DeepMicroClass, a deep learning based contig prediction tool
Project description
DeepMicroClass
This is the repository for DeepMicroClass, a deep learning based method that classifies metagenomic contigs into five sequence classes, e.g., viruses infecting prokaryotic or eukaryotic hosts, eukaryotic or prokaryotic chromosomes.
The paper corresponding to this repository has been published at NAR Genomics and Bioinformatics. The exact code version used in the paper can be found at . If you find DeepMicroClass useful, please consider to cite:
- Shengwei Hou, Tianqi Tang, Siliangyu Cheng, Yuanhao Liu, Tian Xia, Ting Chen, Jed A Fuhrman, Fengzhu Sun, DeepMicroClass sorts metagenomic contigs into prokaryotes, eukaryotes and viruses, NAR Genomics and Bioinformatics, Volume 6, Issue 2, June 2024, lqae044, https://doi.org/10.1093/nargab/lqae044
Please direct any questions to Dr. Fengzhu Sun.
Installation
Consider using virtualenv
or conda
to create a virtual environment for clean environment. The following workflow works on our tested server with GPU Nvidia Quadro RTX 6000 and CUDA version 10.1. Since DeepMicroClass is using Pytorch, it is possible that our Pytorch version is not compiled with the version of your CUDA drive. Therefore, we suggest you install Pytorch compiled with your own device.
You can find the Pytorch that has been compiled with your version of the CUDA driver from Previous PyTorch Versions | PyTorch. Most Pytorch versions are available only for specific CUDA versions. For example, pytorch=1.0.1 is not available for CUDA=9.2
Or, you can update your CUDA driver. To query the CUDA version that matches your driver, just open a terminal and run:
nvidia-smi
To check the CUDA version in your environment variables, which is the CUDA version you are probably using, just run:
nvcc -V
To use GPU (CPU is also optional)
Create a new environment:
conda create -n DeepMicroClass_python=3.10 python=3.10
conda activate DeepMicroClass_python=3.10
Check if pip
is under the new environment:
which pip
Your pip
should be under the new environment:
/home/xiatian/anaconda3/envs/DeepMicroClass_python=3.10/bin/pip
If not, install pip
under the new environment:
conda install pip
Then check it again.
After that, install DeepMicroClass by running:
pip install DeepMicroClass
Then, install Pytorch packages suitable for CUDA=11.6 and python=3.10.13.
conda install pytorch==1.13.1 torchvision==0.14.1 torchaudio==0.13.1 pytorch-cuda=11.6 -c pytorch -c nvidia
These versions are compatible with some Python versions like 3.10.13 and 3.9.18. If you were going to install Pytorch packages suitable for your CUDA, you should try other versions of Python in the first step.
If you encountered errors like:
LibMambaUnsatisfiableError: Encountered problems while solving:
- package torchvision-0.14.1-py310_cpu requires python >=3.10,<3.11.0a0, but none of the providers can be installed
- package torchaudio-0.13.1-py310_cpu requires python >=3.10,<3.11.0a0, but none of the providers can be installed
Could not solve for environment specs
The following packages are incompatible
...
You may need to try other versions of Python.
To use CPU ONLY
CPU mode works for python=3.9-3.12
. The workflow is likely the same.
conda create -n DeepMicroClass_python=3.10 python=3.10
conda activate DeepMicroClass_python=3.10
Check if pip
is under the new environment:
which pip
Your pip
should be under the new environment:
/home/your_username/anaconda3/envs/DeepMicroClass_python=3.10/bin/pip
If not, install pip
under the new environment:
conda install pip
Then check it again.
After that, install DeepMicroClass by:
pip install DeepMicroClass
Install module requests
:
pip install requests
Testing
To test DeepMicroClass, You can simply run:
DeepMicroClass test
It will use test.fa
as input, and use all default settings to classify the input sequences, then write the test result in the current working directory.
To find the expected output, You can run:
DeepMicroClass -h
Usage
Prediction
usage: DeepMicroClass predict [-h] --input INPUT --output_dir OUTPUT_DIR [--model MODEL] [--encoding {onehot,embedding}] [--mode {hybrid,single}] [--single-len SINGLE_LEN] [--device {cpu,cuda}]
options:
-h, --help show this help message and exit
--input INPUT, -i INPUT
Path to the input fasta file
--output_dir OUTPUT_DIR, -o OUTPUT_DIR
Path to the output directory
--model MODEL, -m MODEL
Path to the trained model
--encoding {onehot,embedding}, -e {onehot,embedding}
Encoding method
--mode {hybrid,single}, -md {hybrid,single}
Prediction mode
--single-len SINGLE_LEN, -sl SINGLE_LEN
Length to use in the single mode
--device {cpu,cuda}, -d {cpu,cuda}
Device to use
The most straightforward way to run the prediction is to use the following command:
DeepMicroClass predict -i <input_fasta> -o <output_dir>
Prediction File Format
Each row of the output prediction file contains the scores from the neural network for class label assignment. The higher the score, the more likely the class label for a given sequence.
Sequence Name
: the name of the sequence in the input fasta fileEukaryote
: the score of the current sequence being Eukaryote.EukaryoteVirus
: the score of the current sequence being EukaryoteVirusPlasmid
: the score of the current sequence being PlasmidProkaryote
: the score of the current sequence being prokaryoteProkaryoteVirus
: the score of the current sequence being ProkaryoteVirus
As an example:
Sequence Name Eukaryote EukaryoteVirus Plasmid Prokaryote ProkaryoteVirus
NC_008199#3k#12001#15000 0.0001718740027172454 1.2298997305038642e-05 0.1367506879128939 0.4601334227976679 5.402931783348322
To assign a label to a given sequence, we find the maximum of all the assigned scores. In the above case, ProkaryoteVirus
has the highest score, so the sequence is assigned as belonging to the class ProkaryoteVirus
.
Sequence Extraction
You can extract sequences from a given class of your interest with the following command
DeepMicroClass extract --tsv <output.tsv> --fasta <input.fa> --class <class of interest> --output <output.fa>
The naming of class of interest follows the header of the output.
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 Distribution
File details
Details for the file deepmicroclass-1.0.5.tar.gz
.
File metadata
- Download URL: deepmicroclass-1.0.5.tar.gz
- Upload date:
- Size: 5.8 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.11.4
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | b6ef92e7b215caecac4df4b1e255555a2bccb4a220da4d1a009de367d211afb0 |
|
MD5 | 49379fb63f172338bc89714936d16aed |
|
BLAKE2b-256 | bf2c3e629b9cbb63f9f3ed31a1822856809149ae9dad36785bcd702b04c34fac |
File details
Details for the file DeepMicroClass-1.0.5-py3-none-any.whl
.
File metadata
- Download URL: DeepMicroClass-1.0.5-py3-none-any.whl
- Upload date:
- Size: 5.8 MB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.11.4
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 0bdba9bacbe6149234f5051a9f51e658a0f0e8334fcfe0132cc60663ea5b475e |
|
MD5 | 68467014dde6baa2d52fae479aef0e7b |
|
BLAKE2b-256 | ef172ab80f108aeeb75248fa73f301d35be09f95806e03f63a6f27d5dc3f17bc |