ChIA-Rep 2.1.0

A package for measuring reproducibility of ChIA-PET data.

ChIA-Rep

A Python package for assessing the reproducibility of ChIA-PET datasets.

Methods Overview

Preprocessing

• Keep a loop if one or both of the anchors overlap with a peak, where the peak file may be filtered to keep the top x peaks
• Weigh the loop PET count by the anchor intensity

Graph representation and comparison

• For a non-overlapping window, bin the loops into bins of fixed size
• Create an adjacency matrix for each window, where nodes are bins and edges are sums of weighted loop PET counts
• Convert each adjacency matrix into a probability vector by reading row-by-row
• Compute the Jensen-Shannon divergence or the Earth Mover's Distance (EMD) between two probability vectors
• Transform each value to be between -1 (dissimilar) and 1 (similar)
• Take a weighted average of values for all windows in a chromosome
• Genome-wide reproducibility value is computed by averaging over all chromosomes

Example

• Given two ChIA-PET datasets, create adjacency matrices A1 and A2

Adjacency matrix A1

	bin1	bin2	bin3	bin4
bin1	3	2	0	1
bin2		1	5	3
bin3			10	9
bin4				20

Adjacency matrix A2

	bin1	bin2	bin3	bin4
bin1	4	5	1	4
bin2		3	2	3
bin3			7	9
bin4				27

Probability vectors p_A1 and p_A2

• p_A1 = (0.06, 0.05, 0, 0.02, 0.02, 0.009, 0.06, 0.19, 0.17, 0.37)
• p_A2 = (0.06, 0.08, 0.02, 0.06, 0.05, 0.03, 0.05, 0.11, 0.14, 0.42)

Results

• ChIA-Rep can clearly distinguish between replicates and non-replicates
• Generally, replicates have positive values and non-replicates have negative values
• Can take 0 as a threshold to determine the similarity

Usage

Dependencies:

# Automatically included when pip installing
install_requires=[
	'numpy>=1.17.0',  
	'scipy>=1.3.1',  
	'prettytable>=0.7.2',  
	'pybedgraph>=0.5.31',  
	'matplotlib>=3.1.1']

Installation:

git clone https://github.com/c0ver/chia_rep.git    
pip3 install chia_rep/

pip3 install chia_rep

Input data and compare:

Example script is included in example/script.py.

Testing

pytest  # Runs the tests in test/test_reproducibility.py

Contact

Contact Minji ([e-mail]) for general questions, and report software issues in the Issues page.