Single-cell and spatial Hi-C

Single-cell Hi-C (and Dip-C, scMicro-C, sci-Hi-C) gives each cell a sparse list of contacts — pairs of loci that were close in that nucleus; the multi-omics variants add the cell’s transcriptome or accessibility (HiRES, GAGE-seq, scHiCAR, Paired / Droplet Hi-C), and spatial Hi-C measures contacts per spot of a tissue section. Contacts are not coordinates: U-Chrom keeps them as contact maps linked to the cells, and computes coordinates from them when you want structures.

Step	API
read contacts: .pairs, phased pairs (hickit / Dip-C), the Stevens 2017 / NCC files	uchrom.io.read_pairs, read_phased_pairs; the reconstruction functions read the files directly
link per-cell contact maps (.scool) and pseudo-bulk maps (.cool / .mcool) to the cells	cd.link_scool, cd.link_cool; cd.load_linked_scool(cell=...)
link RNA / ATAC matrices (.h5ad, .h5mu)	cd.link_anndata, cd.link_mudata; cd.linked_adata
3-D structures of one cell, an ensemble of models (native engine, CPU or GPU)	EMber: uchrom.recon.sc.reconstruct_ember, uc.tl.reconstruct_sc; NucDynamics: reconstruct_nucdyn
contacts of a structure (to compare with the input)	uchrom.io.count_contacts
cell features from contact maps; Higashi / FastHigashi	uchrom.emb.scool_cell_features, schicluster_features; uchrom.emb.higashi (extra emb)
spots of a section in the tissue	cd.set_cell_positions, cd.cell_positions, cd.link_spatialdata

The atlas holds single-cell and spatial Hi-C datasets ready to open — HiRES, dscHi-C, GAGE-seq, Droplet and Paired Hi-C, scHiCAR, Uni-C, Stevens 2017 with its published structures, spatial Hi-C, spatial ATAC-Hi-C and Hi-C-RNA sections (datasets).

Tutorials

• Single-cell 3-D genome reconstruction: NucDynamics and EMber
• Contact-map cell embedding with FastHigashi

The scHiCAR part of cell embeddings embeds the same cells by their contact maps, their RNA and their accessibility.

Guides

• Single-cell reconstruction: methods and parameters

Then

Reconstructed models are coordinates like traced ones: distance maps and contact frequencies of the models (per-locus features), comparisons with imaging of the same cell type, and the web browser, which shows each cell’s models next to its own contact map. Linked contact maps, pseudo-bulk maps per cell type and embedded copies for sharing: data model & storage.

API: uchrom.recon, uchrom.emb, uchrom.io.