search

Graylab / IgFold

Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies
Other
319 stars 60 forks source link

ValueError: min() arg is an empty sequence #53

Closed xy127 closed 1 year ago

xy127 commented 1 year ago

Hello, I used python version of 3.11.4 to create a conda environment and pip install igfold. I successfully finished all the example codes without issue. But when I replace with my own sequence, the following error occurs:

  File "/home/ubuntu/user_space/predictantibody.py", line 10, in <module>
    igfold.fold(
  File "/home/ubuntu/user_space/miniconda3/envs/igfold/lib/python3.11/site-packages/igfold/IgFoldRunner.py", line 106, in fold
    model_out = fold(
                ^^^^^
  File "/home/ubuntu/user_space/miniconda3/envs/igfold/lib/python3.11/site-packages/igfold/utils/folding.py", line 221, in fold
    process_prediction(
  File "/home/ubuntu/user_space/miniconda3/envs/igfold/lib/python3.11/site-packages/igfold/utils/folding.py", line 139, in process_prediction
    renumber_pdb(
  File "/home/ubuntu/user_space/miniconda3/envs/igfold/lib/python3.11/site-packages/igfold/utils/abnumber_.py", line 63, in renumber_pdb
    assert len(chain_res) == len(numbering)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError

sequences = {
    "H": "EVQLVESGGGLVQPGGSLRLSCAASDFSLTTYGVHWVRQAPGKGLEWLGVIWSGGSTDYNAAFISRLTISKDNSKNTVYLQMNSLRAEDTAVYYCARDYGSTYVDAIDYWGQGTLVTVSSASTKGPSVFPLAPSSKSTSGGTAALGCLVKDYFPEPVTVSWNSGALTSGVHTFPAVLQSSGLYSLSSVVTVPSSSLGTQTYICNVNHKPSNTKVDKKVEPKSCDKTHTCPPCPAPELLGGPSVFLFPPKPKDTLMISRTPEVTCVVVDVSHEDPEVKFNWYVDGVEVHNAKTKPREEQYGSTYRVVSVLTVLHQDWLNGKEYKCKVSNKALPAPIEKTISKAKGQPREPQVYTLPPSREEMTKNQVSLSCAVKGFYPSDIAVEWESNGQPENNYKTTPPVLDSDGSFFLVSKLTVDKSRWQQGNVFSCSVMHEALHNHYTQKSLSLSPGK",
    "L": "DIVLTQSPDSLAVSLGERATINCRASESVESYGNRYMTWYQQKPGQPPKLLIYRAANLQSGVPDRFSGSGSGTDFTLTISSLQAEDVAVYYCQQSNEDPWTFGQGTKVEIKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVDNALQSGNSQESVTEQDSKDSTYSLSSTLTLSKADYEKHKVYACEVTHQGLSSPVTKSFNRGEC"
}
jeffreyruffolo commented 1 year ago

Hello, it looks like your sequences contain many extra residues beyond the variable domain (Fv). IgFold isn’t trained to predict this much of the structure, but if you truncate to just the variable domain it should work.