This repository contains parts of our disclosed data.
Code for inference.
We take "dBET6 fragments.mol" for testing.
The training set "dataset.smi" is constructed from PROTAC-DB and ZINC.
Code in "postprocess.py" reveals some functions we designed to filtering ourputs generated from deep encoder-decoder network.
The files "docking protein.pdb" and "docking ligands.smi", including predicted molecules passing the postprocess step, are input for docking validation. Top 10 best RMSD outputs are listed in "top 10 docking compounds.sdf".
Please contact Chu-Chung Lin [email protected].