Fold2Seq

[ICML2021] Fold2Seq: A Joint Sequence(1D)-Fold(3D) Embedding-based Generative Model for Protein Design

Environment file:

Data and Feature Generation:

  • Go to data/ and check the README there.

How to train the model:

  • go to src/ and run:

python train.py --data_path $path_to_the_data_dictionary --lr $learning_rate --model_save $path_to_the_saved_model

How to generate sequences:

  • go to src/ and run:

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:

fold2seq3

GitHub

https://github.com/IBM/fold2seq