lsj2408 / Transformer-M

[ICLR 2023] One Transformer Can Understand Both 2D & 3D Molecular Data (official implementation)
https://arxiv.org/abs/2210.01765
MIT License
201 stars 24 forks source link

How to encode proteins in the PDBbind task? #3

Closed fanxiaoyu0 closed 1 year ago

fanxiaoyu0 commented 1 year ago

Very enlightening work. Congratulations on your great achievements in the OGB Challenge! In addition, I noticed that you have made fine-tuning on the PDBbind dataset. How should you encode the protein information? Because proteins usually contain more heavy atoms, do you directly use Transformer-M to encode proteins?

1874Mercury commented 1 year ago

Also looking forward to the answer to this question...

lsj2408 commented 1 year ago

It is a good catch! For the PDBBind task, each data instance is a protein-ligand complex. Since the size of a protein is much larger than that of a ligand, each protein-ligand complex only contains the protein atoms around the ligand. Thus, the model receives the complex as input and predicts its binding affinity.