Chromosomal abnormalities that occur during cell division can affect the cell’s health, in some cases causing a normal cell to become cancerous. The role of these aberrations in the progression of cancer, and the baseline rate at which they occur, remain poorly understood. To investigate the origins of chromosomal instability, researchers from the European Molecular Biology Laboratory (EMBL) created an autonomous platform to examine the cellular context, mutation rates, and triggers leading to the spontaneous formation of chromosomal abnormalities. The researchers tested the platform on cell line models that mimicked the early stages in tumor evolution. This tool, assisted by artificial intelligence (AI), could provide valuable insight into the molecular origins of cancer, laying the groundwork for new approaches to genetic research and potential strategies for preventing cancer. “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,” researcher Jan Korbel said. “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.” MAGIC, short for machine learning-assisted genomics and imaging convergence, operates like a fully automated game of laser tag, spotting cells that exhibit a particular visible feature, like the presence of micronuclei, and marking them using a laser and a photoconvertible dye. Courtesy of Daniela Velasco/EMBL. Researcher Marco Cosenza said that the AI-assisted platform combines genomics, microscopic imaging, and robotic automation. Called MAGIC, short for machine learning-assisted genomics and imaging convergence, the platform integrates live-cell imaging, using confocal microscopy, with machine learning to assess cellular irregularities on-the-fly. The automated microscope captures a series of images of the cell sample, and a machine learning algorithm scans the images. The system targets micronuclei — enclosed compartments inside cells that contain a small portion of the cell’s DNA, which is separated from the bulk of the genome. Cells with micronuclei tend to produce new chromosomal abnormalities, which make them more likely to turn cancerous. The researchers trained the algorithm used by MAGIC on manually annotated datasets of micronuclei-containing cells. When the algorithm identifies cells with micronuclei, it shares their location with the microscope and instructs the microscope to shine light specifically on those cells, permanently tagging them with a laser. The system uses a photoconvertible dye to photolabel the target cells. The tagged cells are isolated from the still-living, heterogeneous cell population using methods like flow cytometry. The isolated cells then undergo single-cell sequencing and systematic phenotype analyses. The researchers used MAGIC to analyze chromosomal abnormalities in cultured cells originally derived from normal human cells. They determined a baseline chromosomal abnormality mutation rate, observing that slightly more than 10% of all cell divisions resulted in spontaneous chromosomal abnormalities of some kind. They further found that the mutation rate nearly doubled in cells where a specific gene, known to be a tumor suppressor, was deficient. The team also investigated the presence and location of double-stranded DNA breaks within a chromosome, and found that the targeted introduction of DNA double-strand breaks along chromosome arms triggered distinct chromosomal abnormalities. MAGIC enables automated analysis of several tens of thousands of cells per experiment, permitting the isolation of rare cell morphologies at large numbers. In total, the researchers isolated 2898 single cells and sequenced 2192 single-cell genomes, generating a large dataset for investigating chromosomal abnormalities. By automating the labor-intensive, time-consuming, error-prone process previously used to detect cells with micronuclei, MAGIC could enable scientists to study these cells at a scale and speed previously not possible. MAGIC is highly versatile and adaptable. For this study, the algorithm was trained to recognize cells that had micronuclei, but in theory, the algorithm could be trained on many kinds of datasets to detect different cellular features. “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,” Korbel said. “Our system, therefore, has potential to advance future discoveries in numerous areas of biology.” The research was published in Nature (www.doi.org/10.1038/s41586-025-09632-5).