AI Breakthrough Predicts Cancer Metastasis with Unprecedented Accuracy
Table of Contents
Geneva-based researchers have developed an artificial intelligence tool, MangroveGS, capable of predicting cancer metastasis and recurrence with nearly 80% accuracy, offering a potential paradigm shift in patient care and clinical trial design.
The elusive question of why some tumors spread while others remain localized has long plagued the medical community. Understanding the mechanisms driving metastasis – the process by which cancer cells break away from the primary tumor and establish new tumors elsewhere in the body – is crucial for improving patient outcomes. Now, scientists at the University of Geneva (UNIGE) believe they’ve unlocked key insights, identifying gene expression signatures that can assess the probability of metastasis and translating those findings into a powerful predictive tool. The results, published in Cell Reports in July 2025, signal a major step forward in personalized oncology.
“The origin of cancer is often attributed to ‘anarchic cells’,” explains a leading researcher on the project. “However, cancer should rather be understood as a distorted form of development.” The team’s work suggests that cancer isn’t a random process, but rather a reactivation of developmental programs suppressed during normal tissue formation, driven by genetic and epigenetic changes.
This reframing of cancer as a distorted developmental process highlights the importance of understanding its underlying logic. “The challenge is therefore to find the keys to understanding its logic and form,” the researcher continued. “And, in the case of metastases, to identify the characteristics of the cells that will separate from the tumor to create another one elsewhere in the body.”
The Deadly Spread of Metastasis
Metastasis remains the primary cause of cancer-related deaths, particularly in cases of colon, breast, and lung cancer. Currently, the first indication of metastatic spread is often the detection of circulating tumor cells in the bloodstream or lymphatic system – a point at which intervention is often too late. While the genetic mutations driving the initial tumor formation are relatively well understood, the factors determining why certain cells metastasize while others do not have remained largely mysterious.
A significant hurdle in this research has been the challenge of analyzing a cell’s complete molecular identity without destroying it, as the process of observation often requires the cell to be fixed or otherwise altered. “The difficulty lies in being able to determine the complete molecular identity of a cell – an analysis that destroys it – while observing its function, which requires it to remain alive,” one researcher noted.
To overcome this obstacle, the UNIGE team employed a novel approach. “To this end, we isolated, cloned and cultured tumor cells,” explains a senior lecturer involved in the study. “These clones were then evaluated in vitro and in a mouse model to observe their ability to migrate through a real biological filter and generate metastases.”
Gene Expression Gradients and the Power of AI
Analysis of gene expression in approximately thirty clones derived from two primary colon tumors revealed significant gene expression gradients correlated with migratory potential. Crucially, the researchers found that assessing metastatic potential isn’t about the characteristics of a single cell, but rather the collective interactions within a group of related cancer cells.
These gene expression signatures were then integrated into an artificial intelligence model, dubbed Mangrove Gene Signatures (MangroveGS). “The great novelty of our tool…is that it exploits dozens, even hundreds, of gene signatures. This makes it particularly resistant to individual variations,” explains a PhD student and co-first author of the study. After rigorous training, MangroveGS achieved an impressive accuracy rate of nearly 80% in predicting both the occurrence of metastasis and cancer recurrence in colon cancer patients – a substantial improvement over existing predictive tools.
Furthermore, the model demonstrated the ability to extrapolate beyond colon cancer, successfully predicting the metastatic potential of other cancers, including stomach, lung, and breast cancer. .
Transforming Clinical Practice and Accelerating Research
The implications of MangroveGS are far-reaching. The tool requires only standard tumor samples, allowing for analysis of RNA sequencing data at the hospital level. A secure, encrypted portal then transmits the resulting metastatic risk score to oncologists and patients. “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 lead researcher stated.
The tool also promises to revolutionize clinical trial design. “It also offers the possibility of optimising the selection of participants in clinical trials, reducing the number of volunteers required, increasing the statistical power of studies, and providing therapeutic benefits to the patients who need it most.”
This groundbreaking research was supported by the Swiss National Science Foundation (SNSF), the Swiss Cancer Research Foundation, and the DIP of the State of Geneva.
