Genet is a convolutional model for shotgun metagenomic classification. Code is provided for training GeNet using the dataset from [1].
To install the required Python packages, install Anaconda and create a virtual environment by executing:
bash install_requirements.sh
Extact the NCBI taxonomy:
cd data
tar -xvzf nodes.dmp.tar.gz
rm nodes.dmp.tar.gz
To train the model, run the following:
source activate genet
cd code
python genet_train.py --dir_path=../data
The first execution of the code will download the required genomes from NCBI.
The resulting weights and tensorboard log will be saved in GeNet/saved_models.