This paper describes an application of e-graphs to a problem from software testing. Given a large and complex input for a program, we use a combination of term refinement and equality rules applied to the input to distill out a much smaller test suite. This smaller test suite is intended to capture all necessary features of the large, complex input for testing the program of interest. In this work we are considering testing of program analysis and transformation tools, where the test suite is intended to capture the set of language features that a tool must support to handle the large scale programs it is intended to operate upon. We describe our e-graph implementation that is based on a JSON term representation and supports term deletion, rule application for establishing equivalences as well as instantiating new terms based on pattern matches, and extraction of terms after refinement. We will present the preliminary results of this ongoing research project at the workshop.