Dense sub-graphs of Protein-Protein Interaction (PPI) graphs are assumed to be potential functional modules and play an important role in inferring the functional behavior of proteins. PPI graphs are known to be scale free, and this property makes the process of isolation of dense sub-graphs very hard. This paper describes a new graph theoretic clustering algorithm that detects densely connected regions in large PPI graph. The method is based on finding bounded diameter sub-graphs around a seed node. The algorithm has the advantage over other graph clustering methods of being very simple and efficient.