Related Experiment Video
Updated: Jul 1, 2025
08:14
A Method to Define the Effects of Environmental Enrichment on Colon Microbiome Biodiversity in a Mouse Colon Tumor Model
Published on: February 28, 2018
8.8K
Explainable artificial intelligence for microbiome data analysis in colorectal cancer biomarker identification
Pierfrancesco Novielli1,2, Donato Romano1,2, Michele Magarelli1
1Dipartimento di Scienze del Suolo, della Pianta e degli Alimenti, Università degli Studi di Bari Aldo Moro, Bari, Italy.
Frontiers in Microbiology
|March 1, 2024
Summary
Machine learning models can identify colorectal cancer (CRC) by analyzing gut microbiome data. Explainable AI and SHAP analysis pinpoint specific bacteria associated with CRC, aiding in early diagnosis and targeted interventions.
Area of Science:
- Microbiome research
- Artificial intelligence in medicine
- Cancer diagnostics
Background:
- Colorectal cancer (CRC) is linked to gut microbiome dysbiosis.
- New early diagnostic methods are crucial due to CRC's high mortality.
- Machine learning (ML) can analyze host-microbiota interactions.
Purpose of the Study:
- To develop an explainable AI framework for CRC classification using gut microbiota and demographic data.
- To identify specific microbial taxonomic markers associated with CRC.
- To enhance the interpretability of ML models in disease prediction.
Main Methods:
- Implemented an explainable artificial intelligence (XAI) framework.
- Utilized gut microbiota data and demographic information for classification.
- Employed the Shapley Method Additive Explanations (SHAP) algorithm for variable importance.
Main Results:
- Random Forest (RF) algorithm showed the best performance in classifying CRC subjects (precision 0.729 ± 0.038).
- SHAP analysis identified key bacterial contributors to CRC classification.
- Confirmed associations of specific bacteria like *Fusobacterium*, *Peptostreptococcus*, and *Parvimonas* with CRC.
Conclusions:
- Gut microbiota data, analyzed via explainable AI, shows promise for CRC classification.
- The RF algorithm is suitable for CRC detection based on microbial signatures.
- SHAP analysis provides interpretable insights into bacterial roles in CRC, paving the way for targeted interventions.
