DNA methylation-based classification of hematolymphoid neoplasms

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Accurate pathological diagnosis of hematolymphoid neoplasms (HLN) is challenging due to complexity and heterogeneity.
• Genome-wide DNA methylation profiling is a powerful tool for tumor classification across various cancers.

Purpose of the Study:

• To explore the utility of DNA methylation profiling for classifying hematolymphoid neoplasms.
• To develop and validate a machine learning-based DNA methylation classifier for HLN.

Main Methods:

• Generated the largest cross-platform HLN methylome cohort (1,156 samples).
• Identified 44 reproducible methylation classes (MCs) aligned with WHO entities.
• Developed and validated a machine learning classifier, assessing performance and impact of tumor purity.

Main Results:

• 44 MCs were identified, correlating with WHO 5th edition/ICC entities and revealing clinically relevant subgroups.
• The DNA methylation classifier achieved 97% concordance with original diagnoses in high-confidence cases.
• In discrepant high-confidence cases, the classifier aided in revising diagnoses in the majority (5/8).
• Tumor purity significantly impacted classifier performance in low-confidence samples.

Conclusions:

• Distinct methylation signatures characterize hematolymphoid tumor types, enabling accurate classification.
• DNA methylation profiling serves as a valuable adjunct diagnostic tool for HLN.
• This pilot study lays the groundwork for a future clinical methylation-based classifier for HLN.