Microbial Signature for Colorectal Cancer Identified Using Machine Learning

Researchers from the European Molecular Biology Laboratory (EMBL) in Heidelberg, Germany, the University of Trento in Italy, and other international collaborators leveraged a machine learning algorithm to identify a subset of gut bacteria associated with colorectal cancer, the third most common cancer worldwide.

They performed a meta-analysis of eight studies of gut bacteria and colorectal cancer, spanning seven countries and three continents, and found that the microbiome changes associated with colorectal cancer are robust, despite differences in environment or diet.

The team used a tool called mOTUs2, which can accurately identify and count bacterial species in metagenomes. “Our method’s strongest feature is that it enables us to quantify bacteria that don’t even have a genomic reference,” says Georg Zeller, the corresponding author of a study published in Nature Medicine. “These species haven’t been isolated and cultivated yet – many of them can’t easily be grown in the lab. But, with mOTUs, we can capture them anyway, which leads to a more complete picture of the microbiome.”

They identified 29 bacteria, which shared common abilities of breaking down proteins and mucins, suggesting potential link between cancer-associated gut microbes and diets rich in meat and fat. This analysis establishes a generalizable and predictive microbiome signature that may serve as a future non-invasive diagnostic for colorectal cancer.