VFM is developed for identifing phages from metagenomic bins or contigs, which has two versions： bin-VFM for metagenomic bins and unbin-VFM for metagenomic contigs.
This project is composed of predicting and training scripts.The details are as follows.
- Predicting scripts - Predict bins or contigs as phages or bacteria
bin-VFM_predict.py
unbin-VFM_pred.py - Training scripts - Train a new model using user's own data
train_bin-VFM.py
train_unbin-VFM.py
VFM: Identification of Bacteriophages From Metagenomic Bins and Contigs Based on Features Related to Gene and Genome Composition(https://ieeexplore.ieee.org/document/8924706)
The system must be Linux, with Python3 installed on. Some python packages and modules should be installed in this way:
pip3 install numpy pandas scipy biopython scikit-learn
VFM depends on some bioinformatic tools which should be installed properly:
- Prokka - For gene prediction
- HMM tools - For hidden Markov models associated with gene/protein families
- BLAST+ - For sequence alignments
VFM can be set up by:
git clone https://github.com/liuql2019/VFM
Pressing the download button can fix it too. The gene databases pVOG and COG must be downloaded from the link: https://www.jianguoyun.com/p/Ddnq2oQQ7I_kBxi4kvUB
Unzip the file using the command:
unzip database.zip -d /{the path to VFM}/VFM/models
In order to generate metagenomic bins, binning tools such as COCACOLA should be applied to metagenomic contigs.Then run the script bin-VFM_predict.py to predict the bins.
To predct metagenomic bins as phages or bacteria, bin-VFM is used as follows:
python3 bin-VFM_predict.py -d BINS_DIR [-t THREADS] [-m MODEL]
where the parameter -d means the directory of the bins, -t means cpu number and -m means user-trained model for predicting. If the parameter -t is omitted, default value 1 will be used. Omitting parameter -m means choosing the default model that has been stored in VFM before release.
Similarly, unbin-VFM is used to predict metagenomic contigs by:
python3 unbin-VFM_pred.py -f FA_FILE [-t THREADS] [-m MODEL]
where the parameter -f means the fasta file in which the contigs are stored. The meanings of -t and -m are the same as above.
The data in the folder data/test is used to demenstrate how to run VFM.
Run bin-VFM as follows:
python3 bin-VFM_predict.py -d ./data/demo/bins -t 4
Run unbin-VFM as follows:
python3 unbin-VFM_pred.py -f ./data/demo/contigs.fasta -t 4
Users could train their own model(s) by scripts train_bin-VFM.py and train_unbin-VFM.py.
Run train_bin-VFM.py as follows:
python3 train_bin-VFM.py -vir VIR_DIR -bac BAC_DIR [-cpus THREADS] -model MODEL_NAME
where the parameters -vir and -bac are folders of phage and bacterium bins for training, -model is the model name. The paremeter -cpus is cpu number which may be omitted.
Use train_unbin-VFM.py as follows:
python3 train_unbin-VFM.py -vir VIR_FILE -bac BAC_FILE [-cpus THREADS] -model MODEL_NAME
where the parameters -vir and -bac are fasta files of phage and bacterium contigs for training.The other parameters are the same as above.
The training sets and test sets used by VFM can be downloaded from https://www.jianguoyun.com/p/DYIe6QgQ7I_kBxihkPUB.
Unzip the package by
unzip train_test_data.zip
We are extremely grateful to the authors of MARVEL and virFinder for the contribution of their open source code，some of which was used in VFM.
Codes here may be modified and used for any purpose.This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.