Evaluation and pre-trained models.
- torchdiffeq : https://github.com/rtqichen/torchdiffeq.
- pytorch >= 1.10.0
- torch-scatter
- torch-sparse
- torch-cluster
- torch-spline-conv
- torch-geometric == 2.0.4
- astropy
- networkx
- tqdm
- Biopython
Make sure to have the correct version of each library as some errors might arise due to version-mismatch. The libraries-version of the local conda env are in env_list.txt
Datasets are placed in "data/" folder. Note that Structural Antibody Dataset is an evolutionaey dataset in which new structures are added each week. The model has been evaluated on a same specification as other competing methods in benchmarks.
Unconditional Antibody Sequence and structure generation:
python evaluation_uncond.py --cdr ${CDR-region-antibody}(1/2/3)
Conditional Antibody Sequence and structure generation:
python evaluation_cond.py --cdr ${CDR-region-antibody}(1/2/3)