Turn on cookies to sign in
Signing in saves your progress to your Curie account. We can only do that with cookies on — turn them on to continue.
Glioblastoma, the deadliest brain cancer in adults, has long stumped researchers because its cells constantly shift between states — making treatment nearly impossible. Two new studies reveal that this shape-shifting actually follows predictable patterns, and a new AI tool called scDORI can map and potentially redirect those patterns. The findings open the door to reprogramming tumor cells into less dangerous forms.
Glioblastoma, the most lethal adult brain cancer, has long been considered nearly untreatable in part because its cells constantly morph between different states — some resembling nerve cells, others stem cells — helping them dodge therapies. The prevailing assumption was that this "state-shifting" was essentially random. Two landmark studies published in Nature by researchers at EMBL, DKFZ, and the Wellcome Sanger Institute now challenge that assumption entirely.
By building the largest-ever molecular atlas of glioblastoma — analyzing DNA, epigenetic marks, and RNA from 12 patients across more than 1 million cells — the teams found that cell state transitions follow stereotypical, predictable paths, and that different cell states occupy distinct spatial niches within every tumor studied. To decode this complexity, they developed scDORI, a deep learning tool that infers gene regulatory networks at single-cell resolution, using the tumor's own natural variation as a training dataset.
The researchers also identified "safeguard" regulators that lock cells in low-plasticity states. Overexpressing one such regulator, MYT1L, reduced tumor malignancy and improved survival in mouse models — hinting at future therapeutic strategies.
Key Takeaways:
Why it matters: If tumor cell behavior can be predicted and redirected, it could transform how glioblastoma — and potentially other cancers — are treated, moving from blunt suppression to precise biological reprogramming.
