xDeep-AcPEP

Deep Learning Method for Anticancer Peptide Activity Prediction through Convolutional Neural Network and Multi-Task Learning

In this repository:

1. We provide python scripts for reproducing the experiments of 5-folds cross-validation model comparison.
2. We provide our final production models for peptide activity prediction.

Requirements

• Anaconda 4.7.0
• Python 3.6.9
• Scikit-learn 0.21.3
• Pytorch 1.2.0 with CUDA 10.0
• Scipy 1.4.1
• Pandas 1.0.2
• Numpy 1.18.1
• Openpyxl 3.0.6

Model and Data

1. Models and data used for reproducing experiments are available at: Here
2. Final production models for peptide activity prediction are available at: ./prediction/model/

Run the script

1. Reproducing Experiments

The script is located in model_comparison_CV folder

python reproduce.py -mo <model_folder_path> -da <data_folder_path> -o <output_folder_path>

Example:

python reproduce.py -mo ./model/ -da ./data/ -o ./

2. Final production model prediction

The script is located in prediction folder

python prediction.py -t <tissue_type> -m <model_folder_path> -d <fasta_file_path> -o <output_folder_path>

where:
<tissue_type> could be selected from breast, cervix, colon, lung, prostate and skin.

Example:

python prediction.py -t breast -m ./model/ -d ./test_breast.fasta -o ./result/

Note: The input peptide data must in the form of the following FASTA format.

>AmphiArc1
KWVKKVHNWLRRWIKVFEALFG
>AmphiArc2
KIFKKFKTIIKKVWRIFGRF
>Gradient2
AWLKRIKKFLKALFWVWVW 
>AmphiArc3
AFRHSVKEELNYIRRRLERFPNRL

References

1. We used iFeature to extract all peptide features. (Github, Paper)