In today’s computing world, graphs have become increasingly important in modeling sophisticated structures, entities and their interactions, with broad applications including Bioinformatics, Computer Vision, Web analysis etc. For an example, multiple gene expressions samples over the same set of genes are recorded to strengthen the evidence of co-expression patterns. These can be modeled by forming a set of graphs for these samples. The problem is how to dig into such multiple sources of information to make better inferences. In this paper, I have presented an efficient method to find useful subnetworks from graph networks. The idea is to create a summary graph from these networks and then find subnetworks using this graph. I have given a detailed comparison between an already existing approach called vertex-vertex summary graph approach and the approach discussed in this paper. The results I have found are more promising than for the existing approach.