Conditional diffusion for single-cell Hi-C data imputation and enhancement with scHiC-Diff.

scHiC-Diff improves recovery of chromatin interaction patterns in single-cell Hi-C data by modeling contact matrices with a diffusion transformer.

• Why it matters: Single-cell Hi-C data are limited by technical dropouts that obscure true 3D genome structures, hindering accurate analysis of cell-to-cell variability.
• What they did: The authors developed scHiC-Diff, a diffusion-based framework that uses a forward perturbation process and a reverse denoising process, modeling contact matrices as masked images with a Diffusion Transformer, tested on multiple datasets.
• The result: scHiC-Diff outperforms existing methods in recovering chromatin interactions and enhances downstream analyses, including chromatin organization reconstruction and loop detection, in simulated and real datasets.