Case Study

There are many case studies where one single line is performed. Suppose you are under the main folder of DeepProtein where it contains a folder called train.

1a. Protein Function (Property) Prediction

We take Beta-lactamase dataset as an example.

CNN Case:

$ python train/beta.py --target_encoding CNN --seed 7 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

GNN Case:

$ python train/beta.py --target_encoding PyG_GCN --seed 7 --wandb_proj DeepProtein --lr 0.00001 --epochs 100 --compute_pos_enc True

1b. Protein Protein Interaction

We take PPI Affinity dataset as an example.

CNN Case:

$ python train/ppi_affinity.py --target_encoding CNN --seed 42 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

GNN Case:

$ python train/ppi_affinity.py --target_encoding PyG_GCN --seed 42 --wandb_proj DeepProtein --lr 0.00001 --epochs 100 --compute_pos_enc True

Pair/PPI torch-geometric support is now available for PyG_GCN, PyG_GAT, PyG_GraphSAGE, PyG_GIN, PyG_ChebNet, and PyG_TAGConv, including optional Laplacian positional encoding with compute_pos_enc=True.

1c. Protein Localization Prediction

We take SubCellular dataset as an example.

CNN Case:

$ python train/subcellular.py --target_encoding CNN --seed 7 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

GNN Case:

$ python train/subcellular.py --target_encoding PyG_GAT --seed 7 --wandb_proj DeepProtein --lr 0.00001 --epochs 100 --compute_pos_enc True

1d. Antigen Epitope Prediction

We take IEDB dataset as an example.

CNN Case:

$ python train/IEDB.py --target_encoding Token_CNN --seed 7 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

1e. Antibody Paratope Prediction

We take SAbDab Liberis dataset as an example.

CNN Case:

$ python train/SAbDab_Liberis.py --target_encoding Token_CNN --seed 7 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

1f. Antibody Developability Prediction (TAP)

We take TAP dataset as an example.

CNN Case:

$ python train/TAP.py --target_encoding CNN --seed 7 --wandb_proj DeepProtein --lr 0.0001 --epochs 100

GNN Case:

$ python train/TAP.py --target_encoding PyG_GAT --seed 7 --wandb_proj DeepProtein --lr 0.00001 --epochs 100 --compute_pos_enc True