The target in its original stoichiometry (A4B8C8) was too large to model within the server window, so a reduced model (A1B2C2) was generated instead.
The relaxation step with Amber failed for these models, so the unrelaxed models were released instead. NB: because of a bug in AF2, unrelaxed models start from chain B instead of chain A
Another set of models (A1B1C2) was erroneusly generated, which are also included as a separate archive
"models.tar.gz" will point to the set that has been submitted to the prediction center (A1B2C2)