netbalance 0.2.0

Balance and visualize multi-cluster interaction networks

netbalance

netbalance is a Python library and CLI tool for balancing and visualizing multi-cluster interaction (association) networks.

Installation

pip install netbalance

Quick start

Balancing an interaction network using entity-balanced sampling:

cat data/virus-host.csv | netbalance --seed 0 --max-iter 1000

Balancing an interaction network using uniform random sampling with a 1:1 negative-to-positive ratio:

cat data/virus-host.csv | netbalance -m balanced --negative-ratio 1.0

Visualizing a bipartite network:

cat data/virus-host.csv | netbalance viz -o virus-host.png

Data format

All inputs are CSVs where the first N columns are entity names (one per cluster) and the last column is a binary label (1 = positive interaction, 0 = negative):

A,B,interaction
A1,B2,1
A1,B4,0

The viz subcommand requires exactly 2 entity columns (bipartite).

Python API

If you prefer to use the Python API instead of the CLI, see the example notebook.

Example usage

Balancing and visualizing a virus-host interaction network

The following cammand uses the entity-balanced sampling method to balance a virus-host interaction network, then visualizes the balanced network as a bipartite graph.

cat data/virus-host.csv | netbalance --seed 0 --max-iter 20000 --delta 0.001| netbalance viz -o virus-host.png

The resulting graph is shown below:

Entity-balanced sampling without the degree-ratio guided intialization

cat data/virus-host.csv | netbalance --seed 0 --max-iter 20000 --delta 0.001 --no-heuristic