I will give an introduction to the long-standing O-ring theory and a refined version called double water exclusion for modelling protein binding sites.The O-ring hypothesis stated that a binding hotspot is usually surrounded by a ring of residues which have insignificant contribution to the binding free energy.Our double water exclusion hypothesis says that the binding hotspot is the cluster of biclique residues that are not very much exposed to bulky water.In this talk, I also describe an extended version, called tripartite model, for the computational implementation of protein binding hotspots.This protein-water-protein tripartite model is novel as immobilized water molecules are included in this graph model.I will illustrate how this model is used to predict protein binding hotspots in binding interfaces, and why the mutation of the 2009 H1N1 virus did not cause a bigger pandemic disaster than the Spanish Flu pandemic in 1918.