This document summarizes techniques for representing and analyzing large graphs in the GMine graph analysis system. It discusses how traditional graph representation methods like adjacency matrices and lists do not scale well for large graphs with millions of nodes and edges. The GMine system uses a graph-tree hierarchy and center-piece subgraphs to summarize large graphs and enable interactive visualization and mining. It provides an algorithm for extracting important subgraphs based on node importance scores to aid in analysis and exploration of large real-world graphs.