Epiclomal: Probabilistic clustering of sparse single-cell DNA methylation data.

Area of Science:

• Computational Biology
• Genomics
• Cancer Research

Background:

• Single-cell DNA methylation data is sparse and contains many missing values.
• Existing clustering methods struggle with the high dimensionality and missing data in single-cell CpG datasets.
• Understanding clonal heterogeneity in tumors is crucial for cancer research.

Purpose of the Study:

• To develop a probabilistic clustering method for simultaneous clustering and imputation of sparse single-cell DNA methylation data.
• To evaluate the performance of the new method against existing approaches using synthetic and real datasets.
• To apply the method to discover novel sub-clonal methylation patterns in cancer genomes.

Main Methods:

• Developed Epiclomal, a probabilistic clustering method based on a hierarchical mixture model.
• Applied Epiclomal to synthetic and published single-cell CpG datasets.
• Utilized newly generated single-cell 5mCpG sequencing data from aneuploid tumor genomes.

Main Results:

• Epiclomal outperforms non-probabilistic methods in clustering and imputing sparse single-cell DNA methylation data.
• The method effectively handles the missing data characteristic of single-cell CpG sequences.
• Epiclomal identified sub-clonal methylation patterns ('epiclones') in aneuploid tumor genomes, revealing novel clonal structures.

Conclusions:

• Epiclomal provides a robust solution for analyzing sparse single-cell DNA methylation data, including imputation of missing values.
• The discovered epiclones offer a new dimension for clonal analysis in cancer, complementing copy number-based lineage tracing.
• This method advances the understanding of epigenetic heterogeneity and clonal evolution in cancer.