Why do some tumors travel to other parts of the body while others stay put? Doctors still don’t know all the reasons, but finding out is key to better care.
Researchers at the University of Geneva looked at cells from colon cancer. They discovered several important factors that decide if a tumor will spread. They also found patterns in gene activity that can tell how risky a tumor is.
Using these patterns, the team built a computer program called MangroveGS. The program turns the gene signals into strong predictions for many kinds of cancer. The findings, published in *Cell Reports*, could help doctors give more tailored treatments and find new drug targets.
Seeing Cancer as a Skewed Growth Process
Professor Ariel Ruiz i Altaba explains that cancer isn’t just “out‑of‑control cells.” Instead, it is a twisted version of normal development. Changes in DNA and other cell parts can reactivate programs that are usually turned off after early growth, and these programs drive tumor formation.
Cancer follows its own set of rules. The challenge is to learn those rules and spot the cells that break away to form new tumors elsewhere.
Finding Cells That Travel
Most cancer deaths come from metastasis – the spread of cancer to new places. By the time cancer cells appear in blood or lymph, the disease often has already started to move. Many mutations cause tumors, but none alone explain why some cells leave the main tumor.
The scientists grew tumor cells in the lab, copied them, and tested them. They watched how the cells moved through a real‑life filter and formed new tumors in mice.
Gene Patterns That Signal Spread
They measured hundreds of genes in about thirty cell clones from two colon tumors. Clear patterns emerged that matched each cell’s ability to move. The risk of spreading depended on groups of cells working together, not just one cell alone.
AI Tool Gives a Risk Score
All the gene patterns were fed into an artificial‑intelligence system. The new tool, Mangrove Gene Signatures (MangroveGS), looks at dozens, even hundreds, of gene signals. This makes it tough for any single variation to fool the test.
After learning from data, the model predicted metastasis and cancer return with about 80 % accuracy, better than older methods. The same gene clues also helped predict spread in stomach, lung, and breast cancers.
Moving Toward Personalized Care
MangroveGS can be used with real tumor samples from hospitals. Doctors sequence the tumor’s RNA, run the test, and receive a risk score quickly through a secure system.
This score can stop doctors from giving strong treatments to low‑risk patients, sparing them side effects and costs. At the same time, it can alert doctors to watch high‑risk patients more closely and treat them earlier.
The tool could also help choose the right people for clinical trials, making studies smaller, stronger, and more helpful for those who need new therapies.