Train test splits can be found in chainsaw_cath1363_train_test_splits.json
Ground truth domains, Chainsaw predictions and performance metrics on the CATH1363 test set can be found in chainsaw_model_v3_on_cath1363_test.csv
Chainsaw predictions on the CATH1363 test set can be reproduced with the following command:
python scripts/get_predictions.py --structure_directory path/to/pdbs --model_dir weight/model_v3 --config_dir conf/model_v3 --renumber_pdbs -o output.tsv --use_first_chain --no_post_processing
Note that the predictions in chainsaw_model_v3_on_cath1363_test.csv assume zero-indexed pdbs with consecutive indices for all residues in the pdb file.
| separates different domains
_ separates different segments in the same domain