All of the protein structure prediction servers (that I am familiar with) do not appear to work for metalloproteins in general, e.g. I-Tasser. Since metalloproteins are half of all proteins this is a major limitation. The protein I am trying to model has a known Zn finger.