AI Model Predicts Survival in Childhood Brain Tumors
Researchers say the MANTIS model could help tailor radiation treatment intensity for children with medulloblastoma and ependymoma.
Updated on Nov. 5, 2026 in Cancer
Researchers developed an artificial intelligence model called MANTIS to predict survival outcomes for children with medulloblastoma and ependymoma. The study was published in Nature Genetics.
Why it matters
The model generates individualized survival estimates that may help guide treatment decisions, including how intensive radiation therapy should be. Its predictions were reported to outperform current clinical indicators.
The MANTIS model was trained on DNA methylation data from 2,786 patients and reached a concordance index of up to 0.80 ± 0.07, outperforming current clinical indicators. Its clinical use beyond the study has not yet been established.
The players
Dana-Farber Cancer Institute
A Boston-based cancer research and treatment center affiliated with Harvard Medical School, whose researchers are among the study's authors.
The details
MANTIS, which stands for Methylation profiling and Artificial Neural networks for Time-dependent Individualized Survival predictions, analyzes DNA methylation patterns and copy number changes to produce 10-year survival probabilities for individual patients. In a related application, the model was adapted to test radiation dose levels, finding that about 60% of average-risk medulloblastoma patients would not face a significant survival disadvantage from lower-dose craniospinal irradiation (P = .683).
Timeline
Study published in Nature Genetics; article reporting it appeared on ascopost.com on November 5, 2026.
Health Landscape
The study extends methylation-based profiling, long used to classify pediatric brain tumors, into individualized survival prediction and radiation dose planning. If validated clinically, it could shift treatment decisions for children with medulloblastoma and ependymoma toward more personalized intensity levels.
For families of children with medulloblastoma or ependymoma, a validated tool like MANTIS could inform how much radiation treatment is needed. Any change to standard care would depend on further clinical study before reaching patients.
The takeaway
Personalized survival prediction may eventually let clinicians reduce treatment intensity for patients who do not need the most aggressive approach. Such changes will require clinical validation before becoming routine.
Further reading
Explore more on Cancer research and treatment advances.
Source note: This article includes information reported by Ascopost.
