated bacterial biomarker profiles.

Conclusions:

• Shapley Additive Explanations (SHAP) provides a powerful tool for personalized colorectal cancer (CRC) biomarker discovery.
• SHAP enhances understanding of the gut microbiome's role in CRC by revealing individual-specific bacterial influences.
• This personalized approach holds promise for improved CRC risk assessment and tailored therapeutic strategies.