Convert TIMED-style benchmark predictions to json file

convert_to_json.py

# Written by Rokas Petrenas (https://github.com/rokaske199)
from benchmark import get_cath
from pathlib import Path
import json

PATH_TO_PDB=Path('/scratch/datasets/pdb/')
cath_location = "cath-domain-description-file.txt"
cath_df = get_cath.read_data(cath_location)
#select only monomers
new_df=get_cath.filter_with_user_list(cath_df,"/home/s1706179/project/sequence-recovery-benchmark/monomers_af2.txt")
#choose your model
path_to_file=Path("/home/s1706179/project/sequence-recovery-benchmark/publication_data/default.csv")
predictions = get_cath.load_prediction_matrix(new_df, path_to_file.with_suffix('.txt'), path_to_file)

sequences={}
for x in predictions:
    sequences[x]=get_cath.most_likely_sequence(predictions[x])

with open('monomers_af.json', 'w') as f:
    json.dump(sequences, f)

monomers_af2.txt: https://gist.github.com/universvm/449a72b1f7f3cd033fe3381c695b4b49