AI System Helps Scientists Unlock 100-Year Cancer Mystery

For over a century, scientists suspected abnormal chromosomes could trigger cancer, but finding the rare cells that show these early warning signs was nearly impossible.

Now, researchers at EMBL Heidelberg have built an AI system called MAGIC that can automatically hunt down and mark cells with chromosomal defects. The technology finally gives scientists a way to test ideas about cancer's origins that date back to 1900.

German scientist Theodor Boveri first proposed that abnormal chromosomes inside cells might cause cancer after studying cells under his microscope. His theory made sense, but proving it was frustratingly difficult. Only a tiny fraction of cells show chromosomal problems at any moment, and many die off quickly through natural selection.

For decades, researchers had to manually search for these rare cells one by one under microscopes. They could isolate only a few cells at a time for deeper study, making large scale research almost impossible.

MAGIC changes everything by working like an automated game of laser tag for cells. The system scans cell samples and uses machine learning to identify those containing micronuclei, which are small DNA containing structures that signal future chromosomal trouble. When it spots a problem cell, the system zaps it with a laser beam that marks it with a special dye.

"This project combined a lot of my interests in one," said Marco Cosenza, the research scientist who helped design MAGIC. "It involves genomics, microscopic imaging, and robotic automation."

The tagged cells can later be separated from healthy ones and analyzed in detail. In less than a day, MAGIC can examine close to 100,000 cells compared to the handful researchers could study manually.

The team used their new tool to study cells originally derived from normal human tissue. They discovered that slightly more than 10% of cell divisions naturally produce chromosomal abnormalities, a finding that would have taken years to uncover using old methods.

Why This Inspires

This breakthrough matters because chromosomal abnormalities drive some of the most aggressive cancers. "They're highly linked to patient death, metastasis, recurrence, chemotherapy resistance, and fast tumor onset," said Jan Korbel, senior scientist at EMBL.

Understanding how and when these chromosomal errors happen could help researchers develop better ways to catch cancer earlier or even prevent it from starting. The technology also shows how AI can solve problems that have stumped scientists for generations.

What started as one researcher's curiosity during COVID lockdown has become a powerful tool for answering questions that are over 100 years old.