Code

[11140321]

Hello, I have some questions about the use of this code. I read your paper and it only says that protein processing is carried out at three different levels. However, there are many folders in your code and I am not sure about them. What does it mean, and how does the process work?

[limei0307]

Hello @11140321,

Thanks for your interest in our work.

For ProNet, you can find the example code here. You can use the hyperparameter level to choose one of the three levels.

Details are in line 205
parser.add_argument('--level', type=str, default='backbone', help='Choose from \'aminoacid\', \'backbone\', and \'allatom\' levels')

The detailed model architecture is in this folder.

Let us know if you have other questions. Thanks.

[11140321]

Hello, I would like to ask if I want to use the code for drug target affinity prediction, can I extract the coordinate information based on the target's PDB file and process the above code?

[limei0307]

Hello @11140321,

Yes, you can.

You have to write your own dataset code to process your PDB files.

To use our model, no matter what's your raw data, eventually, you have to provide
data.x: one-hot amino acid type embedding,
data.coords_ca: alpha carbon atom coordinate,
as shown here.

If you want to use backbone level, you have to further provide data.bb_embs as shown here.

For allatom level, you have to further provide data.side_chain_embs as shown here.

Let us know if you have other questions. Thanks.