ValueError: operands could not be broadcast together with shapes (19,19) (2861,2861)

[AlvinLeopold]

Hi there, Thanks for your guys develop such powerful tool.

Actually, I have some models predicted by colabfold, to make sure the data consistency, I decided use my own model on Umol. But when I tried to upload my own Model predicted by colabfold, I recieve the error in this step, btw, I could foud the "_pred_raw.pdb" file were generated, but "generate_best_conformer" seems like could not handle it. `--------------------------------------------------------------------------- ValueError Traceback (most recent call last) in <cell line: 9>() 107 #Get a conformer 108 pred_ligand = read_pdb(RAW_PDB) --> 109 best_conf, best_conf_pos, best_conf_err, atoms, nonH_inds, mol, best_conf_id = generate_best_conformer(pred_ligand['chain_coords'], LIGAND) 110 111 #Align it to the prediction

/content/Umol/src/relax/align_ligand_conformer_colab.py in generate_best_conformer(pred_coords, ligand_smiles) 102 nonH_pos = pos[nonH_inds] 103 conf_dmat = np.sqrt(1e-10 + np.sum((nonH_pos[:,None]-nonH_pos[None,:])2,axis=-1)) --> 104 err = np.mean(np.sqrt(1e-10 + (conf_dmat-pred_dmat)2)) 105 conf_errs.append(err) 106

ValueError: operands could not be broadcast together with shapes (19,19) (2861,2861)`

Any advice would be helpful, Best regard !

[patrickbryant1]

Hi,

It appears you have some big shape mismatches here. I am not sure why this, but suggest making sure you are using the same protein and ligands at all steps.

[AlvinLeopold]

Thanks for your replied!

Now, I can predict them on colab notebook. But I still have some problem on local.

Mainly about the process of "convert id_esmfold.pdb to id_pred_raw.pdb"

The script seems like dont need input pdb structure file, does it can predict automatically? Or did I missed someting?

[patrickbryant1]

Hi, great. No PDB structure is needed as input - only your protein sequence and ligand SMILES string. See the example. Hope this helps!