Chromatin Interaction Neural Network (ChINN) - Modified Version

Adapted for Comparative Analysis with HiCDiffusionLooping

This version of ChINN has been modified for research purposes, specifically to compare with the HiCDiffusionLooping model. Please note that it deviates from the original implementation of ChINN.

Chromatin Interaction Neural Network (ChINN) only uses DNA sequences of the interacting open chromatin regions. ChINN is able to predict CTCF-, RNA polymerase II- and HiC- associated chromatin interactions between open chromatin regions.

ChINN was able to identify convergent CTCF motifs, AP-1 transcription family member motifs such as FOS, and other transcription factors such as MYC as being important in predicting chromatin interactions.

ChINN also shows good across-sample performances and captures various sequence features that are predictive of chromatin interactions.

Environment

git clone https://github.com/SFGLab/chinn

# enter the directory. Use the repository's root directory as working directory.
cd chinn

# setup python env
conda create --name chinn python=3.8.20
pip install -r requirements.txt torch==1.9.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html
conda activate chinn

download singularity images for data preprocessing

singularity pull tools.sif library://m10an/genomics/tools
singularity pull samtools.sif library://millironx/default/samtools

Dataset

# https://ftp.1000genomes.ebi.ac.uk/vol1/ftp/technical/reference/GRCh38_reference_genome/GRCh38_full_analysis_set_plus_decoy_hla.fa

singularity exec samtools.sif samtools faidx GRCh38_full_analysis_set_plus_decoy_hla.fa
singularity exec samtools.sif samtools faidx GRCh38_full_analysis_set_plus_decoy_hla.fa $(seq -f 'chr%g' 1 22) chrX chrY > out_dir/hg38.fa
singularity exec samtools.sif samtools faidx out_dir/hg38.fa

data
├── gm12878_ctcf/hg38_lifted  # uplifted version of original data
│   ├── pairs.bedpe
│   ├── peaks.bed
│   └── dnase.bed
└── gm12878_ctcf/hg38         # data used for training HiCDiffusionLooping
    ├── 4DNFI9SL1WSF.bedpe    # https://data.4dnucleome.org/files-processed/4DNFI9SL1WSF/
    ├── peaks.bed             # https://data.4dnucleome.org/files-processed/4DNFIV1N7TLK/ and https://data.4dnucleome.org/files-processed/4DNFIW1VY2CW
    └── ENCFF759OLD.bed       # https://www.encodeproject.org/files/ENCFF759OLD/

# gm12878_ctcf/hg38/peaks.bed is combination of two peaks:
# cat 4DNFIV1N7TLK.bed 4DNFIW1VY2CW.bed | sort -k1,1 -k2,2n | uniq > peaks.bed

Note: How up lifting of original hg19 data was made described in data/gm12878_ctcf/hg38_lifted directory

Note: to run the scripts, use the root directory of this repository as the working directory.

Pipeline

# prepare output directory
mkdir out_dir

singularity exec samtools.sif samtools faidx out_dir/hg38.fa
export PYTHONPATH=$PWD

bash preprocess/pipe.sh data/gm12878_ctcf/hg38/4DNFI9SL1WSF.bedpe \
                        data/gm12878_ctcf/hg38/ENCFF759OLD.bed \
                        data/gm12878_ctcf/hg38/peaks.bed \
                        gm12878_ctcf \
                        out_dir

python data_preparation.py -m 1000 -e 500 \
                      --pos_files out_dir/gm12878_ctcf.clustered_interactions.both_dnase.bedpe \
                      --neg_files out_dir/gm12878_ctcf.neg_pairs_5x.from_singleton_inter_tf_random.bedpe \
                      -g out_dir/hg38.fa \
                      -n gm12878_ctcf_distance_matched -o out_dir

python train_distance_matched/train_distance_matched.py \
                       out_dir/gm12878_ctcf_distance_matched_singleton_tf_with_random_neg_seq_data_length_filtered \
                       gm12878_ctcf_model \
                       out_dir

python data_preparation.py -m 1000 -e 500 \
    --pos_files out_dir/gm12878_ctcf.clustered_interactions.both_dnase.bedpe \
    --neg_files out_dir/gm12878_ctcf.neg_pairs_5x.from_singleton_inter_tf_random.bedpe \
                out_dir/gm12878_ctcf.extended_negs_with_intra.bedpe \
    -g out_dir/hg38.fa \
    -n gm12878_ctcf_extended -o out_dir

for i in train valid test; 
do 
  python generate_factor_output.py \
                  out_dir/gm12878_ctcf_model.model.pt \
                  out_dir/gm12878_ctcf_distance_matched_singleton_tf_with_random_neg_seq_data_length_filtered_${i}.hdf5 \
                  gm12878_ctcf_${i} \
                  out_dir;
done

python train_extended/train_extended.py out_dir gm12878_ctcf out_dir/

python predict.py -m out_dir/gm12878_ctcf_model.model.pt -c out_dir/gm12878_ctcf_depth6.gbt.pkl --data_file out_dir/gm12878_ctcf_extended_singleton_tf_with_random_neg_seq_data_length_filtered_test.hdf5 --output_pre out_dir/gm12878_ctcf_extended_test -d

python predict_bedpe.py -m out_dir/gm12878_ctcf_model.model.pt \
                -c out_dir/gm12878_ctcf_depth6.gbt.pkl \
                --pos_files data/positives.bedpe \
                --neg_files data/negatives.bedpe \
                -g out_dir/hg38.fa \
                --min_size 1000 -e 500 -d \
                --output_pre out_dir/hg38