agave233 / SIGN

Official implementation of 'Structure-aware Interactive Graph Neural Networks for the Prediction of Protein-Ligand Binding Affinity' (KDD 2021)
36 stars 8 forks source link

Testing SIGN on CSAR and new data #4

Open NECCCCCCC opened 2 years ago

NECCCCCCC commented 2 years ago

Hi,

In the paper, SIGN was tested on the PDBBind 2016 and CSAR, but I can't find any codes related to CSAR in the respository. Can you provide a script or any other methods to evaluate SIGN on CSAR. It can be very helpful. Moreover, if I have a new complex not from PDBBind or CSAR (e.g., a protein strcuture file and a ligand file which have been docked), how can I use SIGN to predict its binding affnity. Thanks.

erikedlund commented 2 years ago

Unfortunately I get the sense the authors aren't interested in having anyone actually use this tool. I've reached out here and directly to the corresponding scientist on the paper and received no response.

zhoujingbo commented 2 years ago

Document about our model can be found here: https://github.com/PaddlePaddle/PaddleHelix/tree/dev/apps/drug_target_interaction/sign

Besides, we may miss your email. You can send it again to the corresponding author.

erikedlund commented 2 years ago

Thanks, I found the documentation. I see instructions on how to train a model but not how to actually use it on an actual target not in the training data.

agave233 commented 2 years ago

Unfortunately I get the sense the authors aren't interested in having anyone actually use this tool. I've reached out here and directly to the corresponding scientist on the paper and received no response.

Hi, sorry for checking your issues till now. I will update the document details and the related codes for predicting the binding affinity for new protein-ligand complex as soon as possible. Also, the processing code for CSAR will also be updated. This work will be done before this weekend. Thanks a lot for your attention to our work!

erikedlund commented 2 years ago

Thank you, I will look forward to checking it out!

AlyciaBHZ commented 2 years ago

Thank you, I will look forward to checking it out!

Hi I am having the same trouble! I just started my research on 'predicting binding affinities for mhc and peptides using structural information'. I am wodering can you give me some expertise or recommendation on finding open sources project of GNN and binding affinities prediction. Thanks a lot!

erikedlund commented 2 years ago

Hi @AlyciaBHZ , one other cool related library I've found is DeepRank-GNN. Check it out here: https://github.com/DeepRank/Deeprank-GNN

erikedlund commented 2 years ago

Any updates on this?