How to sample mols on all pockets of the test set of Crossdocked dataset?

[youmingzhao91]

Hi, I am trying to set DiffSBDD cond method as one of baselines in my work, I would like to sample mols given 100 pockets from the test set of Crossdocked dataset on my own machine. What can I do?

python generate_ligands.py <checkpoint>.ckpt --pdbfile <pdb_file>.pdb --outdir <output_dir> --resi_list <list_of_pocket_residue_ids>

How can I specify --resi_list <list_of_pocket_residue_ids> or --ref_ligand :?

Does the following command only work for the Bindingmoad dataset? I only need to run your sample code on the test set of Crossdoced.

python test.py <checkpoint>.ckpt --test_dir <bindingmoad_dir>/processed_noH/test/ --outdir <output_dir> --fix_n_nodes

Could you please give me some instructions for my goal?

[youmingzhao91]

In addition, I have preprocessed the CrossDocked dataset in the same way as Pocket2Mol.

[yuanqidu]

Thanks for your interest in our work.

The following command would work for crossdocked dataset as well for sampling over the test set

python test.py <checkpoint>.ckpt --test_dir <bindingmoad_dir>/processed_noH/test/ --outdir <output_dir> --fix_n_nodes

[youmingzhao91]

Thanks! I'm using generate_ligands.py with a for-loop to do that. It works now.

[arneschneuing]

Hi, I would just like to add that this repository only contains checkpoints for BindingMOAD at the moment. Please keep this in mind when sampling molecules for the CrossDocked test set.

[youmingzhao91]

Where can I get the checkpoint for the CrossDocked dataset? Can you upload it here?

[youmingzhao91]

Hi, I would just like to add that this repository only contains checkpoints for BindingMOAD at the moment. Please keep this in mind when sampling molecules for the CrossDocked test set.

The following checkpoint

https://github.com/arneschneuing/DiffSBDD/blob/main/checkpoints/full_atom.ckpt

is only for BindingMOAD? However, I am using it to sample molecules for the CrossDocked test set right now.

[arneschneuing]

Yes, the model was trained on the Binding MOAD dataset