AUTHOR=Murovec Boštjan , Deutsch Leon , Stres Blaž 

TITLE=Predictive modeling of colorectal cancer using exhaustive analysis of microbiome information layers available from public metagenomic data

JOURNAL=Frontiers in Microbiology

VOLUME=Volume 15 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2024.1426407

DOI=10.3389/fmicb.2024.1426407

ISSN=1664-302X

ABSTRACT=This study aimed to compare the microbiome profiles of patients with colorectal cancer (CRC, n=380) and colorectal adenomas (CRA, n=110) with those of generally healthy participants (n=2461) from diverse studies. The overarching objective was to conduct a real-life experiment and construct a robust machine learning model applicable to the general population. A total of 2951 stool samples underwent a comprehensive analysis utilizing the in-house MetaBakery pipeline, encompassing diverse data matrices, including microbial taxonomy, functional genes, enzymatic reactions, metabolic pathways, and predicted metabolites. The study revealed no statistically significant difference in microbial diversity among individuals, with distinct clusters identified for healthy, CRC, and CRA groups through linear discriminant analysis (LDA). Machine learning analysis demonstrated consistent model performance, showcasing the potential of microbial taxa, functional genes, enzymatic reactions, and metabolic pathways as prediagnostic indicators for CRC and CRA. Notable biomarkers on the taxonomy level and microbial functionality (gene families, enzymatic reactions and metabolic pathways) associated with CRC, were identified. The research presents promising avenues for practical clinical applications, with potential validation on external clinical datasets in future studies.