AI-based precision prognostics and therapy personalization for childhood brain tumors.

MANTIS, an AI model using genome-wide DNA methylation arrays, predicts 10-year individualized survival for medulloblastoma and ependymoma in 2,786 patients with concordance indices up to 0.80 ± 0.07.

• Why it matters: Childhood brain tumor treatment must balance recurrence and death risk against long-term therapy sequelae, yet current clinical indicators do not predict outcomes accurately enough for individualized decisions.
• What they did: The study built MANTIS from methylation array data and used artificial neural networks to generate time-dependent survival probabilities over 10 years, then compared performance to existing clinical indicators.
• The result: Extending MANTIS to estimate individualized treatment effects identifies ~60% of average-risk medulloblastoma patients with no significant survival disadvantage (P = 0.683) when given less intensive craniospinal irradiation, linked to improved neurocognitive outcomes.

AI summary of the abstract · may contain errors · not medical advice · read the paper.