Goal:

Reconstruct chromatin compartments from affordable single-cell epigenetic data.

Key questions:

1. How to overcome the sparsity of epigenetic data?

2. How to model long-range chromatin interactions?

3. How to quantify chromatin conformation switches?

Approach:

1. Metacell construction

2. Correlation graph

3. Embedding alignment

Bin:

fixed-length genome regions (chr1:1-1,000,000).

Metacell:

group of similiar cells constructed to mitigate sparsity.

Problem Formulation:

Given a scATAC-seq matrix F (N bins, c metacells), we determine a binary vector H in {0,1}^N to represent the compartment type.

​

scENCORE: leveraging single-cell epigenetic data to

predict chromatin conformation using graph embedding

Briefings in Bioinformatics, 2024

Zhang Lab

Code

Data

scENCORE recovers Hi-C

chromatin map from bulk ATAC-seq

scENCORE reconstructs A/B compartments validated by cell-specific Hi-C data

Chromatin Conformation Benchmark

scENCORE identifies key compartment switching events between different cell types

scENCORE highlights extensive cell-type-specific chromatin re-structuring events in brain disorders

scENCORE & scHi-C