Here is a link to a paper that might be helpful:

http://pubs.acs.org/doi/abs/10.1021/ic401037x

You can read the attached paper where they have developed a method that uses sequence and structural information to predict the residues in metal ion–binding sites.

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0039252

Metal ion is crucial for formation of ICODs (Internal Carbon optimised domains) in the protein that lead to stability. If one develops a method based on carbon domains that would be appropriate for prediction of metal ion binding sites.