Example dataset

We have provided an example dataset to test your TwinC installation. You can follow this dataset's format to prepare your data and label files for training TwinC in a new tissue or cell line. 

cd twinc/
wget https://zenodo.org/records/15839236/files/TwinC_example_V1.tar.gz

After downloading, you can uncompress the file, and it will populate the data folder. 

tar -xvzf TwinC_example_V1.tar.gz

This dataset contains:

  1. hg38.no_Y_MT.fa: Fasta file for the human genome containing chromosomes 1-22 and X.

  2. hg38.no_Y_MT.fa.fai: Index for the fasta file.

  3. hg38.no_Y_MT.memmap: One-hot-encoded memory map for the human genome.

A -> [1, 0, 0, 0]
C -> [0, 1, 0, 0]
G -> [0, 0, 1, 0]
T -> [0, 0, 0, 1]
  1. train_labels.txt: Training labels. A text file containing genomic coordinates from chromosomes A and B, a contact label (0 -> no contact, 1 -> contact) and KR-normalized contact frequency. The columns in the files are as follows:
1. chrA_Name: Name of the first chromosome.
2. chrA_Start: Start coordinate in chrA.
3. chrA_End: End coordinate in chrA.
4. chrB_Name: Name of the second chromosome.
5. chrB_Start: Start coordinate in chrB.
6. chrB_End: End coordinate in chrB.
7. Contact_Label: Whether the two loci are in contact (0->No, 1->Yes).
8. Contact_Frequency: KR-Normalized contact frequency.

  1. val_labels.txt: Validation labels.

  2. test_labels.txt: Test labels.