Researchers at the University of Geneva have developed MangroveGS, an AI model that predicts cancer metastasis risk with nearly 80% accuracy. The tool analyzes gene expression patterns in tumor cells, initially from colon cancer, and applies to other types like breast and lung. Published in Cell Reports, it aims to enable more personalized treatments.
Researchers led by Ariel Ruiz i Altaba, professor in the Department of Genetic Medicine and Development at the University of Geneva (UNIGE) Faculty of Medicine, studied colon tumor cells to understand metastasis. They found that cancer spread follows structured biological programs rather than random processes, reactivating early developmental pathways through genetic and epigenetic changes. Metastasis accounts for most deaths in colon, breast, and lung cancers, but detecting it early remains challenging as it often begins before cells are found in blood or lymphatics. No single mutation fully explains why some cells migrate while others do not. To address this, the team isolated, cloned, and grew about thirty cell clones from two primary colon tumors. These were tested in vitro and in mouse models for migration and metastasis formation. Analysis of hundreds of genes revealed expression patterns linking groups of cancer cells to metastatic potential, rather than individual cells. These patterns were integrated into MangroveGS, an AI tool that uses dozens or hundreds of gene signatures for robust predictions. Aravind Srinivasan noted, 'The great novelty of our tool, called Mangrove Gene Signatures (MangroveGS), is that it exploits dozens, even hundreds, of gene signatures. This makes it particularly resistant to individual variations.' The model predicts metastasis and colon cancer recurrence with nearly 80% accuracy, outperforming prior methods, and applies to stomach, lung, and breast cancers. It processes hospital tumor samples via RNA sequencing, generating risk scores shared securely with doctors and patients. Ruiz i Altaba stated, 'This information will prevent the overtreatment of low-risk patients, thereby limiting side effects and unnecessary costs, while intensifying the monitoring and treatment of those at high risk.' The study appears in Cell Reports (2026; 45(1):116834).
