EpiClass

EpiClass employs methylation density–based binary classification to predict and optimize the performance of DNA methylation biomarkers for clinical diagnostics, with emphasis on blood-based cancer testing.

Key Features:

• Methylation Density Binary Classification: Implements a binary classifier based on statistical differences in single-molecule sample methylation density distributions to identify optimal thresholds distinguishing case and control samples.
• Single-Molecule Methylation Analysis: Uses intramolecular methylation density information from individual DNA molecules to inform classification decisions.
• Compatibility with Bisulfite Sequencing Data: Applies to reduced representation bisulfite sequencing (RRBS) and whole genome bisulfite sequencing (WGBS) datasets for biomarker evaluation.
• cfDNA Threshold Optimization: Predicts classifier performance by deriving precise methylation density thresholds from circulating cell-free DNA (cfDNA) profiles.
• Validation Using ZNF154 and Ovarian Carcinoma Data: Includes in silico simulations using methylation density profiles from the ZNF154 locus and was developed/validated with RRBS data from ovarian carcinoma tissue DNA and controls.
• Experimental Confirmation and Performance Metrics: Performance predictions were experimentally confirmed using digital methylation density analysis on plasma samples, achieving 91.7% sensitivity and 100.0% specificity in an independent validation cohort and outperforming CA-125 for epithelial ovarian carcinoma (EOC) detection.
• Broad Cancer Applicability: Demonstrated potential to identify multiple cancer types and to perform as well as or better than alternative methylation-based classifiers.

Scientific Applications:

• Liquid Biopsy Diagnostics: Optimizes methylation biomarker selection and classification for blood-based cancer testing using cfDNA.
• Cancer Detection and Monitoring: Enhances detection of epithelial ovarian carcinoma (EOC) and has potential utility across diverse cancer types.
• Biomarker Benchmarking: Enables in silico and experimental benchmarking of methylation-based classifiers using RRBS/WGBS and locus-specific profiles (e.g., ZNF154).

Methodology:

Binary classification based on statistical differences in single-molecule methylation density distributions; in silico simulations using methylation density profiles from the ZNF154 locus; use of RRBS and WGBS datasets and cfDNA-derived methylation density profiles to predict performance and determine optimal thresholds.