Exception 'NoneType' object is not subscriptable occurred for complex APBB1_24. Skipping.

[mechealeth]

Hi ,

I trying to Run SPOC on my AlphaPullDown result. My input folder /mnt/Q71SY5_med25_and_APBB1_24 as follwoing :

And I run the command : python3 run.py /mnt/Q71SY5_med25_and_APBB1_24

but I got the error info as following :

I'm not sure where the error is coming from and would appreciate any suggestions.

Many thanks

MinC

[gladshire]

Encountered the same issue. Looks to be a bug in their code. It occurs when you have amino nitrogens with pLDDTs less than 50, followed by an atom with a pLDDT greater than 50. 'get_data_from_structure()' will skip creating a 'residue' variable for the sub-50 nitrogen atom. When it then hits the 'CA' after that, it tries to subscript the non-existent 'residue' variable with 'atoms':

for atom_line in get_lines_from_pdb_file(pdb_filepath):
    if atom_line[0:4] != 'ATOM':
        continue
    
    atom_type = atom_line[13:16].strip()
    chain = atom_line[20:22].strip()
    aa_type = aa_3c_to_1c[atom_line[17:20]]
    is_nitrogen = atom_type == 'N'
    #in AlphaFold output PDB files, pLDDT values are stored in the "bfactor" column 
    bfactor = float(atom_line[60:66])
    if(is_nitrogen):
        #Keep track of what the absolute index of this residue in the file is 
        abs_res_index += 1
        if chain != last_chain_2: 
            if last_chain_2 is not None:
                pdb_res_sequence += ':'
            last_chain_2 = chain
        
        pdb_res_sequence += aa_type
        plddts.append(bfactor)

    if bfactor < min_plddt:
        #No need to examine this atom since it does not meet the pLDDT cutoff, skip it
        continue

    atom = atom_from_pdb_line(atom_line)
    if is_nitrogen:

        #Every amino acid residue starts PDB entry with exactly one "N" atom, so when we see one we know we have just encountered a new residue
        if chain != last_chain:
            #Are we in a new chain? If so, increment chain index and create new list in "residues"
            chain_index += 1
            last_chain = chain
            N_coords.append([])
            residues.append([])
            chains.append(chain)

        residue = {"chain":chain, "atoms":[],'c_ix':int(atom_line[22:26]), "a_ix":abs_res_index, "type":aa_type, "plddt":bfactor}
        residues[chain_index].append(residue)

        #add nitrogen atom coordinates to coordinates list to allow for fast broad searching later
        N_coords[chain_index].append(atom['xyz'])

    residue['atoms'].append(atom)

Note how 'residue' is only created if the 'N' atom has a pLDDT greater than min_plddt.

Hopefully it gets fixed soon.

[NicholasStevers]

Hello,

I was getting the same error as above, but I might have a solution.

If you use the DeepMind AF2-multimer code from their github the outputs are not what SPOC is expecting as the pae files do not contain pTM or ipTM scores. Rather, all scores are stored in the pkl files.

Here is a script I have written to extract and create equivalent jsons of the pae, pTM, and ipTM from the pkl files.

import os
import glob
import pickle
import json
import numpy as np

def convert_pkl_to_json(pkl_file: str) -> None:
    """Load an AlphaFold2 result_model_*.pkl file and save pae/ptm/iptm to a .json."""
    with open(pkl_file, "rb") as f:
        data = pickle.load(f)
    
    required_keys = ["predicted_aligned_error", "ptm", "iptm"]
    if not all(k in data for k in required_keys):
        print(f"Skipping {pkl_file} - missing required keys.")
        return

    # Convert NumPy arrays/floats into vanilla Python structures
    json_dict = {
        "pae": data["predicted_aligned_error"].tolist(),
        "ptm": float(data["ptm"]),
        "iptm": float(data["iptm"])
    }
    
    json_file = "pae_" + os.path.splitext(pkl_file)[0] + ".json"
    with open(json_file, "w") as f:
        json.dump(json_dict, f)
    print(f"Created {json_file}")

if __name__ == "__main__":
    # Adjust pattern if your pkl files use different naming
    pkl_files = glob.glob("result_model_*.pkl")
    for pkl_file in pkl_files:
        convert_pkl_to_json(pkl_file)

This will take a list of pkl files:
result_model_1_multimer_v3_pred_0.pkl
result_model_1_multimer_v3_pred_1.pkl
result_model_1_multimer_v3_pred_2.pkl
result_model_1_multimer_v3_pred_3.pkl
result_model_1_multimer_v3_pred_4.pkl
result_model_2_multimer_v3_pred_0.pkl
result_model_2_multimer_v3_pred_1.pkl
result_model_2_multimer_v3_pred_2.pkl
result_model_2_multimer_v3_pred_3.pkl
result_model_2_multimer_v3_pred_4.pkl
result_model_3_multimer_v3_pred_0.pkl
result_model_3_multimer_v3_pred_1.pkl
result_model_3_multimer_v3_pred_2.pkl
result_model_3_multimer_v3_pred_3.pkl
result_model_3_multimer_v3_pred_4.pkl
result_model_4_multimer_v3_pred_0.pkl
result_model_4_multimer_v3_pred_1.pkl
result_model_4_multimer_v3_pred_2.pkl
result_model_4_multimer_v3_pred_3.pkl
result_model_4_multimer_v3_pred_4.pkl
result_model_5_multimer_v3_pred_0.pkl
result_model_5_multimer_v3_pred_1.pkl
result_model_5_multimer_v3_pred_2.pkl
result_model_5_multimer_v3_pred_3.pkl
result_model_5_multimer_v3_pred_4.pkl

