I am using abintio protocol to predict a cytochrome structure. After I generated the model, it did not give HEME locations. The pdb output only contains the regular protein residues, but missing the location of hemes. I am wondering is there anyway to tell Rosetta also predict the hemes?
Unfortunately, there isn't. Rosetta's ab initio protocol doesn't have a general method for placing ligands or covalently-attached non-canonical modifications. You might be able to script a specialized protocol for this in RosettaScripts or PyRosetta.
Is there a particular reason you want to use ab initio? If you have homolog structural information (and I'm guessing with cytochromes there's plenty of structurally related proteins you can use as templates) you'll get much better results with less computational time if you use a homology modeling approach.
The RosettaCM homology modeling protocol should be able to handle ligands (particularly hemes).