“…Therefore, accurate identification of the protein-ion-binding sites is important for understanding the mechanism of protein function and for new drug discovery. Many computational methods have been proposed in the last two decades for predicting general ligand-protein binding sites, which can be roughly grouped into two categories of sequencebased (Capra and Singh, 2007;Chen et al, 2014Chen et al, , 2016Magliery and Regan, 2005;Rausell et al, 2010) and structure-based (Brylinski and Skolnick, 2008;Capra et al, 2009;Hendlich et al, 1997;Laskowski, 1995;Roche et al, 2011;Roy and Zhang, 2012;Wass et al, 2010;Yang et al, 2013b) approaches. The sequence-based methods mostly rely on residue conservation analyses under the assumption that ligand-binding residues are functionally important and therefore should be conserved in the evolution.…”