AcrNET: Predicting Anti-CRISPR with Deep Learning

Pytorch implementation and trained model for AcrNET: Predicting Anti-CRISPR with Deep Learning.

Environment Setup

• Download the repository.

  git clone https://github.com/banma12956/AcrNET.git
  cd AcrNET

• Install anaconda or conda.
• Create an environment for AcrNET and activate it.

  conda env create -f environment.yml
  conda activate AcrNET

Run a simple demo

We provide a trained model and a simple dataset for demonstration. Run the following code and try:

python test.py

Train

We also provide the five-fold cross-validation training code we used in our experiment:

python train_five_fold.py

The data we used in experiments can be downloaded at https://drive.google.com/file/d/1LK6y9g75ktlJEOy3CXZcPpZjQ4h4l-Ws/view?usp=sharing

Data

You can also upload your own protein sequence data and use our trained model to make prediction.

DeepAcr needs structure information, evolutionary information and Transformer feature as input features, as shown in folder data. RaptorX, POSSUM, ESM-1b can provide corresponding calculation.