Researchers at the University of Geneva have used cells from intestinal tumors to identify criteria that influence the risk of metastases: They determined gene expression signatures to assess the probability and developed an artificial intelligence tool called MangroveGS. This converts the data into predictions for numerous types of cancer and achieves a reliability of almost 80 percent. The results pave the way for more precise treatments and the discovery of new therapeutic targets. The study was published in the journal Cell Reports.

Cancer arises from the activation of programs that are normally suppressed during organ and tissue development. Metastases remain the main cause of cancer death, especially in colorectal, breast and lung cancer. Previous methods of prediction are often based on mutations that explain the primary tumor, but no single genetic change clarifies why some cells migrate and others do not. The challenge is to capture the molecular identity of a cell without destroying it, while observing its function.

The researchers isolated, cloned and cultured tumor cells from two primary colorectal tumors. These clones were examined in vitro and in a mouse model for their migration potential. Analysis of the expression of several hundred genes revealed gene expression gradients that are closely correlated with the potential for metastasis. The risk assessment is not based on the profile of a single cell, but on the sum of the interactions between related cancer cells.

The MangroveGS tool integrates dozens to hundreds of gene signatures and is therefore resistant to individual variations. It achieved nearly 80 percent accuracy in predicting metastases and relapses in colorectal cancer, outperforming existing tools. Signatures from colorectal cancer can also be applied to other types of cancer such as stomach, lung or breast cancer.

Tumor samples are sufficient for analysis. The highlight: The cells are sequenced and the risk score is transmitted to oncologists via an encrypted portal. This helps to avoid low-risk overtreatment, minimise side effects and reduce costs. High-risk patients can be monitored and treated more intensively. In addition, the tool optimizes the selection of participants for clinical trials, reduces the number of volunteers required and increases statistical value.

The method thus opens up new perspectives for precision medicine. By taking into account the life course and the realities of care, inequalities in treatment can be reduced. The study underlines that cancer must be understood as an orderly, albeit cumbersome, development. This could lead to new approaches to block metastasis.

The researchers see potential for broader applications. The tool could be integrated into routine examinations and improve prognosis. It addresses the lack of tools that use multiple signatures. The high accuracy is based on the robustness against variations. The study could influence guidelines and lead to more targeted therapies. In oncology, it could increase the chances of survival by detecting risks early.

The work combines cell analysis, gene expression and AI. The results are based on clones that reflect the migration potential. The method is scalable and could be implemented in hospitals. The researchers are planning further validations for other types of cancer. The study highlights the need to view cancer as a dynamic process in order to develop effective strategies.

Original Paper:

“Emergence of high-metastatic potentials and prediction of recurrence and metastasis”
Aravind Srinivasan, Arwen Conod, Yann Tapponnier, Marianna Silvano, Luca Dall’Olio, Céline Delucinge-Vivier, Isabel Borges-Grazina, Ariel Ruiz i Altaba
Cell Reports
10.1016/j.celrep.2025.116834

Editor: X-Press Journalistenbüro GbR

Gender Notice. The personal designations used in this text always refer equally to female, male and diverse persons. Double/triple naming and gendered designations are used for better readability. ected.