How I can use pre-trained model to generate new chemistry with my pdb file (protein+ligand)

[sll513]

Thanks,

[INHYEOKSONG]

editing generate.config file , Currently, that's how I'm creating it.

[mattragoza]

You should create a new data file (refer to data/it2_tt_0_lowrmsd_valid_mols_test0_1000.types for an example) that references the structure files you want to use. Then you should copy/edit config/generate.config so that the data_file setting refers to your new data file and the data_root setting is correctly set as well. Then run python generate.py generate.config.

To create a new data file, the receptor and input ligand should first be split into separate files, and the ligand should be split into separate files for each pose. Then you should put them in the data file with the row format "{class_label:d} {affinity:f} {RMSD:f} {receptor_path} {ligand_path}". The only two relevant columns are receptor_path and ligand_path, which should be paths to the structure file relative to the data_root variable. I.e. if you have a structure file at /users/me/data/rec1/rec1.pdb and your data file contains a reference to rec1/rec1.pdb, then you should set data_root to /users/me/data.