epimutacions

epimutacions detects epimutations in DNA methylation data to identify rare, locus-specific alterations in methylation patterns that may underlie rare diseases.

Key Features:

• Statistical approaches: Implements two previously reported methods plus MANOVA, multivariate linear models, isolation forest, robust Mahalanobis distance, quantile, and beta approaches for epimutation detection.
• Case versus reference comparison: Compares a case sample against a reference panel of healthy individuals to identify outlier methylation loci.
• Performance at low sample sizes: Demonstrates superior performance for epimutation detection at low sample sizes compared with the R package ramr.
• Validation and benchmarking: Validated and benchmarked using three public datasets containing experimentally validated epimutations and compared to ramr.
• Experimental design insights: Analyzes INMA and HELIX child cohorts to identify technical and biological factors affecting epimutation detection and to inform experimental design and preprocessing.
• Clinical application: Applied to a cohort of children with autism to identify novel recurrent epimutations in candidate genes.
• Annotation and visualization: Includes functions to annotate and visualize identified epimutations.

Scientific Applications:

• Rare disease epimutation discovery: Identification of rare, locus-specific DNA methylation alterations potentially linked to rare diseases.
• Low-sample-size studies: Detection and analysis of epimutations in studies with limited sample sizes.
• Method benchmarking and validation: Benchmarking and validation of epimutation detection methods using public datasets with experimentally validated events.
• Cohort analyses and experimental design: Investigation of technical and biological confounders in population cohorts (INMA, HELIX) and guidance for preprocessing and study design.
• Clinical genomics and candidate gene discovery: Discovery of recurrent epimutations in candidate genes within autism cohorts for clinical research.

Methodology:

Implements two previously reported methods and additional approaches—MANOVA, multivariate linear models, isolation forest, robust Mahalanobis distance, quantile, and beta approaches—to compare a case sample against a reference panel of healthy individuals; validated on three public datasets with experimentally validated epimutations and includes functions for annotation and visualization.