DNA Methylation-Based Classification of Kidney Neoplasms.

Area of Science:

• Genomic Medicine
• Oncology
• Computational Biology

Background:

• Renal neoplasms exhibit significant heterogeneity, complicating accurate diagnosis.
• Interobserver variability and overlapping microscopic features hinder classification.
• A subset of renal tumors remains unclassifiable even with advanced diagnostic techniques.

Purpose of the Study:

• To leverage genome-wide DNA methylation signatures for improved renal neoplasm classification.
• To develop and validate a machine learning-based classifier for renal tumor diagnosis.
• To explore the potential of DNA methylation profiling in identifying novel subtypes and aiding clinical reclassification.

Main Methods:

• Analysis of genome-wide DNA methylation profiles from over 2000 renal neoplasms.
• Utilizing machine learning models to train and validate a diagnostic classifier on 1284 samples.
• Testing the classifier on an independent dataset of 287 renal neoplasms.

Main Results:

• Identification of 23 coherent groups correlating with known renal neoplasm types.
• Discovery of novel, clinically relevant subtypes within existing tumor categories.
• Achieved >90% concordance between expected tumor type and DNA methylation-based classification on an external dataset.
• Identified cases where methylation classification prompted potential reclassification of the original histologic diagnosis.

Conclusions:

• Demonstrates the feasibility of a DNA methylation classifier as a clinically applicable tool for renal neoplasm diagnosis.
• Highlights the potential of DNA methylation profiling to resolve diagnostic ambiguities and refine tumor classification.
• Supports the integration of molecular data, specifically DNA methylation, into routine diagnostic workflows for renal cancers.