ContigNet: Phage-bacteria contig interaction prediction with convolutional neural network

The paper has been published at https://doi.org/10.1093/bioinformatics/btac239

Version: 1.0.1

Authors: Tianqi Tang, Shengwei Hou, Jed Fuhrman, Fengzhu Sun

Maintainer: Tianqi Tang tianqit@usc.edu

Description

This is the repository containing the software ContigNet and related scripts for the paper “Phage-bacteria contig interaction prediction with convolutional neural network”.

ContigNet is a deep learning based software for phage-host contig interaction prediction. Traditional methods can work on contigs however the performance is poor. Existing Deep learning based methods are not able to solve the particular question regarding interaction prediction between two contigs.

Installation

The software is available at PyPI and Bioconda now, to install, run

pip install ContigNet

conda install -c bioconda contignet

To install the software from source, download and enter the repository by

git clone https://github.com/tianqitang1/ContigNet
cd ContigNet

To install required dependencies a Anaconda or Miniconda installation is recommended for managing virtual environments. After a conda distribution is installed, create and activate a conda virtual environment with the following commands

conda create --name ContigNet
conda activate ContigNet
pip install .

Usage

usage: ContigNet [-h] [--host_dir HOST_DIR] [--virus_dir VIRUS_DIR]
                    [--output, -o OUTPUT] [--cpu]

ContigNet, a deep learning based phage-host interaction prediction tool

optional arguments:
  -h, --help            show this help message and exit
  --host_dir HOST_DIR   Directory containing host contig sequences in fasta
                        format (default: demo/host_fasta)
  --virus_dir VIRUS_DIR
                        Directory containing virus contig sequences in fasta
                        format (default: demo/virus_fasta)
  --output, -o OUTPUT   Path to output file (default: result.csv)
  --cpu                 Force using CPU if specified (default: False)

Examples

Test new contigs

Suppose the phage and host sequences are stored in phage and host directories respectively, running

ContigNet --host_dir host --virus_dir phage

and the likelihood of each phage interacting with each host will be output to result.csv.

For Windows machine, run

python -m ContigNet --host_dir host --virus_dir phage

Browse training directory for the instructions of running the training and testing process for the paper.

Copyright and License Information