Artem Shmatko

Abstract Molecular testing is essential for classifying central nervous system (CNS) tumors, with methylation profiling providing the highest diagnostic granularity. However, this requires more resources and time than conventional hematoxylin and eosin (H&E) histopathology, which is widely available globally. Here we propose Hetairos, an artificial intelligence algorithm that predicts 102 methylation-based CNS tumor subtypes from digital H&E slides.

Correction to: Nature https://doi.org/10.1038/s41586-025-09529-3 Published online 17 September 2025 In the version of this article initially published, in the Methods “Exponential waiting time model” section, there was a typographical error in the equation now reading “lossj = – log P(j) = cross_entropy(logits, tokens),” where a minus sign originally appeared before the cross-entropy term. The equation has been amended in the HTML and PDF versions of the article.

Abstract Decision-making in healthcare relies on understanding patients’ past and current health states to predict and, ultimately, change their future course1,2,3. Artificial intelligence (AI) methods promise to aid this task by learning patterns of disease progression from large corpora of health records4,5. However, their potential has not been fully investigated at scale.