toxify

deep learning approach to classify animal toxins

Installation

1. Get the repository for toxify

git clone https://github.com/tijeco/toxify.git

2. Change into toxify repository

cd toxify

3. Create conda environment with dependencies needed to run toxify

conda env create -f requirements.yml

4. Activate toxify conda environment

source activate toxify_env

5. Install toxify to the toxify conda environment

python setup.py install

6. Run toxify

toxify predict <input.fasta>

7. view results in <input.fasta>_toxify_predictions/predictions_proteins.csv
8. Deactivate toxify conda environment

source deactivate

Custom training dataset

If you wish to train a different model with your own training data, that can be accomplished with toxify using the train sub command.

There are a few variables that can be modified in terms of setting up the model.

• -pos can be followed by a list of protein sequence fasta files that you wish to constitute the positive dataset

• -neg can be followed by a list of protein sequence fasta files that you wish to constitute the negative dataset

• -maxLen is the maximum length (integer) of protein seqeunce to be included in the training set

• -units is the number (integer) of gated recurrent units to be used in the model

• -epochs is the number (integer) of training epochs to run for the model

• lr is the learning_rate (float, 0 < lr < 1)