Cancer is terrifying for many reasons, but here’s the one that haunts oncologists: they often can’t tell which tumors will stay put and which will spread. By the time metastasis is detected — cancer cells colonizing distant organs — the window for effective intervention has often closed.

A new AI tool from the University of Geneva is changing that equation. Called MangroveGS (Mangrove Gene Signatures), it predicts whether a cancer is likely to metastasize with nearly 80% accuracy — and it works across multiple cancer types. The research, published this week in Cell Reports, could reshape how doctors decide who needs aggressive treatment and who can be safely monitored.

Cancer Isn’t Chaos — It’s a Hijacked Program

The most fascinating insight from this research challenges conventional thinking. We imagine tumors as anarchic — rogue cells dividing without rules. But lead researcher Ariel Ruiz i Altaba frames it differently: “Cancer should rather be understood as a distorted form of development.”

Genetic and epigenetic changes reactivate biological programs that should’ve stayed dormant after early development. The tumor isn’t random. It’s following a twisted version of the body’s own playbook.

If cancer follows structured rules, those rules can be decoded. And if decoded, predicted.

Why Metastasis Prediction Has Been So Hard

Metastasis kills. It’s responsible for the majority of cancer deaths in colon, breast, and lung cancers. Scientists understand what causes primary tumors to form. But no single genetic alteration explains why some cancer cells break away and migrate while others don’t.

There’s also a brutal research paradox: analyzing a cancer cell’s molecular identity destroys it. Observing its behavior requires keeping it alive. You can’t easily do both.

The Geneva team found a workaround. They isolated and cloned individual tumor cells from two primary colon cancers — roughly thirty clones — then tested them both in vitro and in mouse models to observe actual migratory behavior.

How MangroveGS Works

The team analyzed expression of hundreds of genes across cloned cell populations and identified clear gene expression gradients that tracked each clone’s ability to spread.

The key discovery: metastatic potential isn’t about one cell’s profile. It’s about how groups of related cancer cells interact. Collective behavior, not individual mutations, signals risk.

“The great novelty of our tool is that it exploits dozens, even hundreds, of gene signatures,” explains co-first author Aravind Srinivasan. “This makes it particularly resistant to individual variations.”

Think of it like weather forecasting: instead of one thermometer reading, MangroveGS cross-references hundreds of data points. That redundancy is what pushes accuracy to ~80% — significantly outperforming existing tools.

It Transfers Across Cancer Types

Here’s the headline finding: gene signatures MangroveGS learned from colon cancer transferred to stomach, lung, and breast cancers.

Most AI models in oncology are narrowly trained — they work for one cancer type and collapse elsewhere. MangroveGS’s patterns appear to reflect fundamental metastatic biology, not colon-cancer-specific quirks. If that holds in further validation, we’re looking at the first general-purpose metastasis risk screener.

Oncology has never had one.

What This Means for Patients

Low-risk patients: Less overtreatment. Many patients receive aggressive chemo “just in case.” Reliable low-risk identification could spare them unnecessary side effects.

High-risk patients: Earlier, more aggressive intervention — before spread is clinically detectable.

Clinical trials: Better participant selection means smaller, faster, more conclusive studies that accelerate drug development.

The workflow fits real clinical settings: hospital tumor samples get RNA-sequenced, MangroveGS generates a metastasis risk score, and results are shared securely with medical teams. The researchers anticipate clinical availability within a year.

The Honest Caveat

Let’s be clear about what 80% accuracy means in a life-or-death context. A 20% error rate is significant. Some high-risk patients could be falsely reassured. Some low-risk patients could receive unnecessary treatment.

But context matters: existing metastasis prediction tools perform substantially worse. An 80% accurate tool used as one input among many in clinical decision-making — not the sole arbiter — is a genuine step forward.

The model still needs large-scale prospective clinical trials. The 80% figure comes from retrospective data. Real-world performance in diverse populations could differ.

AI Is Quietly Transforming Oncology

MangroveGS joins a growing wave of AI tools reaching the validation stage — moving from “interesting paper” to “something an oncologist might actually use.” Detection, drug response prediction, target identification, trial optimization — AI is touching every stage of cancer care.

The philosophical shift matters too. Viewing cancer as a structured program rather than cellular anarchy opens new doors. If the rules are orderly, machine learning is exactly the right tool to find patterns humans can’t see in high-dimensional gene expression data.

What Comes Next

The team will likely pursue prospective clinical validation — testing on new patients in real time. Expect partnerships with cancer centers and possible regulatory filings.

The bigger question: can we build a unified AI platform that predicts metastatic risk across all solid tumors from a single biopsy? MangroveGS’s cross-cancer performance suggests we’re closer than anyone expected.

The era of “wait and see” in oncology may be giving way to “predict and prepare.” And that shift could save a lot of lives.

Sources: University of Geneva, Cell Reports, ScienceDaily