New AI Model Could Help Doctors Predict Cancer Spread With Remarkable Accuracy
A new breakthrough in cancer research uses artificial intelligence to analyze gene activity in tumors and predict the likelihood of cancer spreading, potentially helping doctors personalize treatment.
A major breakthrough in cancer research could change how doctors predict and treat cancer. Scientists have developed a new artificial intelligence system that can estimate whether a tumor is likely to spread to other parts of the body. The tool analyzes patterns in how genes behave inside cancer cells and can estimate the risk of metastasis, the process by which cancer spreads.
The research was conducted by scientists at the University of Geneva (UNIGE) and published in the journal Cell Reports. The team developed an AI tool called MangroveGS, which converts complex genetic data into predictions about cancer spread.
Why some cancers spread, and others don’t
One of the biggest questions in cancer research is why some tumors stay in one place while others spread throughout the body. Doctors know a lot about the mutations that cause tumors to form, but they still don’t fully understand what makes certain cancer cells break away and form new tumors elsewhere.
Researchers from the University of Geneva (UNIGE) studied this question using cells from colon cancer tumors as part of their ongoing cancer research efforts. Their work helped identify patterns in gene activity that are linked to how likely a tumor is to spread. Using this information, the team created the AI tool MangroveGS, which can turn these genetic signals into predictions about the risk of cancer spreading.
Cancer may follow hidden biological programs. According to the scientists, cancer is often described as cells behaving “chaotically.” But researchers involved in this cancer research believe cancer is actually more organized than that.
During early development of the body, certain biological programs guide how tissues and organs grow. Later in life, these programs normally shut down. However, genetic and epigenetic changes can sometimes reactivate them in the wrong place and time. When that happens, these programs can drive the growth of tumors.
“Cancer is often described as anarchic cells, but it is better understood as a distorted form of development,” said Ariel Ruiz i Altaba, a professor at the University of Geneva, who led the study.
Understanding this process is a major focus of modern cancer research because it could help scientists figure out why some cancer cells gain the ability to spread.
Studying the cells that spread:
Metastasis is responsible for most cancer deaths, especially in colon, breast, and lung cancers, making it one of the most urgent challenges in cancer research. Usually, doctors detect metastasis when cancer cells appear in the bloodstream or lymphatic system. By that stage, the disease may already be spreading.
To study this process more closely, the researchers isolated individual tumor cells from colon cancer patients. They grew these cells in the lab and tested how easily they could move and form new tumors in experimental models, an approach commonly used in advanced cancer research.
They then measured the activity of hundreds of genes in about 30 cloned cancer cells. The scientists discovered clear patterns in gene activity that were strongly linked to how easily the cells could move and spread, an important finding for cancer research.
Interestingly, the results showed that metastasis is not determined by just one cell. Instead, it seems to depend on how groups of related cancer cells interact with each other inside the tumor.
AI predictions with high accuracy
The researchers used these gene patterns to train their AI system, MangroveGS, combining artificial intelligence with cutting-edge cancer research. Instead of relying on just a few genetic markers, the tool analyzes dozens or even hundreds of gene signatures. This makes the system more reliable and less sensitive to individual variations between patients, something researchers in cancer research have long been trying to achieve.
After training, the AI predicted the chances of colon cancer spreading or returning with nearly 80% accuracy, which is better than many existing prediction methods.
The scientists also found that the same gene patterns could help predict metastasis in other cancers, including stomach, lung, and breast cancers, highlighting its broader potential in cancer research.
How could this help patients?
In the future, MangroveGS could become part of routine cancer care, a goal many experts in cancer research are working toward. Doctors would simply take a tumor sample and analyze the RNA inside the cells. The data would be processed by the system to produce a metastasis risk score.
This score could help doctors make better treatment decisions, something that has long been a key objective of cancer research.
For example:
- Patients with a low risk of metastasis might avoid unnecessary aggressive treatments and side effects.
- Patients with high risk could receive closer monitoring and stronger therapies earlier.
The tool could also help researchers design better clinical trials in cancer research by selecting patients who are most likely to benefit from new treatments.
Overall, this AI system could lead to more personalized cancer treatment and help scientists discover new ways to stop cancer from spreading, marking a significant milestone in ongoing cancer research.
