Training data should be placed in the directory training_data/, and this directory should reside in the project root directory.

The format of the training data files, be it naming or content, are as per what was given to us.

Similarly, testing data should be placed in the directory testing_data/, and this directory should reside in the project root directory.

The format is also as per what was given to us.

The following assumes that you are using Python3 and have libraries like numpy and keras installed. You may also need to install GraphViz as the training step also tries to print out a visualization of the model.

For both models, you should expect the following output:

• A graph with loss and accuracy plots saved as a .png file
• Model weights saved as a .h5 file
• Model visualization saved as a .png file

To start training the Dual-stream 3DCNN, simply run the following command.

python train.py

To start training the MLp model, simply run the following command.

python dist_train.py

First make sure you change the WEIGHTS_FILENAME variable in predict_utils.py. Then run the following command.

python predict.py

The predictions of the top ten ligand candidates for binding to any protein will be printed in the file test_predictions.txt.