Issue with alphafold=True in Preprocessing for AF2-predicted structures

[chenanyangming]

Hello,

Thank you for sharing this excellent tool! I’m trying to use ANTIPASTI to predict affinity for antigen-antibody complex structures predicted by AlphaFold2 (AF2).

While following the [Tutorial] Predicting affinity using ANTIPASTI.ipynb, I noticed that when setting alphafold=True in the Preprocessing:

preprocessed_data = Preprocessing(..., alphafold=True)

the generated list_of_residuesfiles appear incorrect. For example, inspecting the provided notebooks/test_data/list_of_residues folder shows seemingly erroneous files like 1_af.npy.

print(np.load('../notebooks/test_data/list_of_residues/1_af.npy', allow_pickle=True))
['START-Ab']

Could you please clarify:

• Whether this is a known issue when processing AF2 predictions?
• If modifications are needed in the Preprocessingclass code to ensure correct output when alphafold=True?

For context, my input PDBs are AF2-predicted complexes . Any guidance on proper preprocessing for such cases would be greatly appreciated!

Best regards!