Phynteny: Synteny-based annotation of bacteriophage genes

Approximately 65% of all bacteriophage (phage) genes cannot be attributed a known biological function. Phynteny uses a long-short term memory model trained on phage synteny (the conserved gene order across phages) to assign hypothetical phage proteins to a PHROG category.

Phynteny is still a work in progress and the LSTM model has not yet been optimised. Use with caution!

NOTE This version of Phynteny will only annotate phages with 120 genes or less due to the architecture of the LSTM. We aim to adjust this in future versions.

Dependencies

Phynteny installation requires Python 3.8 or above. You will need the following python dependencies to run Phynteny and its related support scripts. The latest tested versions of the dependencies are:

• python - version 3.10.0
• sklearn - version 1.2.2
• biopython - version 1.81
• numpy - version 1.21.0 (Windows, Linux, Apple Intel), version 1.24.0 (Apple M1/M2)
• tensorflow - version 2.9.0 (Windows, Linux, Apple Intel), tensorflow-macos version 2.11 (Apple M1/M2)
• pandas - version 2.0.2
• loguru - version 0.7.0
• click - version 8.1.3
  

We recommend GPU support if you are training Phynteny. This requires CUDA and cuDNN:

• CUDA toolkit - version 11.2
• cuDNN - version 8.1.1

Installation

Currently Phynteny can be installed from this repository

git clone https://github.com/susiegriggo/Phynteny.git --branch main --depth 1 
cd Phynteny 
pip install . 

Install Models

Once you've installed Phynteny you'll need to download the pre-trained models

install_models.py 

If you would like to specify a particular location to download the models run

install_models.py -o <path/to/database_dir>

If for some reason this does not work. you can download the pre-trained models from Zenodo and untar in a location of your choice.

Usage

Phynteny takes a genbank file containing PHROG annotations as input. If you phage is not yet in this format, pharokka can take your phage (in fasta format) to a genbank file with PHROG annotations. Phynteny will then return a genbank files and a table containing the details of the predictions made using phynteny. Each prediction is accompanied by a 'phynteny score' which ranges between 1-10 and a recalibrated confidence score.

Reccomended

phynteny test_data/test_phage.gbk  -o test_phynteny

Custom

If you wish to specify your own LSTM model, run:

phynteny test_phage.gbk -o test_phage_phynteny -m your_models -t confidence_dict.pkl 

Details of how to train the phynteny models and generate confidence estimates is detailed below.

Train Phynteny

Phynteny has already been trained for you on a dataset containing over 1 million prophages! If you feel inclined to generate your own Phynteny model using your own dataset, instructions and training scripts are provided here.

Performance

Coming soon: Notebooks demonstrating the performance of the model

Bugs and Suggestions

If you break Phynteny or would like to make any suggestions please open an issue or email me at susie.grigson@flinders.edu.au

Wow! How can I cite this incredible piece of work?

The Phynteny manuscript is currently in preparation. In the meantime, please cite Phynteny as:

Grigson, S. R.,  Mallawaarachchi, V., Roach, M. R., Papudeshi, B., Bouras, G., Decewicz, P., Dinsdale, E. A. & Edwards, R. A. (2023). Phynteny: Synteny-based annotation of phage genomes. DOI: 10.5281/zenodo.8128917

If you use pharokka to annotate your phage before using Phynteny please cite it as well:

Bouras, G., Nepal, R., Houtak, G., Psaltis, A. J., Wormald, P. J., & Vreugde, S. (2023). Pharokka: a fast scalable bacteriophage annotation tool. Bioinformatics, 39(1), btac776.