Please follow this to install and configure RefHiC.
We suggest users run RefHiC-SR on GPU.
You can run RefHiC-SR on CPU, but it is much slower than on GPU. Model training on CPU is almost impossible.
NB: Our reference panel is for hg38 at 5kb. You can follow the instruction to create your own reference panel.
Enhancing chr1 in test.mcool:
refhic sr pred test.mcool output.bedpe --chrom chr1
It outputs enhanced contact maps in a bedpe file. ### output format It contains tab separated fields as follows:
Chr1 Start1 End1 Chr2 Start2 End2 enhancedIF IF
| Field | Detail |
|---|---|
| Chr1/Chr2 | chromosome names |
| Start1/Start2 | start genomic coordinates |
| End1/End2 | end genomic coordinates (i.e. End1=Start1+resol) |
| enhancedIF | RefHiC-SR enhanced Interaction frequency |
| IF | Interaction frequency |
NB: RefHiC-SR comes with a pretrained model. You don't need to train your own model in most case.
- Preprocessing:
refhic util ceptraindata comma_separated_list_of_downsampled_mcool_files full_coverage.mcool val/hicMatrix --chrom validation_chrs
refhic util ceptraindata comma_separated_list_of_downsampled_mcool_files full_coverage.mcool train/hicMatrix --chrom training_chrs
refhic util ceptraindata comma_separated_list_of_downsampled_mcool_files full_coverage.mcool test/hicMatrix --chrom test_chrs
- Training:
refhic sr train TRAINDATA VALDATA PREFIX
You can use scripts in https://doi.org/10.5281/zenodo.7761968 to reproduce our experiments.
If you use RefHiC in your work, please cite our paper:
Zhang, Y., Blanchette, M. Reference panel guided super-resolution inference of Hi-C data
A GitHub issue is preferable for all problems related to using RefHiC-SR.
For other concerns, please email Yanlin Zhang or Mathieu Blanchette ([email protected], [email protected]).