This repo executes a minimally modified version of the Activity-By-Contact (ABC) model presented in https://www.nature.com/articles/s41588-019-0538-0#Sec1 for predicting enhancer promoter functional contacts based off of Hi-C and ChIPseq data. The modifications are due to lack of good datasources for ATAC-seq data in Mbd3KO cell lines.

Requirements

• numpy
• pandas
• itertools
• cython
• straw (https://github.com/aidenlab/straw/wiki/Python)
• tqdm (https://tqdm.github.io/)
• pyBigWig (https://github.com/deeptools/pyBigWig)

Installation

From the main directory run:

python cython_setup.py build_ext --inplace

Notes

I use 0-indexing of columns. This is relevant, for example when running ABC_score.py

Required input files:

• Transcripts (TSS) (.tsv with columns ['chrom','start','end','strand','id'] detailing the genomic positions of each TSS we are profiling) Optionally this file can contain further columns detailing the expression of each transcript profiled using RNA-seq
• H3K27ac ChIPseq file(s)
  • narrowPeak format files are used to identify regulatory element regions
  • bigwig or narrowPeak format files can be used to assess regulatory element 'strength'

Optional input files

• Hi-C data (.hic format generated using either Juicer https://github.com/aidenlab/juicer/wiki or Pre https://github.com/aidenlab/juicer/wiki/Pre. NOTE: This algorithm assumes sufficient unique contacts to bin the genome at 5kb resolution)
• H3K4me3 ChIpseq data (narrowPeak or bigwig format) can be used to filter out TSS sites without active promoter sites (see make_promoter_info.py)

scripts

• make_promoter_info.py - uses the TSS inputs to create promoter regions.
• make_candidate_regions.py - Generates putative candidate elements and their strengths using the promoter_info produced using make_promoter_info.py as well as any number of H3K27ac narrowPeak files.
• linear_links.py - creates master list of possible enhancer promoter links by linking those promoters with regulatory that fall within some linear threshold distance along the chromatin backbone. Defaults to 5Mb.
• ABC_score.py - inputting ~KW-normalised per-chromosome sparse hi-C matrices, promoter info, enhancer info, promoter expression and enhancer k27ac scores, calculates and stores a file detailing enhancer-promoter links with an ABC score above a given threshold from the master list of linear enhancer-promoter links.

suggested directory structure

/topdir/
/topdir/make_promoter_info.py
/topdir/make_candidate_regions.py
/topdir/linear_links.py
/topdir/ABC_score.py


/topdir/utils/


/topdir/data/
/topdir/data/processed/
/topdir/data/processed/links/

/topdir/data/raw/
/topdir/data/raw/input.hic
/topdir/data/raw/transcripts.tsv
/topdir/data/raw/tracks/k27ac/{k27ac narrowPeak files}
/topdir/data/raw/tracks/k4me3/{k4me3 narrowPeak files}

Example Usage

With the directory structure above an example usage would be:

python make_promoter_info.py --transcripts /topdir/data/raw/transcripts.tsv --outpath /topdir/data/processed/ --k4 /topdir/data/raw/tracks/k4me3/example_k4_track_1.narrowPeak /topdir/data/raw/tracks/k4me3/example_k4_track_2.narrowPeak --chromosomes 1 8 10

python make_candidate_elements.py --promoters /topdir/data/processed/promoter_info.tsv --k27ac_files /topdir/data/raw/tracks/k27ac/example_k27_track_1.narrowPeak /topdir/data/raw/tracks/k27ac/example_k27_track_2.narrowPeak --outpath /topdir/data/processed/ --chromosomes 1 8 10

python linear_links.py --regions1 /topdir/data/processed/promoter_info.csv --regions2 /topdir/data/processed/candidate_elements.csv --outpath data/processed/links/ --region1names promoter --region2names regulatory_element --distance_thresh 5000

python ABC_score.py --promoters /topdir/data/processed/promoter_info.csv --rnaexpression_column 4 --regelements /topdir/data/processed/candidate_elements.csv --k27ac /topdir/data/raw/tracks/example_k27_track_1.bigwig --chip_format bigwig --linearlinks /topdir/data/processed/links/linear_promoter_regulatory_element_links.npz --contacts /topdir/data/raw/input.hic --binsize 5000 --normalisation KR --threshold 0.05 --outpath /topdir/data/processed/links --outname my_links --tsv 1