[ICML2021] Fold2Seq: A Joint Sequence(1D)-Fold(3D) Embedding-based Generative Model for Protein Design
![Fold2Seq Architecture](https://github.com/IBM/fold2seq/raw/main/fold2seq1.png)
Environment file:
Data and Feature Generation:
- Go to data/ and check the README there.
How to train the model:
python train.py --data_path $path_to_the_data_dictionary --lr $learning_rate --model_save $path_to_the_saved_model
How to generate sequences:
python inference.py --trained_model $path_to_the_trained_model --output $path_to_the_output_file --data_path $path_to_the_data_dictionary
Fold2Seq generated structures against natural structures:
![Fold2Seq structures](https://github.com/IBM/fold2seq/raw/main/fold2seq3.png)