error: "local variable 'encode' referenced before assignment"

[LedaKatopodi]

hi~

running deepimmuno-cnn on mode "multiple" returns this error:

UnboundLocalError: local variable 'encode' referenced before assignment

which can be traced back to line 156 of deepimmuno-cnn.py when running peptide_data_aaindex - specifically when running:

encode_pep = peptide_data_aaindex(peptide,after_pca)

replacing the function peptide_data_aaindex in place solved the issue for me - so instead of running the line above I run instead:

# encode_pep = peptide_data_aaindex(peptide,after_pca)    # [10,12]
    length = len(peptide)
    if length == 10:
        encode = aaindex(peptide,after_pca)
    elif length == 9:
        peptide = peptide[:5] + '-' + peptide[5:]
        encode = aaindex(peptide,after_pca)
    encode = encode.reshape(encode.shape[0], encode.shape[1], -1)
    encode_pep = encode

not sure whether anyone else can replicate this error, but am raising the issue in case this is actually a bug.

thanks for building this tool!

cheers,
Leda

[frankligy]

Hi @LedaKatopodi,

Thanks for bringing this up and sorry for the inconvenience you were experiencing.

The UnboundLocalError usually comes from the fact that there are peptides not of length 9 or 10, because as you can see from the function peptide_data_aaindex, if the length doesn't fit into 9 or 10, no encode will be defined, so when later executing reshape function, it will throw an error.

But you said replacing the code solves your problem, which I can not intuit an explanation off hand, because if there are peptides of other lengths, the same issue should persist. But Maybe I missed something here.

Would you mind first checking the peptide length in your input? And happy to help further!

Best,
Frank

[LedaKatopodi]

hi @frankligy~

I also thought it might be the length but for that specific batch I only loaded 9aa-long antigens.

thanks for looking into this

best,
Leda