KO-Identification

Requirements

See requirements.txt

Getting Started

# get ProtT5 embedding (input format: fasta, output format: h5)
python prott5_seq2embedding.py seq_file.fa embedding_file.h5

# use the trained classifier to make predictions (output format: csv)
python prediction.py model/mlp_pipe.pt embedding_file.h5 cls_result.csv

# cluster (result file format: csv)
python cluster.py embedding_file.h5 reference.h5 cluster_result.csv

# train and test the classifier (The paths to the model and dataset can be modified in the python file)
python cls/mlp_train.py
python cls/mlp_test.py

Models

If you want to apply the model directly, you should use mlp_pipe.pt.

└── Model
    ├── att_cls.pt          # attention model
    ├── lstm_cls.pt         # LSTM model
    ├── mlp_cls.pt          # MLP model (used for classifier testing)
    └── mlp_pipe.pt         # MLP model (used for testing the entire pipeline)

Data availability

Publicly available datasets were analyzed in our paper. These datasets were collected from the KEGG database, the PDB database, and the AFDB database.