Epiclomal

Epiclomal performs probabilistic clustering and imputation of sparse single-cell DNA methylation (CpG, 5mCpG) data to identify epigenetic subpopulations and handle missing values.

Key Features:

• Hierarchical mixture model: Uses a probabilistic hierarchical mixture model to represent single-cell CpG methylation patterns.
• Joint clustering and imputation: Simultaneously clusters cells and imputes missing CpG methylation states within the same probabilistic framework.
• Missing data handling: Explicitly models and imputes missing values typical of sparse single-cell CpG/5mCpG sequencing data.
• Validation and benchmarking: Validated on synthetic and published single-cell CpG datasets and reported to outperform non-probabilistic methods.
• Epiclone detection: Identifies "epiclones," i.e., sub‑clonal methylation clusters that may align with or extend beyond copy number variation–defined clonal lineages.
• Application to aneuploid genomes: Applicable to analysis of aneuploid tumor genomes for sub‑clonal methylation pattern discovery.
• Implementation: Implemented in R and Python.

Scientific Applications:

• Single‑cell methylation analysis: Analysis of single‑cell CpG and 5mCpG sequencing datasets to characterize epigenetic heterogeneity.
• Clonal and sub‑clonal analysis in cancer: Uncovering sub‑clonal methylation patterns and defining epiclones within aneuploid tumor genomes to complement copy number variation–based lineage analysis.
• Study of cellular heterogeneity: Characterizing epigenetic cellular heterogeneity and its role in tumorigenesis and evolution.

Methodology:

Employs a probabilistic hierarchical mixture model to perform simultaneous clustering and imputation of sparse single‑cell CpG methylation data, validated on synthetic and published single‑cell CpG datasets and compared to non‑probabilistic methods.