All the trained models are on the website - https://caliber.math.biu.ac.il/, and can be predicted directly through the website. You can also use a git code for this
The test data can be in 3 different formats (sequnces in /fasta format, PDB IDs, PDB Files) main.py --mode predict --init XXX --model XXX --epi XXX --test_seq_input XXX main.py --mode predict --init XXX --model XXX --epi XXX --test_pdb_path XXX main.py --mode predict --init XXX --model XXX --epi XXX --test_pdb_list XXX
For example: main.py --mode predict --init Random --model BiLSTM --epi Nonlinear --test_seq_input data/nonlinear_test.fasta
main.py --mode train --init XXX --model XXX --epi XXX --train_path XXX --test_path XXX
For example: main.py --mode train --init Random --model BiLSTM --epi Linear --train_path data/linear_train.csv --test_path data/linear_test.csv
| Parameter | Description | Required | Options |
|---|---|---|---|
| --mode | pradiction using the pre trained model or training | True | "predict", "train" |
| --init | Choose the protein-encoding | True | "Kidera" ,"Random", "Kidera+bio", "ESM-2" |
| --model | Choose the model architecture | True | "BiLSTM" ,"GCN", "Boosting" |
| --epi | Choose the epitope sequences | True | "Linear", "Nonlinear", "Both" |
| --train_path | Path to the training data | only for train | path for the trainig data |
| --test_path | Path to the testing data | only for train | path for the test data |
| --test_seq_input | Path to predict data in FASTA format | False | path for the data in string |
| --test_pdb_path | Path to the predict data of PDB files | False | path for folder with PDB files |
| --test_pdb_list | Path to the predict data of PDB IDs in a list | False | path for file including PDB IDs |