Protein　classification　is　a　well　established　research　field　concerned　with　the　discovery　of　molecule＇s　properties　through　informational　techniques.　Graph-based　kernels　provide　a　nice　framework　combining　machine　learning　techniques　with　graph　theory.　In　this　paper　we　introduce　a　novel　graph　kernel　method　for　annotating　functional　residues　in　protein　structures.　A　structure¬　is　first　modeled　as　a　protein　contact　graph,　where　nodes　correspond　to　residues　and　edges　connect　spatially　neighboring　residues.　In　experiments　on　classification　of　graph　models　of　proteins,　the　method　based　on　Weisfeiler　Lehman　shortest　path　kernel　with　complement　graphs　outperformed　other　state-of-art　methods.