Advancing CNS tumor diagnostics with expanded DNA methylation-based classification

Area of Science:

• Neuro-oncology
• Genomics
• Bioinformatics

Background:

• DNA methylation profiling is crucial for classifying central nervous system (CNS) tumors, as recognized by the World Health Organization (WHO).
• Previous versions of the Heidelberg classifier have aided in CNS tumor diagnosis.
• Understanding CNS tumor heterogeneity is key for accurate classification.

Purpose of the Study:

• To introduce the Heidelberg CNS Tumor Methylation Classifier version 12.8 (v12.8).
• To expand the classification of CNS tumors to a subclass level.
• To enhance diagnostic precision in neuro-oncology.

Main Methods:

• Training the classifier on 7,495 DNA methylation profiles.
• Utilizing a random forest algorithm for classification.
• Incorporating data from a large online repository and global collaborations.

Main Results:

• The v12.8 classifier recognizes 184 CNS tumor subclasses, a significant expansion from 91 classes.
• Achieved 95% accuracy at the subclass level.
• Probabilistic scores provide reliable confidence measures for classifications.
• Hierarchical output supports interpretation across multiple taxonomic levels.

Conclusions:

• The v12.8 classifier offers improved precision and practical utility over previous versions and WHO-based methods.
• This advancement supports personalized neuro-oncology through more granular tumor classification.
• The classifier aids clinical decision-making by providing detailed diagnostic information.