Ver2.2.0 ValueError: The number of positions must match the number of atoms

[zjq1011]

Hello. I used ver. 2.2.0 and I still met the error in the title, the same as #246. It failed after producing one unrelaxed model with some clashes in it. Does anyone have any idea about that?

[Aroids]

I too am experiencing this issue, but while using Alphafold (multimer) version 2.2.2 to predict a homohexamer (~3500 resi total) on HPC.

I've experienced it with three of the four proteins (homologous, ~40 seqid) I've tried to predict so, including one with a nearly complete PDB structure (which I verified Alphafold does use as a template). In one case rerunning the job was enough to 'resolve' the issue, but in all other cases the error recurs.

For my purposes relaxation is necessary so I'm hesitant to turn if off.