Researchers from EMBL, LUMC, and collaborators reanalysed nearly 6,800 gut microbiome profiles in colorectal cancer patients and controls.

One of the largest single-disease gut microbiome meta-analyses to date, the new study identified a robust colorectal cancer microbiome signature that was consistent across populations, sequencing methods, and age-of-onset groups. 

A machine-learning classifier could distinguish colorectal cancer from non-cancer microbiomes across datasets. 

The colorectal cancer microbiome signature was linked to lower dietary fibre intake and could be reduced through fibre-focused dietary interventions.

Researchers from Donghua University and Yanshan University have developed an enhanced neural network called EF-YOLO, which significantly improves the automated detection of colorectal polyps—a key precursor to colorectal cancer. By integrating advanced modules such as transformer, deformable convolution, coordinate attention, and a newly designed multi-scale aggregation mechanism, the model achieves a mean average precision (mAP) of 96.60% on standard medical datasets. This represents a major step forward in reducing missed detections during colonoscopy and supporting early clinical intervention.