3CPDD

Code base

our code is based on ABdockgen

Dependencies

• PyTorch
• tqdm
• sidechainnet
• prody
• torchsparse

3D convolution

Data

Data for testing the basic functions are in folder data

To download the pretraining data set please run, this will take about 1 hour and generate a jsonl file with size 5G

python pdbfile/pretrain_data

Training

training script can be launched by

python dock_train.py --L_target 20 --save_dir ckpts/HERN-dock
python dock_train.py --L_target 20 --sparse_encoder --save_dir ckpts/HERN-dock

mkdir outputs
python predict.py ckpts/HERN-dock-sparse/model.best data/rabd/test_data.jsonl

It will produce a PDB file for each epitope in the test set.

Context given paratope Design

The training script can be launched by

python lm_train.py --L_target 20 --save_dir ckpts/HERN-gen
python lm_train.py --L_target 20 --sparse_encoder --save_dir ckpts/HERN-gen

At test time, we can generate new CDR-H3 paratopes specific to a given epitope:

python generate.py ckpts/HERN_gen.ckpt data/rabd/test_data.jsonl 1 > results/HERN.txt

The above script will generate one CDR-H3 sequence per epitope. You can sample more candidates by changing this parameter.