Prepare Data

1. Download the dataset from here and put it into the directory "./data/dscript/data"
2. Run the following script to prepare the source data and get protein embeddings by ProtT5.

bash scripts/dscript/0.prepare-dataset.sh

Train

To train the PPITrans (with and without pre-trained protein embedding), run

bash scripts/dscript/1.train-ppi.sh

Predict

To predict the results of the testset, run

bash scripts/dscript/2.predict-ppi.sh

Evaluate

To evaluate the performance of PPITrans, run

bash scripts/dscript/3.evaluate-ppi.sh

Run probes

The directory ./experiments contains the source code for all probes, and subdirectories correspond to probes. In each subdirectory, run

bash run.bash

to conduct the corresponding probe, and results will appear in the "results" directory. The "plot.ipynb" can be used to plot statistics.