Deep learning tool for protein orthologous group predictions
Project description
DeepNOG: protein orthologous groups prediction
Predict orthologous groups of proteins on CPUs or GPUs with deep networks. DeepNOG is both faster and more accurate than assigning OGs with HMMER.
The deepnog
command line tool is written in Python 3.7+.
Current version: 1.1.0
Installation guide
The easiest way to install DeepNOG is to obtain it from PyPI:
pip install deepnog
Alternatively, you can clone or download bleeding edge versions from GitHub and run
pip install /path/to/DeepNOG
If you plan to extend DeepNOG as a developer, run
pip install -e /path/to/DeepNOG
instead.
Usage
DeepNOG can be used through calling the above installed deepnog
command with a protein sequence file (FASTA).
Example usages:
- deepnog proteins.faa
- OGs prediction of proteins in proteins.faa will be written into out.csv
- deepnog proteins.faa --out prediction.csv
- Write into prediction.csv instead
- deepnog proteins.faa --tab
- Instead of semicolon (;) separated, generate tab separated output-file
The individual models for OG predictions are not stored on GitHub or PyPI,
because they exceed file size limitations (up to 200M).
deepnog
automatically downloads the models, and puts them into a
cache directory (default ~/deepnog_data/
). You can change this directory
by setting the DEEPNOG_DATA
environment variable.
For help and advanced options, call deepnog --help
,
or see the user & developer guide.
File formats supported
Preferred: FASTA (raw or gzipped)
DeepNOG supports protein sequences stored in all file formats listed in https://biopython.org/wiki/SeqIO but is tested for the FASTA-file format only.
Databases supported
- eggNOG 5.0, taxonomic level 1 (root level)
- eggNOG 5.0, taxonomic level 2 (bacteria level)
- (for additional level, please create an issue)
Neural network architectures supported
- DeepEncoding (=DeepNOG in the research article)
Required packages (and minimum version)
- PyTorch 1.2.0
- NumPy 1.16.4
- pandas 0.25.1
- Biopython 1.74
- tqdm 4.35.0
- pytest 5.1.2 (for tests only)
Acknowledgements
This research is supported by the Austrian Science Fund (FWF): P27703, P31988, and by the GPU grant program of Nvidia corporation.
Citation
A research article is currently in preparation.
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distributions
Hashes for deepnog-1.1.0-py3-none-win_amd64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 1287f9681fea0314c2bdd3076343db25b8e3d7996db76ac36344242c75dd1ef8 |
|
MD5 | a75b5a3a8033c686542b37e28bfd2a86 |
|
BLAKE2b-256 | 2ca6fbc6e997307f0d5586cda030a16ce54779857ffcd64e24d13336a5b45fa1 |