This is a description of the data folder.
TestData
Here you can quickly test the model by the testdata.
We provide less than 50 proteins, and then you can simply run preprocessing.sh and next training steps by setting {datapath} to data/testdata. This still takes several minutes.
- Structual feature of node2vec is saved to testdata/emb_graph_bp[/mf/cc].npy.
- Sequence feature is saved to testdata/esm_feature.npy. If you don't want to download esm2_t33_650M_UR50D, you can use the esm_feature.npy we provide.
Your Dataset
If you want to use your own dataset, please follow the steps below:
- The PPI file is placed in the datadir/PPIdata directory, with a file name such as: 3702.protein.links.v11.0
- Download the ID mapping file from the string database to the uniprot database from Uniprot, and put it in datadir/IDmap-uniprot-string, with the file name and format like uniprotkb_3702.xlsx
- Download all_uniprot_goa.gaf from GOA website, and put it into datadir/GO. We provide get_goa_spiece function in data_preprocessing.py to extract spieces goa you need, and you need to enable it in main function of data_preprocessing.py
Before you run preprocessing.sh, don't forget modify TAXIDS in utils.py
This may take a long time.
Our Dataset
We also provide the full dataset we used in the training, which can be downloaded from here