ProteinMPNN_dataset

[jdhenaos]

A new Dataset for ProteinMPNN model was added.

The Dataset was evaluated using the next piece of code:

from torch_geometric.loader import DataLoader

path = "./sample_data"
params = {
    "LIST"    : f"{path}/pdb_2021aug02_sample/list.csv",
    "VAL"     : f"{path}/pdb_2021aug02_sample/valid_clusters.txt",
    "DIR"     : f"{path}/pdb_2021aug02_sample",
    "DATCUT"  : "2030-Jan-01",
    "RESCUT"  : 3.5, #resolution cutoff for PDBs
    "HOMO"    : 0.70 #min seq.id. to detect homo chains
}

train_dataset = PMPNNDataset(root=path,params=params,set_type='train')
validation_dataset = PMPNNDataset(root=path,params=params,set_type='val')

train_loader = DataLoader(train_dataset, batch_size=32, shuffle=True,
                          num_workers=2)
validation_loader = DataLoader(validation_dataset, batch_size=32, shuffle=True,
                          num_workers=2)