Contact Map / Residue Importance for MUSE-PPI

[orekhovrv]

Hi, congratulations on the publication and many thanks for publishing the code!

I'm playing with the MUSE-PPI model and was able to train it, but I can't wrap my head around how contact map or residue importance you reported in the paper can be extracted from this model.

Could you please add a bit more details on this part?

[a1e-the]

can you tell me how to solve"FileNotFoundError: [Errno 2] No such file or directory: '/data/user/raojh/worksapce/MUSE_dev/data/snap_dpi/raw/pdb/P49862.tensor" i already downloaded edge_list_12.npy and x_list.pt from website, but dont know how to use them

[Kuzaiherba]

Hi @a1e-the, I managed to run training of PPI model, but with a bunch of workarounds:

• to produce .tensor & .dssp files from .pdb files, I used this script: https://github.com/biomed-AI/GPSite/blob/main/script/feature_extraction/process_structure.py
  • there is also a function that allows you to get pdb-structures using ESMFold, but I used AF2.
• in addition, it required a few minor changes in code to run (e.g., transferring some tensors to the same device).

In the end, I wasn't any close to reproducing the results though.