A new artificial intelligence-powered platform developed at the European Molecular Biology Laboratory (EMBL) in Heidelberg is offering insight into how chromosomal abnormalities arise in cancer biology.
The tool, called MAGIC (machine learning-assisted genomics and imaging convergence), was created by researchers in Jan Korbel’s group at EMBL Heidelberg and published in a new Nature paper. MAGIC is an autonomously operated platform that integrates live-cell imaging of micronucleated cells, machine learning on the fly, and single-cell genomics to systematically investigate cancer formation. It enables scientists to track how spontaneous errors occur during cell division—errors that can set the stage for tumor development.
“Chromosomal abnormalities are a main driver for particularly aggressive cancers, and they’re highly linked to patient death, metastasis, recurrence, chemotherapy resistance, and fast tumor onset,” said Korbel, senior scientist at EMBL and senior author of the new study. “We wanted to understand what determines the likelihood that cells undergo such chromosomal alterations, and what’s the rate at which such abnormalities arise when a still normal cell divides.”
But until recently, directly studying these abnormalities as they form was nearly impossible. Only a small fraction of dividing cells show obvious chromosomal defects at any given time, and such cells often die before they can be analyzed. Scientists traditionally had to spot them manually under the microscope—a slow and painstaking task that allowed only a handful of cells to be studied.
A “laser tag” system for abnormal cells
“Recent reports have established that a single DNA lesion can trigger a cascade of alterations, resulting in chromosomal instability and promoting complex cancer formation processes,” the authors write. “We devised MAGIC to gain insights into cancer formation from studying nuclear atypia. We show that MAGIC facilitates investigating spontaneously arising cancers, providing a representative view of the de novo cancer landscape linked to micronucleation in non-transformed cell lines.”
MAGIC operates like a microscopic game of laser tag. It scans thousands of cells using automated microscopy and an AI algorithm trained to recognize a distinctive cellular feature known as a micronucleus. This small, extra nucleus forms when bits of DNA break off during cell division.
Micronuclei are more than just debris; they are signs of trouble. Cells with micronuclei are more likely to produce new chromosomal abnormalities and can eventually transform into cancer cells.
Once the AI detects cells with micronuclei, it instructs the microscope to “tag” them using a photoconvertible dye—a fluorescent molecule that changes color when illuminated with a laser. This light-based tagging allows researchers to track and isolate those specific cells later on using flow cytometry, a technique that separates cells based on their fluorescent signals.
By automating this entire process—from image capture to cell tagging to genetic analysis—MAGIC removes the need for manual selection and dramatically accelerates the pace of research. In less than a day, scientists can now analyze nearly 100,000 cells.
Tracking chromosome errors as they happen
In the Nature study, the EMBL team used MAGIC to monitor chromosomal abnormalities in cultured human cells. They found that more than 10% of normal cell divisions produce some kind of spontaneous chromosomal alteration. When TP53, the gene encoding p53—a key tumor suppressor often mutated in cancers—was disabled, that rate nearly doubled, underscoring how p53 helps maintain genomic stability.
The researchers also pinpointed specific mechanisms behind these abnormalities. One major culprit appears to be dicentric chromosomes—chromosomes with two centromeres instead of one. These unstable structures can get pulled in opposite directions during cell division, leading to chromosome breakage and rearrangement.
In the study, MAGIC enabled automated analysis of several tens of thousands of cells per experiment, permitting the isolation of rare cell morphologies at large numbers, and thus overcoming previous limitations in studying nuclear atypia. In total, the EMBL team isolated 2,898 single cells and sequenced 2,192 single-cell genomes in this study, generating an unprecedented dataset for investigating de novo cancers.
Interestingly, the researchers found a bias toward chromosome losses rather than gains in newly formed abnormalities. This mirrors what has been observed in cancer genomes: across more than 2,600 tumor samples analyzed by the team, 81.5% of chromosomal abnormalities involved losses rather than gains. “Although proteotoxic stress linked to trisomy can select against chromosome gains, our data indicate that the bias is established during cancer formation, preceding proteotoxic effects,” the authors wrote.
MAGIC is not limited to studying cancer-related abnormalities. Because the system’s AI can be retrained to recognize virtually any visible cellular feature, it could be adapted to investigate many other biological questions—from neurodegeneration to aging to genetic diseases.
“As long as you have a feature that can be discriminated visually from a ‘regular’ cell, you can—thanks to AI—train the system to detect it,” said Korbel. “Our system, therefore, has the potential to advance future discoveries in numerous areas of biology.”
