Clinical Evaluation of the CrossNN DNA Methylation Classifier for Central Nervous System Tumors

In this study, we evaluated the performance of the CrossNN classifier in a real-world CNS tumor cohort and compared it with the established Heidelberg classifier to assess its potential as both a non-inferior alternative and a complementary tool to improve diagnostic accuracy.

Methods

A retrospective cohort of CNS tumors profiled using Illumina Human Methylation 930k EPIC v2 BeadChip arrays was used to evaluate the performance of the CrossNN classifier. Classifier outputs were compared with integrated WHO CNS5 (2021) diagnoses. In addition, CrossNN and Heidelberg outputs were harmonized to WHO CNS5 (2021) tumor entity levels and evaluated both sequentially and in parallel to evaluate their complementarity.

Results: A total of 205 samples were included. CrossNN generated correct and confident predictions in 76.1% and showed concordant results with the Heidelberg classifier in 87.3% of cases. Complementary use of both classifiers improved diagnostic accuracy by approximately 10%.

Conclusions: CrossNN demonstrated classification performance non-inferior to the Heidelberg classifier, consistent with previously reported CrossNN accuracy. CrossNN therefore represents a robust, alternative methylation-based classifier for CNS tumor classification. Combined use of CrossNN and Heidelberg further improved diagnostic accuracy while retaining a low rate of incorrect predictions.