Ctrl+K

hicluster embedding

hicluster embedding#

This step will generate cell by 100kb-pair contact matrix which can be used for cell embedding

Command Docs#

usage: hicluster embedding [-h] --cell_table_path CELL_TABLE_PATH --output_dir
                           OUTPUT_DIR [--chrom_size_path CHROM_SIZE_PATH]
                           [--dim DIM] [--dist DIST] [--resolution RESOLUTION]
                           [--scale_factor SCALE_FACTOR] [--cpu CPU]
                           [--norm_sig] [--save_model] [--save_raw]

optional arguments:
  -h, --help            show this help message and exit
  --chrom_size_path CHROM_SIZE_PATH
                        Path to UCSC chrom size fileContain all the
                        chromosomeinformation in two tab-separated columns:1.
                        chromosome name, 2. chromosome length. No header
                        (default: None)
  --dim DIM
  --dist DIST
  --resolution RESOLUTION
                        Resolution for embedding.Consistent with resolution of
                        imputed contact files (default: 100000)
  --scale_factor SCALE_FACTOR
  --cpu CPU
  --norm_sig
  --save_model
  --save_raw

required arguments:
  --cell_table_path CELL_TABLE_PATH
                        Contain all the imputed contact files information in
                        twotab-separated columns: 1. cell_uid, 2. file_path.
                        No header (default: None)
  --output_dir OUTPUT_DIR
                        Path to the output directory of the embedding output
                        (default: None)

Command Example#

Here is an example to generate chrom by chrom contact matrix.

hicluster embedding \
    --cell_table_path cell_table.tsv \
    --output_dir dataset/embedding \
    --dim 50 \
    --dist 1000000 \
    --resolution 100000 \
    --scale_factor 100000 \
    --norm_sig \
    --save_raw \
    --cpu 20

Command Breakdown#

--cell_table_path cell_table.tsv

Specify the file paths of the cool files after imputtaion in this line(e.g. /home/qzeng_salk_edu/project/aging/230711_m3C/impute/100K/chunk0/AMB_220712_18mo_12D_13B_2_P4-1-I15-G2.cool). Here is an example of what the contact table looks like:

cell_1 absolute_cool_file_path_1
cell_2 absolute_cool_file_path_2
cell_3 absolute_cool_file_path_3

The first column indicates the cell name (e.g. AMB_220712_18mo_12D_13B_2_P4-1-I15-G2) whereas the second column indicates the cool file path of the cell. Make sure the two parts are separated by a tab; also make sure the file has no header.

--output_dir dataset/embedding

This is the path to the output folder for your output files and you don’t need to create the folder before running. This command will save output files in output_dir/raw and output_dir/decomp.

The output_dir/raw: npz files of each chromosome, which contains the information of cell x 100kb-pair contacts matrix (e.g. chr1.npz).

The folder “output_dir/decomp”: the concatenated contacts of all chromosomes after performing singular value decomposition (SVD) on each chromosome. (total_chrom_decomp_concat.npz). Additionally, the concatenated decomposition matrices of all chromosomes are further subjected to another round of SVD. (total_decomp.npz)

For information regading loading npz files see [here] https://numpy.org/doc/stable/reference/generated/numpy.savez.html

Contents