We propose a retrieval-augmented diffusion framework, termed RADAb, for efficient antibody design. Our method leverages a set of structural homologous motifs that align with query structural constraints to guide the generative model in inversely optimizing antibodies according to desired design criteria.
You can create a new environment using the requirements.txt by $ conda create --name <env> --file requirements.txt.
Antibody structures in the SAbDab dataset can be downloaded here. Extract chothia folder in all_structures.zip into the ./data folder.
Training process is implemented in train.py. The configuration is in the ./config folder.
You can evaluate the generated structures by sequentially running run_folding.py, run_relax.py, and run_eval.py.