iProphIT

A deep learning approach that identifies the inducible activity of prophages from their DNA sequences.

Requirements

System and software requirements:

• Linux
• Python 3.x (Any Python version compatible with PyTorch，Tested with Python 3.12.)
• biopython
• numpy
• pytorch (If you want to enable GPU acceleration, please install the appropriate GPU-enabled PyTorch version from the official PyTorch website.)

Installation

1. You only need to download iProphIT-classifier.py and iProphIT_model-v1.pth into your working directory.
(iProphIT_model-v1.pth website: (https://doi.org/10.5281/zenodo.17605580)

2. Create a conda environment and install required packages:

conda create -n iprophit python=3.12
conda activate iprophit
conda install -c conda-forge biopython numpy
conda install pytorch

Run iProphIT

1. Download iProphIT-classifier.py， iProphIT_model-v1.pth and put them in your working path.

2. Run iProphIT-classifier.py

python iProphIT-classifier.py -i test_iProphIT.fasta -m iProphIT_model-v1.pth -o ./Result.tsv -t 16

Usage

usage: iProphIT-classifier.py [-h] -i INPUT [-m MODEL] [-o OUTPUT] [-t THREADS] [-b BATCH_SIZE]

options:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Path to the input FASTA file (required)
  -m MODEL, --model MODEL
                        Path to the trained model file (default: ./iProphIT_model-v1.pth)
  -o OUTPUT, --output OUTPUT
                        Output TSV file path (default: ./Result.tsv)
  -t THREADS, --threads THREADS
                        Number of CPU threads for DataLoader (default: 4)
  -b BATCH_SIZE, --batch_size BATCH_SIZE
                          Batch size for prediction (default: 4). Larger values accelerate inference.
                          GPU: Increase for speedup until CUDA OOM, then reduce.
                          CPU: Can use larger values due to more RAM.

Typical output

• Find result in Result.tsv

  ID	Predict	Confidence
  prophage1	active	0.9981
  prophage2	dormant	0.9221

• Explanation
  1.ID is the content of the description line in the genome file.
  2.Predict is the result of identification (active->inducible prophage, dormant->non-inducible prophage).

Using testing data

genome file: OY731326.1 and OY731419.1,
source: Dahlman S. et al., Nature (2025), https://doi.org/10.1038/s41586-025-09614-7

• Run iProphIT-classifier.py

python iProphIT-classifier.py -i test_iProphIT.fasta -m iProphIT_model-v1.pth -o ./Result.tsv -t 16

• Output of the test

ID	Predict	Confidence
OY731326.1	active	0.9989
OY731419.1	active	0.9927

Notes

• Input can accept genome files in formats such as .fasta, .fa, .fna, etc.
• iProphIT will automatically use the GPU if available, as long as you have installed a PyTorch version with CUDA support.

Copyright

Hongbo Zhang, Chen Liu, Hanpeng Liao, Fujian Provincial Key Laboratory of Soil Environmental Health and Regulation, College of Resources and Environment, Fujian Agriculture and Forestry University, Fuzhou, 350002, China.