scHiMe: predicting single-cell DNA methylation levels based on single-cell Hi-C data

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• Methyl-3C technology captures single-cell chromosomal conformations and DNA methylation.
• Limited methyl-3C datasets exist compared to single-cell Hi-C data.
• A computational method is needed to predict methylation from single-cell Hi-C data.

Purpose of the Study:

• To develop a computational tool for predicting base-pair-specific DNA methylation levels from single-cell Hi-C data.
• To leverage existing single-cell Hi-C datasets for methylation analysis.
• To enable a deeper understanding of epigenomic variability at the single-cell level.

Main Methods:

• Developed scHiMe, a graph transformer model.
• Integrated single-cell Hi-C data and DNA nucleotide sequences as input.
• Benchmarked scHiMe on human genome promoters, promoter-exon-intron regions, and random genomic regions.

Main Results:

• scHiMe accurately predicted base-pair-specific methylation levels.
• High consistency was observed between predicted and methyl-3C-detected methylation levels.
• Predicted methylation levels enabled accurate cell type classification, reflecting cell-to-cell variability.

Conclusions:

• scHiMe effectively predicts single-cell DNA methylation from single-cell Hi-C data.
• The tool captures cell-to-cell epigenetic variability present in Hi-C data.
• scHiMe provides a valuable computational resource for epigenomic studies.