Cancer Metastasis Prediction Tool Achieves 80% Accuracy Using Gene Pattern Analysis

Researchers have developed an artificial intelligence tool capable of predicting which cancers will metastasize with approximately 80 percent accuracy, a breakthrough that reframes cancer spread as a programmed biological process rather than a random event.

The system, called MangroveGS, was built by a team at Université de Genève after studying gene patterns in colon tumor cells. The tool converts genetic signals into predictions that work across multiple cancer types, according to findings published in Cell Reports. The research suggests that certain tumors carry identifiable molecular signatures that indicate metastatic potential before spread occurs.

The discovery challenges longstanding clinical assumptions. Rather than treating all aggressive cancers with uniform intensity, the tool could enable physicians to stratify patients based on biological risk profiles, reserving the most aggressive interventions for cases where metastasis is genetically probable. Conversely, patients whose tumors lack high-risk signatures might avoid overtreatment.

(The research was conducted on colon cancer cells initially, then validated across other cancer types. The 80 percent accuracy rate represents performance in controlled research settings; clinical deployment timelines and regulatory pathways were not disclosed in available reports.)

The work arrives as artificial intelligence applications in oncology expand beyond imaging and diagnostics into molecular prediction. Traditional cancer staging relies heavily on anatomical spread and histological grading, methods that offer limited foresight into metastatic behavior. Gene-based prediction tools like MangroveGS represent a shift toward molecular prognostication, though questions remain about how such systems will integrate into existing clinical workflows and whether accuracy holds across diverse patient populations.

The Université de Genève team framed their findings around the concept of cancer as a "distorted development process," suggesting that metastatic programs may be encoded early in tumorigenesis. If validated in larger clinical trials, the approach could influence treatment protocols and open avenues for therapies targeting the genetic switches that activate metastasis.