and generate json equivalents:
pae_result_model_1_multimer_v3_pred_0.json
pae_result_model_1_multimer_v3_pred_1.json
pae_result_model_1_multimer_v3_pred_2.json
pae_result_model_1_multimer_v3_pred_3.json
pae_result_model_1_multimer_v3_pred_4.json
pae_result_model_2_multimer_v3_pred_0.json
pae_result_model_2_multimer_v3_pred_1.json
pae_result_model_2_multimer_v3_pred_2.json
pae_result_model_2_multimer_v3_pred_3.json
pae_result_model_2_multimer_v3_pred_4.json
pae_result_model_3_multimer_v3_pred_0.json
pae_result_model_3_multimer_v3_pred_1.json
pae_result_model_3_multimer_v3_pred_2.json
pae_result_model_3_multimer_v3_pred_3.json
pae_result_model_3_multimer_v3_pred_4.json
pae_result_model_4_multimer_v3_pred_0.json
pae_result_model_4_multimer_v3_pred_1.json
pae_result_model_4_multimer_v3_pred_2.json
pae_result_model_4_multimer_v3_pred_3.json
pae_result_model_4_multimer_v3_pred_4.json
pae_result_model_5_multimer_v3_pred_0.json
pae_result_model_5_multimer_v3_pred_1.json
pae_result_model_5_multimer_v3_pred_2.json
pae_result_model_5_multimer_v3_pred_3.json
pae_result_model_5_multimer_v3_pred_4.json

these can then be used successfully with the run.py from SPOC.

[mechealeth]

Hello,

I was getting the same error as above, but I might have a solution.

If you use the DeepMind AF2-multimer code from their github the outputs are not what SPOC is expecting as the pae files do not contain pTM or ipTM scores. Rather, all scores are stored in the pkl files.

Here is a script I have written to extract and create equivalent jsons of the pae, pTM, and ipTM from the pkl files.

import os
import glob
import pickle
import json
import numpy as np

def convert_pkl_to_json(pkl_file: str) -> None:
    """Load an AlphaFold2 result_model_*.pkl file and save pae/ptm/iptm to a .json."""
    with open(pkl_file, "rb") as f:
        data = pickle.load(f)
    
    required_keys = ["predicted_aligned_error", "ptm", "iptm"]
    if not all(k in data for k in required_keys):
        print(f"Skipping {pkl_file} - missing required keys.")
        return

    # Convert NumPy arrays/floats into vanilla Python structures
    json_dict = {
        "pae": data["predicted_aligned_error"].tolist(),
        "ptm": float(data["ptm"]),
        "iptm": float(data["iptm"])
    }
    
    json_file = "pae_" + os.path.splitext(pkl_file)[0] + ".json"
    with open(json_file, "w") as f:
        json.dump(json_dict, f)
    print(f"Created {json_file}")

if __name__ == "__main__":
    # Adjust pattern if your pkl files use different naming
    pkl_files = glob.glob("result_model_*.pkl")
    for pkl_file in pkl_files:
        convert_pkl_to_json(pkl_file)

This will take a list of pkl files: result_model_1_multimer_v3_pred_0.pkl result_model_1_multimer_v3_pred_1.pkl result_model_1_multimer_v3_pred_2.pkl result_model_1_multimer_v3_pred_3.pkl result_model_1_multimer_v3_pred_4.pkl result_model_2_multimer_v3_pred_0.pkl result_model_2_multimer_v3_pred_1.pkl result_model_2_multimer_v3_pred_2.pkl result_model_2_multimer_v3_pred_3.pkl result_model_2_multimer_v3_pred_4.pkl result_model_3_multimer_v3_pred_0.pkl result_model_3_multimer_v3_pred_1.pkl result_model_3_multimer_v3_pred_2.pkl result_model_3_multimer_v3_pred_3.pkl result_model_3_multimer_v3_pred_4.pkl result_model_4_multimer_v3_pred_0.pkl result_model_4_multimer_v3_pred_1.pkl result_model_4_multimer_v3_pred_2.pkl result_model_4_multimer_v3_pred_3.pkl result_model_4_multimer_v3_pred_4.pkl result_model_5_multimer_v3_pred_0.pkl result_model_5_multimer_v3_pred_1.pkl result_model_5_multimer_v3_pred_2.pkl result_model_5_multimer_v3_pred_3.pkl result_model_5_multimer_v3_pred_4.pkl

and generate json equivalents: pae_result_model_1_multimer_v3_pred_0.json pae_result_model_1_multimer_v3_pred_1.json pae_result_model_1_multimer_v3_pred_2.json pae_result_model_1_multimer_v3_pred_3.json pae_result_model_1_multimer_v3_pred_4.json pae_result_model_2_multimer_v3_pred_0.json pae_result_model_2_multimer_v3_pred_1.json pae_result_model_2_multimer_v3_pred_2.json pae_result_model_2_multimer_v3_pred_3.json pae_result_model_2_multimer_v3_pred_4.json pae_result_model_3_multimer_v3_pred_0.json pae_result_model_3_multimer_v3_pred_1.json pae_result_model_3_multimer_v3_pred_2.json pae_result_model_3_multimer_v3_pred_3.json pae_result_model_3_multimer_v3_pred_4.json pae_result_model_4_multimer_v3_pred_0.json pae_result_model_4_multimer_v3_pred_1.json pae_result_model_4_multimer_v3_pred_2.json pae_result_model_4_multimer_v3_pred_3.json pae_result_model_4_multimer_v3_pred_4.json pae_result_model_5_multimer_v3_pred_0.json pae_result_model_5_multimer_v3_pred_1.json pae_result_model_5_multimer_v3_pred_2.json pae_result_model_5_multimer_v3_pred_3.json pae_result_model_5_multimer_v3_pred_4.json

these can then be used successfully with the run.py from SPOC.

thanks alot ! it also work for me!