Synthetic epigenomes for open cross-platform episignature classifiers

MethaDory trains and releases synthetic-DNAm episignature classifiers, achieving an average F1 of 0.9 on arrays while covering 102 rare disorders.

• Why it matters: Episignature diagnosis is limited by platform-specific testing and by scarce, non-shareable positive controls, blocking independent classifier training and reproducibility.
• What they did: The framework generates in silico synthetic DNAm profiles from control data plus affected-individual summary statistics, then trains 188 classifiers and validates 154 across 915 diagnosed cases and 533 controls.
• The result: MethaDory matches or outperforms existing approaches on arrays and enables applying episignature classifiers to long-read genome data for diagnoses in unsolved and ultra-rapid settings.

AI summary of the abstract · may contain errors · not medical advice · read the paper.